{ "owner": "metehan.ai", "rights": "Copyright 2026 metehan.ai. All rights reserved. Quote with attribution and a link to the source URL.", "count": 104, "unique_claims": 103, "topics": { "crawlers-and-indexing": 14, "ai-visibility-measurement": 9, "retrieval-and-ranking": 36, "chatgpt-search": 18, "llm-internals": 9, "google-ai-mode-and-overviews": 8, "seo-tools-and-automation": 3, "google-discover": 3, "perplexity": 3, "geo-aeo-method": 1 }, "method": "Facts come from research posts only. Each was extracted with a verbatim evidence quote, then scored by a TypeSafe Jev decision model (supported by source, standalone, original, value, kind). Only findings and mechanisms above the quality gate are kept. quality is 0 to 1.", "usage": "Filter with ?topic=, ?q=, ?limit=. Every fact has a verbatim evidence quote from its source URL. Facts sharing a canonical_fact_id report the same claim from different sources; the canonical one lists the others in also_reported_in.", "facts": [ { "id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#1", "fact": "Over one 30-day period, StateGlobe received more than 600,000 bot requests, including around 72,000 from ChatGPT-User.", "evidence": "In the last 30 days alone, the site got more than 600,000 requests from bots. Around 72,000 of them came from ChatGPT-User.", "topic": "crawlers-and-indexing", "source": { "title": "Publish stats, get cited: The tactic that put my data all over the web", "url": "https://metehan.ai/blog/publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web/", "published": "2026-10-07" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#1", "published": "2026-10-07", "age_days": 6, "possibly_outdated": false }, { "id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#2", "fact": "Seven months after launch, StateGlobe had links from more than 330 sites, including Shopify, DHL, and Bluehost, according to Ahrefs and Semrush.", "evidence": "According to Ahrefs and Semrush, the site has links from 330+ other sites. That includes well-known brands like Shopify, DHL, and Bluehost.", "topic": "ai-visibility-measurement", "source": { "title": "Publish stats, get cited: The tactic that put my data all over the web", "url": "https://metehan.ai/blog/publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web/", "published": "2026-10-07" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#2", "published": "2026-10-07", "age_days": 6, "possibly_outdated": false }, { "id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#3", "fact": "ChatGPT rewrites prompts into fanout queries, retrieves candidates from search sources and its cache, reranks them, and opens selected pages to ground answers.", "evidence": "When ChatGPT searches, it doesn't just copy Google's top 10. It rewrites your prompt into its own sub-queries (called fanout queries), pulls candidates from its search sources and its own cache, reranks them, and then opens a handful of pages to ground the answer.", "topic": "retrieval-and-ranking", "source": { "title": "Publish stats, get cited: The tactic that put my data all over the web", "url": "https://metehan.ai/blog/publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web/", "published": "2026-10-07" }, "kind": "mechanism", "quality": 0.72, "canonical_fact_id": "publish-stats-get-cited-the-tactic-that-put-my-data-all-over-the-web#3", "published": "2026-10-07", "age_days": 6, "possibly_outdated": false }, { "id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#1", "fact": "In observed ChatGPT UI retrieval data, indexed pages supplied pre-scored passage chunks, while fresh or unindexed pages were split into hundreds of fragments at runtime.", "evidence": "When a page is already indexed, the passage chunks and their neural relevance scores are carried straight into the fetch payload and normalized to 1.0000. Fresh or unindexed pages are split into hundreds of small fragments at runtime.", "topic": "retrieval-and-ranking", "source": { "title": "Inside the ChatGPT Leak: How ChatGPT Ranks Chunks From Fetched Pages [CHUNKING PROOF]", "url": "https://metehan.ai/blog/inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages/", "published": "2026-09-09" }, "kind": "mechanism", "quality": 0.96, "canonical_fact_id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#1", "published": "2026-09-09", "age_days": 34, "possibly_outdated": false }, { "id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#2", "fact": "Across 76 pre-indexed pages fetched by ChatGPT, chunk count and order matched between search results and fetched-page records on every page.", "evidence": "We checked all 76. Chunk count and order match on every page.", "topic": "chatgpt-search", "source": { "title": "Inside the ChatGPT Leak: How ChatGPT Ranks Chunks From Fetched Pages [CHUNKING PROOF]", "url": "https://metehan.ai/blog/inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages/", "published": "2026-09-09" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#2", "published": "2026-09-09", "age_days": 34, "possibly_outdated": false }, { "id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#3", "fact": "For 68 ChatGPT web, PDF, Reddit, YouTube, and Wikipedia results, the top chunk normalized to 1.0; eight news results normalized to 0.5.", "evidence": "The top chunk is exactly 1.0 for web, PDF, Reddit, YouTube and Wikipedia results (68 pages) and exactly 0.5 for news results (8 pages).", "topic": "retrieval-and-ranking", "source": { "title": "Inside the ChatGPT Leak: How ChatGPT Ranks Chunks From Fetched Pages [CHUNKING PROOF]", "url": "https://metehan.ai/blog/inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages/", "published": "2026-09-09" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "inside-chatgpt-leak-how-ranks-chunks-from-fetched-pages#3", "published": "2026-09-09", "age_days": 34, "possibly_outdated": false }, { "id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#1", "fact": "In Peec's ChatGPT data, site: queries rose from under 0.11% during mid-May through late July 2026 to 26.2% in the week of August 10.", "evidence": "In Peec's ChatGPT query data, site: went from almost unused (under 0.11 percent of weekly queries from mid May to late July 2026) to 10.2 percent in the week of August 3, then 26.2 percent in the week of August 10", "topic": "chatgpt-search", "source": { "title": "Want a real short-term AI visibility hack? Or in a bad actor's hands, becomes a real blackhat GEO hack?", "url": "https://metehan.ai/blog/want-real-short-term-ai-visibility-hack-bad-actors-hands/", "published": "2026-08-28" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#2", "published": "2026-08-28", "age_days": 46, "possibly_outdated": false }, { "id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#2", "fact": "A WHOIS check of ChatGPT site: fanouts found four registerable domains appearing at least 100 times and 104 domains appearing 10 to 99 times.", "evidence": "In our WHOIS pass, 4 registerable domains showed up 100 or more times, and 104 showed up between 10 and 99 times. The rest of the checked set was already taken.", "topic": "chatgpt-search", "source": { "title": "Want a real short-term AI visibility hack? Or in a bad actor's hands, becomes a real blackhat GEO hack?", "url": "https://metehan.ai/blog/want-real-short-term-ai-visibility-hack-bad-actors-hands/", "published": "2026-08-28" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#2", "published": "2026-08-28", "age_days": 46, "possibly_outdated": false }, { "id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#3", "fact": "ChatGPT sometimes generated an unregistered hostname for a site: fanout while citations still pointed to the real official domain.", "evidence": "Citations still go to branddomain.com. The hidden fanout asks for site: brand-domain.com, or a country or .org variant of the same name. That guessed host was never registered.", "topic": "retrieval-and-ranking", "source": { "title": "Want a real short-term AI visibility hack? Or in a bad actor's hands, becomes a real blackhat GEO hack?", "url": "https://metehan.ai/blog/want-real-short-term-ai-visibility-hack-bad-actors-hands/", "published": "2026-08-28" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#3", "published": "2026-08-28", "age_days": 46, "possibly_outdated": false }, { "id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#2", "fact": "In the week of August 10, 26.2 percent of observed ChatGPT queries used a site: parameter, representing 3.9 million of 14.9 million queries.", "evidence": "In the week of August 10 it hit 26.2 percent. That week, 3.9 million of the 14.9 million ChatGPT queries we observed carried a site: parameter.", "topic": "chatgpt-search", "source": { "title": "Reddit lost most of its visibility in ChatGPT, and the site parameter filled the gap", "url": "https://metehan.ai/blog/reddit-lost-most-its-visibility-chatgpt-site-filled-gap/", "published": "2026-08-21" }, "kind": "finding", "quality": 0.97, "canonical_fact_id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#2", "also_reported_in": [ { "id": "want-real-short-term-ai-visibility-hack-bad-actors-hands#1", "url": "https://metehan.ai/blog/want-real-short-term-ai-visibility-hack-bad-actors-hands/" } ], "published": "2026-08-21", "age_days": 53, "possibly_outdated": false }, { "id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#1", "fact": "Reddit's share of ChatGPT citations fell from 4.6 percent in early August to 0.23 percent by August 21, a 95 percent drop.", "evidence": "By August 21 it sat at 0.23 percent, a 95 percent drop from its early August level.", "topic": "chatgpt-search", "source": { "title": "Reddit lost most of its visibility in ChatGPT, and the site parameter filled the gap", "url": "https://metehan.ai/blog/reddit-lost-most-its-visibility-chatgpt-site-filled-gap/", "published": "2026-08-21" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#1", "published": "2026-08-21", "age_days": 53, "possibly_outdated": false }, { "id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#3", "fact": "ChatGPT's site: parameter searches overwhelmingly targeted official sources, while fewer than 1 percent pointed to reddit.com.", "evidence": "In our tracked fanouts, those targets are overwhelmingly official sources: brand websites, product documentation, help centers, and review platforms. Reddit is almost never the target. Fewer than 1 percent of the site: parameter queries we analyzed point at reddit.com.", "topic": "retrieval-and-ranking", "source": { "title": "Reddit lost most of its visibility in ChatGPT, and the site parameter filled the gap", "url": "https://metehan.ai/blog/reddit-lost-most-its-visibility-chatgpt-site-filled-gap/", "published": "2026-08-21" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "reddit-lost-most-its-visibility-chatgpt-site-filled-gap#3", "published": "2026-08-21", "age_days": 53, "possibly_outdated": false }, { "id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#3", "fact": "BGE-reranker-large placed roughly 24 percent of its attention on punctuation and special tokens for a typical passage.", "evidence": "In our measurements, BGE-reranker-large puts roughly 24 percent of its attention on punctuation and special tokens for a typical passage.", "topic": "retrieval-and-ranking", "source": { "title": "Rerankers for GEO/AEO: How AI search chooses passages and sources", "url": "https://metehan.ai/blog/rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources/", "published": "2026-08-06" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#3", "published": "2026-08-06", "age_days": 68, "possibly_outdated": false }, { "id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#1", "fact": "OpenAI File Search defaults to 800-token chunks with 400-token overlap and combines semantic and keyword search before optional reranking.", "evidence": "Documents are chunked (default 800 tokens per chunk with 400 token overlap), embedded with text-embedding-3-large, and at query time the system runs both a semantic (vector) search and a keyword search, then can apply a reranker to the combined candidates.", "topic": "retrieval-and-ranking", "source": { "title": "Rerankers for GEO/AEO: How AI search chooses passages and sources", "url": "https://metehan.ai/blog/rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources/", "published": "2026-08-06" }, "kind": "mechanism", "quality": 0.71, "canonical_fact_id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#1", "published": "2026-08-06", "age_days": 68, "possibly_outdated": false }, { "id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#2", "fact": "ModernBERT supports an 8,192-token context window, 16 times BERT's original context length.", "evidence": "ModernBERT brought encoders up to date: an 8,192 token context window (16 times the original BERT), modern positional embeddings, and much faster inference.", "topic": "llm-internals", "source": { "title": "Rerankers for GEO/AEO: How AI search chooses passages and sources", "url": "https://metehan.ai/blog/rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources/", "published": "2026-08-06" }, "kind": "finding", "quality": 0.7, "canonical_fact_id": "rerankers-for-geo-aeo-how-ai-search-chooses-passages-and-sources#2", "published": "2026-08-06", "age_days": 68, "possibly_outdated": false }, { "id": "reddit-machine-translated-pages-ai-visibility#2", "fact": "ChatGPT's share of Reddit citations pointing to translated pages dropped from 6.14% in April 2026 to 0.66% in May 2026.", "evidence": "ChatGPT's share of Reddit citations pointing to translated pages dropped from 6.14% in April to 0.66% in May.", "topic": "chatgpt-search", "source": { "title": "Translated Reddit Is Winning AI Citations.", "url": "https://metehan.ai/blog/reddit-machine-translated-pages-ai-visibility/", "published": "2026-06-22" }, "kind": "finding", "quality": 0.97, "canonical_fact_id": "reddit-machine-translated-pages-ai-visibility#2", "published": "2026-06-22", "age_days": 113, "possibly_outdated": false }, { "id": "reddit-machine-translated-pages-ai-visibility#3", "fact": "Google AI Overview's translated Reddit citation share reached 52% in Germany and exceeded 70% in Sweden and Norway.", "evidence": "In Germany, 52% of the Reddit threads Google AI Overview cited were translated versions: more than half of every Reddit citation it showed to German users. In Sweden and Norway, it’s over 70%.", "topic": "google-ai-mode-and-overviews", "source": { "title": "Translated Reddit Is Winning AI Citations.", "url": "https://metehan.ai/blog/reddit-machine-translated-pages-ai-visibility/", "published": "2026-06-22" }, "kind": "finding", "quality": 0.97, "canonical_fact_id": "reddit-machine-translated-pages-ai-visibility#3", "published": "2026-06-22", "age_days": 113, "possibly_outdated": false }, { "id": "reddit-machine-translated-pages-ai-visibility#1", "fact": "Across ChatGPT, Google AI Overview, Google AI Mode, and Gemini, 3,699,577 Reddit citations carried the translation parameter from March 1 to June 10, 2026.", "evidence": "We analyzed 64M+ Reddit citations between March 1 and June 10, 2026, across 20 different countries and four AI sources: ChatGPT, Google AI Overview, Google AI Mode, and Gemini. 3,699,577 Reddit citations carried ?tl=.", "topic": "ai-visibility-measurement", "source": { "title": "Translated Reddit Is Winning AI Citations.", "url": "https://metehan.ai/blog/reddit-machine-translated-pages-ai-visibility/", "published": "2026-06-22" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "reddit-machine-translated-pages-ai-visibility#1", "published": "2026-06-22", "age_days": 113, "possibly_outdated": false }, { "id": "http-headers-internal-links#1", "fact": "A Rust crawler processed 65,000 pages from response headers in 99 seconds without parsing HTML.", "evidence": "Then I crawled 65,000 pages in 99 seconds without parsing one byte of HTML.", "topic": "crawlers-and-indexing", "source": { "title": "I Crawled 65,000 Pages of My Own Site Without Parsing a Single Line of HTML", "url": "https://metehan.ai/blog/http-headers-internal-links/", "published": "2026-05-04" }, "kind": "finding", "quality": 0.94, "canonical_fact_id": "http-headers-internal-links#1", "published": "2026-05-04", "age_days": 162, "possibly_outdated": false }, { "id": "http-headers-internal-links#2", "fact": "Analysis of data.stateglobe.com's header-derived link graph found 27,372 pages with no inbound internal links and 27,659 pages unreachable from the homepage.", "evidence": "27,372 pages (41.9%) had zero inbound internal links. Pure orphans, only reachable through the sitemap. On a site that was supposed to be tightly cross-linked.\n27,659 pages were unreachable by walking from /.", "topic": "seo-tools-and-automation", "source": { "title": "I Crawled 65,000 Pages of My Own Site Without Parsing a Single Line of HTML", "url": "https://metehan.ai/blog/http-headers-internal-links/", "published": "2026-05-04" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "http-headers-internal-links#2", "published": "2026-05-04", "age_days": 162, "possibly_outdated": false }, { "id": "http-headers-internal-links#3", "fact": "On data.stateglobe.com, about 17 KB of combined custom headers exceeded a response-header limit and caused hub pages to return HTTP 500 errors.", "evidence": "X-Internal-Links alone was around 10.8 KB. Add X-Headings at ~6.9 KB on top, and the combined response headers on those hub pages crossed ~17 KB. That breached the response-header size limit on the path between origin and edge", "topic": "crawlers-and-indexing", "source": { "title": "I Crawled 65,000 Pages of My Own Site Without Parsing a Single Line of HTML", "url": "https://metehan.ai/blog/http-headers-internal-links/", "published": "2026-05-04" }, "kind": "finding", "quality": 0.85, "canonical_fact_id": "http-headers-internal-links#3", "published": "2026-05-04", "age_days": 162, "possibly_outdated": false }, { "id": "how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool#2", "fact": "NeuralSEO converts text to TTS audio, derives word-level timestamps with faster-whisper, and then uses TRIBE v2 to predict fMRI responses.", "evidence": "Inference: Text goes through TTS audio, then word-level timestamps via faster-whisper, then TRIBE v2 fMRI prediction", "topic": "seo-tools-and-automation", "source": { "title": "How I Turned Meta AI's Brain Scanner Model Into a Free SEO Tool", "url": "https://metehan.ai/blog/how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool/", "published": "2026-03-30" }, "kind": "mechanism", "quality": 0.76, "canonical_fact_id": "how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool#2", "published": "2026-03-30", "age_days": 197, "possibly_outdated": false }, { "id": "how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool#1", "fact": "Meta AI's TRIBE v2 was trained on fMRI recordings from more than 700 volunteers and predicts cortical responses across roughly 20,000 vertices.", "evidence": "It was trained on fMRI recordings, actual brain scans, collected while 700+ volunteers watched videos, listened to audio, and read text. It's a multilanguage model.\nThe model learned to predict how the human cortex responds to any input across roughly 20,000 cortical vertices.", "topic": "llm-internals", "source": { "title": "How I Turned Meta AI's Brain Scanner Model Into a Free SEO Tool", "url": "https://metehan.ai/blog/how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool/", "published": "2026-03-30" }, "kind": "finding", "quality": 0.66, "canonical_fact_id": "how-i-turned-meta-ais-brain-scanner-model-into-a-free-seo-tool#1", "published": "2026-03-30", "age_days": 197, "possibly_outdated": false }, { "id": "i-turned-16-months-of-google-search-console-data-into-a-vector-database-heres-what-i-learned#2", "fact": "A vector database can retrieve a 200-impression query yet miss a 50,000-impression query because embedding similarity does not guarantee numerical correctness.", "evidence": "The vector DB might return a query with 200 impressions because its text happened to be close in embedding space. It might miss a query with 50,000 impressions because the text representation didn't match.", "topic": "retrieval-and-ranking", "source": { "title": "I Turned 16 Months of Google Search Console Data Into a Vector Database.", "url": "https://metehan.ai/blog/i-turned-16-months-of-google-search-console-data-into-a-vector-database-heres-what-i-learned/", "published": "2026-03-18" }, "kind": "mechanism", "quality": 0.85, "canonical_fact_id": "i-turned-16-months-of-google-search-console-data-into-a-vector-database-heres-what-i-learned#2", "published": "2026-03-18", "age_days": 209, "possibly_outdated": false }, { "id": "i-built-a-60-000-page-ai-website-for-10-gptbot-crawled-it-30-000-times-in-12-hours#1", "fact": "GPTBot made more than 29,000 requests to StateGlobe in the first 12 hours, crawling at roughly one request per second.", "evidence": "First 12 Hours\n29,000+ requests from GPTBot alone\nGPTBot was hitting the site at roughly 1 request per second, systematically crawling through pages", "topic": "crawlers-and-indexing", "source": { "title": "I Built a 60,000-Page AI Website for $10: GPTBot Crawled It 30,000+ Times in 12 Hours", "url": "https://metehan.ai/blog/i-built-a-60-000-page-ai-website-for-10-gptbot-crawled-it-30-000-times-in-12-hours/", "published": "2026-03-04" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "i-built-a-60-000-page-ai-website-for-10-gptbot-crawled-it-30-000-times-in-12-hours#1", "published": "2026-03-04", "age_days": 223, "possibly_outdated": false }, { "id": "google-discover-architecture#1", "fact": "Google Discover's client-side parser checks Schema.org JSON-LD before Open Graph, Twitter Card, and generic HTML meta tags.", "evidence": "Structured Data & Open Graph Parsing: The client-side parser extracts page metadata with a defined priority order: Schema.org JSON-LD is checked first, followed by Open Graph tags, then Twitter Card tags, then generic HTML meta tags.", "topic": "google-discover", "source": { "title": "Google Discover Architecture: Clusters, Classifiers, OG Tags, NAIADES - What SDK Telemetry Reveals", "url": "https://metehan.ai/blog/google-discover-architecture/", "published": "2026-02-24" }, "kind": "mechanism", "quality": 0.95, "canonical_fact_id": "google-discover-architecture#1", "published": "2026-02-24", "age_days": 231, "possibly_outdated": false }, { "id": "google-discover-architecture#2", "fact": "Google Discover tracks publisher or domain filtering separately from single-URL filtering.", "evidence": "The system operates two separate filter levels: filter_collection_status (publisher/domain level) and filter_entity_status (single URL level).", "topic": "google-discover", "source": { "title": "Google Discover Architecture: Clusters, Classifiers, OG Tags, NAIADES - What SDK Telemetry Reveals", "url": "https://metehan.ai/blog/google-discover-architecture/", "published": "2026-02-24" }, "kind": "mechanism", "quality": 0.94, "canonical_fact_id": "google-discover-architecture#2", "published": "2026-02-24", "age_days": 231, "possibly_outdated": false }, { "id": "google-discover-architecture#3", "fact": "Google Discover permanently tombstones dismissed content so it does not resurface.", "evidence": "User dismissals are permanent. Content is tombstoned and does not resurface.", "topic": "google-discover", "source": { "title": "Google Discover Architecture: Clusters, Classifiers, OG Tags, NAIADES - What SDK Telemetry Reveals", "url": "https://metehan.ai/blog/google-discover-architecture/", "published": "2026-02-24" }, "kind": "mechanism", "quality": 0.87, "canonical_fact_id": "google-discover-architecture#3", "published": "2026-02-24", "age_days": 231, "possibly_outdated": false }, { "id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#2", "fact": "In Yesilyurt's identical-text test, Arabic text decreased from 70 tokens with GPT-2 encoding to 21 tokens with o200k.", "evidence": "Arabic text went from 70 tokens to 21, a 3.3x compression improvement.", "topic": "llm-internals", "source": { "title": "Reverse-Engineering the OpenAI's GPT-5 Tokenizer: What 200,000 Tokens Reveal About AEO/GEO", "url": "https://metehan.ai/blog/reverse-engineering-the-gpt-5-tokenizer-aeo-geo/", "published": "2026-02-13" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#2", "published": "2026-02-13", "age_days": 242, "possibly_outdated": false }, { "id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#3", "fact": "The o200k regex splits ChatGPT into Chat and GPT before BPE processing begins.", "evidence": "This means ChatGPT is split into Chat + GPT at the regex level, before BPE even runs.", "topic": "llm-internals", "source": { "title": "Reverse-Engineering the OpenAI's GPT-5 Tokenizer: What 200,000 Tokens Reveal About AEO/GEO", "url": "https://metehan.ai/blog/reverse-engineering-the-gpt-5-tokenizer-aeo-geo/", "published": "2026-02-13" }, "kind": "mechanism", "quality": 0.8, "canonical_fact_id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#3", "published": "2026-02-13", "age_days": 242, "possibly_outdated": false }, { "id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#1", "fact": "The o200k_base vocabulary contains 200,000 tokens, including 199,998 mergeable tokens and two special tokens.", "evidence": "Total tokens 200,000 (199,998 mergeable + 2 special)", "topic": "llm-internals", "source": { "title": "Reverse-Engineering the OpenAI's GPT-5 Tokenizer: What 200,000 Tokens Reveal About AEO/GEO", "url": "https://metehan.ai/blog/reverse-engineering-the-gpt-5-tokenizer-aeo-geo/", "published": "2026-02-13" }, "kind": "finding", "quality": 0.75, "canonical_fact_id": "reverse-engineering-the-gpt-5-tokenizer-aeo-geo#1", "published": "2026-02-13", "age_days": 242, "possibly_outdated": false }, { "id": "cc-rank#3", "fact": "A 680-million-citation analysis found Wikipedia led ChatGPT at 7.8%, while Reddit led Perplexity at 6.6% and Google AI Overviews at 2.2%.", "evidence": "Profound (August 2024 - June 2025) - 680 million citations analyzed:\n \n \n Platform Top Source Share of Total Citations \n \n \n ChatGPT Wikipedia 7.8% \n Perplexity Reddit 6.6% \n Google AI Overviews Reddit 2.2%", "topic": "ai-visibility-measurement", "source": { "title": "The Hidden Authority Signal: Why Your CC Rank May Matter More for AI Visibility", "url": "https://metehan.ai/blog/cc-rank/", "published": "2026-01-07" }, "kind": "finding", "quality": 0.73, "canonical_fact_id": "cc-rank#3", "published": "2026-01-07", "age_days": 279, "possibly_outdated": false }, { "id": "cc-rank#1", "fact": "Among 47 LLMs analyzed from 2019 to 2023, 30 models, or 64%, used at least one filtered version of Common Crawl.", "evidence": "64% of 47 LLMs analyzed (2019-2023) used at least one filtered version of Common Crawl (30 out of 47 models)", "topic": "llm-internals", "source": { "title": "The Hidden Authority Signal: Why Your CC Rank May Matter More for AI Visibility", "url": "https://metehan.ai/blog/cc-rank/", "published": "2026-01-07" }, "kind": "finding", "quality": 0.71, "canonical_fact_id": "cc-rank#1", "published": "2026-01-07", "age_days": 279, "possibly_outdated": false }, { "id": "cc-rank#2", "fact": "Common Crawl uses Harmonic Centrality to prioritize crawling, with higher-scoring domains crawled more frequently.", "evidence": "CC uses Harmonic Centrality to determine crawl priority, domains with higher scores are crawled more frequently", "topic": "crawlers-and-indexing", "source": { "title": "The Hidden Authority Signal: Why Your CC Rank May Matter More for AI Visibility", "url": "https://metehan.ai/blog/cc-rank/", "published": "2026-01-07" }, "kind": "mechanism", "quality": 0.7, "canonical_fact_id": "cc-rank#2", "published": "2026-01-07", "age_days": 279, "possibly_outdated": false }, { "id": "chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics#2", "fact": "ChatGPT can use summarized turn messages from previous chats to personalize responses.", "evidence": "The thing is, ChatGPT is returning with \"summarized\" turn messages from previous chats.", "topic": "llm-internals", "source": { "title": "ChatGPT's Personal Memory Affects Everything In Responses & OpenAI is Using Google Analytics", "url": "https://metehan.ai/blog/chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics/", "published": "2025-12-20" }, "kind": "mechanism", "quality": 0.85, "canonical_fact_id": "chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics#2", "published": "2025-12-20", "age_days": 297, "possibly_outdated": false }, { "id": "chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics#3", "fact": "Personalization creates a blind spot when tracking prompts to measure AI visibility in ChatGPT.", "evidence": "So the amplification of personalization is heavy and when it comes to track prompts for AI visibility actually creates a huge blind spot.", "topic": "ai-visibility-measurement", "source": { "title": "ChatGPT's Personal Memory Affects Everything In Responses & OpenAI is Using Google Analytics", "url": "https://metehan.ai/blog/chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics/", "published": "2025-12-20" }, "kind": "mechanism", "quality": 0.62, "canonical_fact_id": "chatgpts-personal-memory-affects-everything-in-responses-openai-is-using-google-analytics#3", "published": "2025-12-20", "age_days": 297, "possibly_outdated": false }, { "id": "how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias#1", "fact": "Metehan Yesilyurt reported that an article about LLM recency bias generated 125,000 Google impressions.", "evidence": "The same recency effect that drove 125K impressions also shapes AI citations in ChatGPT, Perplexity and Gemini.", "topic": "ai-visibility-measurement", "source": { "title": "How I Accidentally Triggered 125K Google Impressions Through LLM \\\"AI News Today Recency\\\" Bias", "url": "https://metehan.ai/blog/how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias/", "published": "2025-12-15" }, "kind": "finding", "quality": 0.86, "canonical_fact_id": "how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias#1", "published": "2025-12-15", "age_days": 302, "possibly_outdated": false }, { "id": "how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias#2", "fact": "On December 16, 2025, Yesilyurt reported that his newly indexed Google post was receiving impressions but had not received clicks.", "evidence": "16 December 2025 Update - It's happening again\nGoogle indexed my post and this is the last 24 hours. Impressions are counting, but no clicks, yet.", "topic": "ai-visibility-measurement", "source": { "title": "How I Accidentally Triggered 125K Google Impressions Through LLM \\\"AI News Today Recency\\\" Bias", "url": "https://metehan.ai/blog/how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias/", "published": "2025-12-15" }, "kind": "finding", "quality": 0.66, "canonical_fact_id": "how-i-accidentally-triggered-125k-google-impressions-through-llm-ai-news-today-recency-bias#2", "published": "2025-12-15", "age_days": 302, "possibly_outdated": false }, { "id": "reverse-engineering-google-ai-mode#2", "fact": "Discovery Engine activates its personalized predicted click-through rate model only after serving more than 100,000 queries.", "evidence": "TIER 3: Personalized PCTR Model Incorporates user-specific signals: user metadata, individual search history, and behavioral patterns.\nOnly activates after 100,000+ queries have been served", "topic": "retrieval-and-ranking", "source": { "title": "Reverse-Engineering Google AI Mode: What Discovery Engine Reveals About Google's AI Search Architecture", "url": "https://metehan.ai/blog/reverse-engineering-google-ai-mode/", "published": "2025-11-26" }, "kind": "mechanism", "quality": 0.8, "canonical_fact_id": "reverse-engineering-google-ai-mode#2", "published": "2025-11-26", "age_days": 321, "possibly_outdated": false }, { "id": "reverse-engineering-google-ai-mode#1", "fact": "Google Discovery Engine caps retrieval chunks at 500 tokens and can include ancestor headings with each chunk.", "evidence": "Chunking Configuration:\nAdvanced chunking can be enabled\nChunk size limit: 500 tokens (maximum, not fixed)\nOption to include ancestor headings in chunks", "topic": "retrieval-and-ranking", "source": { "title": "Reverse-Engineering Google AI Mode: What Discovery Engine Reveals About Google's AI Search Architecture", "url": "https://metehan.ai/blog/reverse-engineering-google-ai-mode/", "published": "2025-11-26" }, "kind": "mechanism", "quality": 0.79, "canonical_fact_id": "reverse-engineering-google-ai-mode#1", "published": "2025-11-26", "age_days": 321, "possibly_outdated": false }, { "id": "how-chatgpt-uses-the-public-suffix-list#1", "fact": "ChatGPT's browser-based source code contains explicit references to Public Suffix List processing for citations and domain handling.", "evidence": "Through technical analysis of ChatGPT's browser-based source code, I discovered explicit references to Public Suffix List processing in their citation and domain handling systems.", "topic": "chatgpt-search", "source": { "title": "How ChatGPT Uses the Public Suffix List to Treat Domains & Subdomains Differently", "url": "https://metehan.ai/blog/how-chatgpt-uses-the-public-suffix-list/", "published": "2025-11-10" }, "kind": "finding", "quality": 0.87, "canonical_fact_id": "how-chatgpt-uses-the-public-suffix-list#1", "published": "2025-11-10", "age_days": 337, "possibly_outdated": false }, { "id": "how-chatgpt-uses-the-public-suffix-list#3", "fact": "The Public Suffix List is infrastructure rather than a ranking factor, defining ownership before algorithms determine what to surface.", "evidence": "The Public Suffix List isn't a ranking factor. It's infrastructure. It sits below the semantic layer, defining who owns what before algorithms decide what to surface.", "topic": "retrieval-and-ranking", "source": { "title": "How ChatGPT Uses the Public Suffix List to Treat Domains & Subdomains Differently", "url": "https://metehan.ai/blog/how-chatgpt-uses-the-public-suffix-list/", "published": "2025-11-10" }, "kind": "mechanism", "quality": 0.62, "canonical_fact_id": "how-chatgpt-uses-the-public-suffix-list#3", "published": "2025-11-10", "age_days": 337, "possibly_outdated": false }, { "id": "100-questions-that-show-aeo-geo-is-different-than-seo#1", "fact": "Metehan Yesilyurt reports finding Reciprocal Rank Fusion in ChatGPT's ranking code.", "evidence": "Having reverse-engineered ChatGPT’s ranking system and discovered RRF in the code, I now have more questions than answers.", "topic": "chatgpt-search", "source": { "title": "100+ Questions That Show AEO/GEO Is Different Than SEO", "url": "https://metehan.ai/blog/100-questions-that-show-aeo-geo-is-different-than-seo/", "published": "2025-10-29" }, "kind": "finding", "quality": 0.83, "canonical_fact_id": "100-questions-that-show-aeo-geo-is-different-than-seo#1", "published": "2025-10-29", "age_days": 349, "possibly_outdated": false }, { "id": "100-questions-that-show-aeo-geo-is-different-than-seo#2", "fact": "The article reports that ChatGPT retrieves 38 to 65 results, while Google indexes trillions of pages.", "evidence": "Google indexes trillions of pages. ChatGPT retrieves 38-65. This isn’t a small difference – it’s a 99.999% reduction", "topic": "retrieval-and-ranking", "source": { "title": "100+ Questions That Show AEO/GEO Is Different Than SEO", "url": "https://metehan.ai/blog/100-questions-that-show-aeo-geo-is-different-than-seo/", "published": "2025-10-29" }, "kind": "finding", "quality": 0.79, "canonical_fact_id": "100-questions-that-show-aeo-geo-is-different-than-seo#2", "published": "2025-10-29", "age_days": 349, "possibly_outdated": false }, { "id": "visual-query-fan-out-screaming-frog#1", "fact": "Google AI Mode's visual search fan-out analyzes subtle image details and secondary objects, then runs multiple queries in the background.", "evidence": "It's a technique where AI Mode performs a comprehensive analysis of an image, recognizing subtle details and secondary objects in addition to the primary subjects, and then runs multiple queries in the background.", "topic": "google-ai-mode-and-overviews", "source": { "title": "[FREE CODE INSIDE] How to Run Visual Query Fan-Out Analysis in Screaming Frog Using OpenAI", "url": "https://metehan.ai/blog/visual-query-fan-out-screaming-frog/", "published": "2025-10-12" }, "kind": "mechanism", "quality": 0.64, "canonical_fact_id": "visual-query-fan-out-screaming-frog#1", "published": "2025-10-12", "age_days": 366, "possibly_outdated": true }, { "id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#1", "fact": "Waseda University researchers tested recency bias in seven AI models by adding artificial publication dates to search results.", "evidence": "A team from Waseda University published a great study testing seven major AI models (GPT-4o, GPT-4, GPT-3.5, LLaMA-3 8B/70B, and Qwen-2.5 7B/72B). They added artificial publication dates to search results and measured what happened.", "topic": "retrieval-and-ranking", "source": { "title": "I Found It in the Code, Science Proved It in the Lab: The Recency Bias That's Reshaping AI Search", "url": "https://metehan.ai/blog/i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search/", "published": "2025-10-03" }, "kind": "finding", "quality": 0.73, "canonical_fact_id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#1", "published": "2025-10-03", "age_days": 375, "possibly_outdated": true }, { "id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#3", "fact": "In the artificial-date reranking test, GPT-3.5-turbo produced a 95-rank jump, while LLaMA3-8B showed a 25.23% preference-reversal rate.", "evidence": "Largest single position jump 61 ranks (Qwen2.5-7B) 95 ranks (GPT-3.5-turbo) \n Preference reversals 8.25% (Qwen2.5-72B) 25.23% (LLaMA3-8B)", "topic": "retrieval-and-ranking", "source": { "title": "I Found It in the Code, Science Proved It in the Lab: The Recency Bias That's Reshaping AI Search", "url": "https://metehan.ai/blog/i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search/", "published": "2025-10-03" }, "kind": "finding", "quality": 0.7, "canonical_fact_id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#3", "published": "2025-10-03", "age_days": 375, "possibly_outdated": true }, { "id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#2", "fact": "Across all tested models and both datasets, ranks 1 through 10 shifted 0.8 to 4.8 years toward fresher content.", "evidence": "Ranks 1-10: +0.8 to +4.8 years fresher (all models, both datasets)", "topic": "retrieval-and-ranking", "source": { "title": "I Found It in the Code, Science Proved It in the Lab: The Recency Bias That's Reshaping AI Search", "url": "https://metehan.ai/blog/i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search/", "published": "2025-10-03" }, "kind": "finding", "quality": 0.68, "canonical_fact_id": "i-found-it-in-the-code-science-proved-it-in-the-lab-the-recency-bias-thats-reshaping-ai-search#2", "published": "2025-10-03", "age_days": 375, "possibly_outdated": true }, { "id": "the-num100-effect-why-deep-research-feels-shrinks#2", "fact": "In the same experiment, Perplexity returned 20 results, down from approximately 52 previously.", "evidence": "Perplexity: 20 results, down from ~52 previously. (They are saying 42 sources in average)", "topic": "perplexity", "source": { "title": "The num=100 Effect: Why Deep Research Feels Shrinks", "url": "https://metehan.ai/blog/the-num100-effect-why-deep-research-feels-shrinks/", "published": "2025-09-16" }, "kind": "finding", "quality": 0.94, "canonical_fact_id": "the-num100-effect-why-deep-research-feels-shrinks#2", "published": "2025-09-16", "age_days": 392, "possibly_outdated": true }, { "id": "the-num100-effect-why-deep-research-feels-shrinks#1", "fact": "In a small deep research experiment, ChatGPT showed 19 sources, compared with more than 40 in an earlier test.", "evidence": "ChatGPT: 19 sources this time, compared to 40+ before.\nThis was from two months ago.", "topic": "chatgpt-search", "source": { "title": "The num=100 Effect: Why Deep Research Feels Shrinks", "url": "https://metehan.ai/blog/the-num100-effect-why-deep-research-feels-shrinks/", "published": "2025-09-16" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "the-num100-effect-why-deep-research-feels-shrinks#1", "published": "2025-09-16", "age_days": 392, "possibly_outdated": true }, { "id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#1", "fact": "In reported tests, ChatGPT retrieved 65 sources for AI news, 52 for tech news, 38 for sports news, and 60 for sneaker brands.", "evidence": "\"latest ai news\" → 65 sources\n\"latest tech news\" → 52 sources\n\"latest sport news\" → 38 sources\n\"trending sneaker brands 2025\" → 60 sources", "topic": "chatgpt-search", "source": { "title": "The RRF Top-n Playbook: How to Get Cited by ChatGPT’s Web Mode (k≈60)", "url": "https://metehan.ai/blog/the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode/", "published": "2025-09-11" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#1", "published": "2025-09-11", "age_days": 397, "possibly_outdated": true }, { "id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#2", "fact": "Reciprocal Rank Fusion gives a document zero score when the document is outside the retrieval window.", "evidence": "RRF (Reciprocal Rank Fusion) only scores what it retrieves. If you're not in the retrieval window, you get zero score. No score means no chance of citations.", "topic": "retrieval-and-ranking", "source": { "title": "The RRF Top-n Playbook: How to Get Cited by ChatGPT’s Web Mode (k≈60)", "url": "https://metehan.ai/blog/the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode/", "published": "2025-09-11" }, "kind": "mechanism", "quality": 0.72, "canonical_fact_id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#2", "published": "2025-09-11", "age_days": 397, "possibly_outdated": true }, { "id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#3", "fact": "In Yesilyurt's RRF model with k=60, two appearances at rank 40 yield a fused score of 0.020.", "evidence": "2× top-40: $latex 2/(60+40)=2/100=0.020$ ✅", "topic": "retrieval-and-ranking", "source": { "title": "The RRF Top-n Playbook: How to Get Cited by ChatGPT’s Web Mode (k≈60)", "url": "https://metehan.ai/blog/the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode/", "published": "2025-09-11" }, "kind": "mechanism", "quality": 0.71, "canonical_fact_id": "the-rrf-top-n-playbook-how-to-get-cited-by-chatgpts-web-mode#3", "published": "2025-09-11", "age_days": 397, "possibly_outdated": true }, { "id": "chatgpt-5-search-configuration#1", "fact": "ChatGPT's retrieval system uses a reranking model named ret-rr-skysight-v3.", "evidence": "At the heart of ChatGPT's retrieval system is a reranking model with the cryptic name ret-rr-skysight-v3.", "topic": "retrieval-and-ranking", "source": { "title": "Inside ChatGPT's GPT 5 Search: What the Configuration Files Reveal About How It Ranks Your Content", "url": "https://metehan.ai/blog/chatgpt-5-search-configuration/", "published": "2025-08-20" }, "kind": "mechanism", "quality": 0.85, "canonical_fact_id": "chatgpt-5-search-configuration#1", "published": "2025-08-20", "age_days": 419, "possibly_outdated": true }, { "id": "chatgpt-5-search-configuration#3", "fact": "ChatGPT's configuration enables query-intent analysis to distinguish requests such as definitions, how-to guides, and comparisons.", "evidence": "With enable_query_intent: true, ChatGPT analyzes what the user is actually trying to accomplish. It's not just matching keywords, it's understanding whether someone wants a definition, a how-to guide, a comparison, or something else entirely.", "topic": "retrieval-and-ranking", "source": { "title": "Inside ChatGPT's GPT 5 Search: What the Configuration Files Reveal About How It Ranks Your Content", "url": "https://metehan.ai/blog/chatgpt-5-search-configuration/", "published": "2025-08-20" }, "kind": "mechanism", "quality": 0.7, "canonical_fact_id": "chatgpt-5-search-configuration#3", "published": "2025-08-20", "age_days": 419, "possibly_outdated": true }, { "id": "chatgpt-5-search-configuration#2", "fact": "ChatGPT's instructions direct the system to use web search for recent events, fresh information, or news.", "evidence": "Use the `web` tool instead when the user asks about recent events / fresh information, or asks about news etc.", "topic": "chatgpt-search", "source": { "title": "Inside ChatGPT's GPT 5 Search: What the Configuration Files Reveal About How It Ranks Your Content", "url": "https://metehan.ai/blog/chatgpt-5-search-configuration/", "published": "2025-08-20" }, "kind": "mechanism", "quality": 0.66, "canonical_fact_id": "chatgpt-5-search-configuration#2", "published": "2025-08-20", "age_days": 419, "possibly_outdated": true }, { "id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#2", "fact": "In the analyzed ChatGPT sample, 75% of queries were task-oriented commands rather than only questions.", "evidence": "Primary Intent Task-Oriented A staggering 75% of queries are not questions, but also commands to perform a specific action.", "topic": "chatgpt-search", "source": { "title": "I Analyzed 1827 Real User Prompts from ChatGPT - Here What I've Found: Agentic Search Will Be The New Hype", "url": "https://metehan.ai/blog/i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search/", "published": "2025-08-10" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#2", "published": "2025-08-10", "age_days": 429, "possibly_outdated": true }, { "id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#1", "fact": "Among 1,827 analyzed ChatGPT queries, the average prompt contained 42 words and the median contained 11 words.", "evidence": "Total Queries Analyzed 1,827 A robust dataset illustrating emerging user behaviors. \n Average Prompt Length (Words) 42 Users are providing detailed instructions, treating the AI as a collaborator, not just a search box. \n Median Prompt Length (Words) 11", "topic": "chatgpt-search", "source": { "title": "I Analyzed 1827 Real User Prompts from ChatGPT - Here What I've Found: Agentic Search Will Be The New Hype", "url": "https://metehan.ai/blog/i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search/", "published": "2025-08-10" }, "kind": "finding", "quality": 0.94, "canonical_fact_id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#1", "published": "2025-08-10", "age_days": 429, "possibly_outdated": true }, { "id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#3", "fact": "Code and development accounted for 40% of all task-oriented ChatGPT queries in the analyzed sample.", "evidence": "A massive 40% of all task-oriented queries were related to code and development. The AI is being used as an indispensable tool for debugging, learning, and increasing productivity.", "topic": "chatgpt-search", "source": { "title": "I Analyzed 1827 Real User Prompts from ChatGPT - Here What I've Found: Agentic Search Will Be The New Hype", "url": "https://metehan.ai/blog/i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search/", "published": "2025-08-10" }, "kind": "finding", "quality": 0.85, "canonical_fact_id": "i-analyzed-1827-real-user-prompts-from-chatgpt-here-what-ive-found-agentic-search#3", "published": "2025-08-10", "age_days": 429, "possibly_outdated": true }, { "id": "perplexity-ai-seo-59-ranking-patterns#2", "fact": "Perplexity's L3 reranker can discard an entire result set when too few results pass its quality checks.", "evidence": "L3 Reranking: l3_xgb_model applies XGBoost quality thresholds; l3_reranker_drop_all_docs_if_count_less_equal discards entire result sets below minimum count (Browser analysis, 2025).", "topic": "retrieval-and-ranking", "source": { "title": "Breaking: Perplexity's 59 Ranking Patterns and Secret Browser Architecture Revealed (With Code)", "url": "https://metehan.ai/blog/perplexity-ai-seo-59-ranking-patterns/", "published": "2025-08-04" }, "kind": "mechanism", "quality": 0.91, "canonical_fact_id": "perplexity-ai-seo-59-ranking-patterns#2", "published": "2025-08-04", "age_days": 435, "possibly_outdated": true }, { "id": "perplexity-ai-seo-59-ranking-patterns#1", "fact": "Perplexity applies an L3 machine learning reranking layer after initial retrieval and scoring for entity searches.", "evidence": "This advanced reranking layer operates after initial retrieval and scoring, applying machine learning models to refine result quality.", "topic": "perplexity", "source": { "title": "Breaking: Perplexity's 59 Ranking Patterns and Secret Browser Architecture Revealed (With Code)", "url": "https://metehan.ai/blog/perplexity-ai-seo-59-ranking-patterns/", "published": "2025-08-04" }, "kind": "mechanism", "quality": 0.84, "canonical_fact_id": "perplexity-ai-seo-59-ranking-patterns#1", "published": "2025-08-04", "age_days": 435, "possibly_outdated": true }, { "id": "perplexity-ai-seo-59-ranking-patterns#3", "fact": "Perplexity tracks weekly engagement with discover_engagement_7d and long-term performance with historic_engagement_v1.", "evidence": "`discover_engagement_7d`: Tracks weekly user engagement patterns\n`historic_engagement_v1`: Considers long-term performance history", "topic": "perplexity", "source": { "title": "Breaking: Perplexity's 59 Ranking Patterns and Secret Browser Architecture Revealed (With Code)", "url": "https://metehan.ai/blog/perplexity-ai-seo-59-ranking-patterns/", "published": "2025-08-04" }, "kind": "mechanism", "quality": 0.74, "canonical_fact_id": "perplexity-ai-seo-59-ranking-patterns#3", "published": "2025-08-04", "age_days": 435, "possibly_outdated": true }, { "id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#2", "fact": "Metehan Yesilyurt's Chrome DevTools analysis found a standard Reciprocal Rank Fusion implementation in ChatGPT's search code.", "evidence": "While inspecting ChatGPT's search implementation, I found this revealing code:\nThis confirms ChatGPT uses standard RRF to combine search results.", "topic": "chatgpt-search", "source": { "title": "Is RRF the Secret to Dominating AI Citations? I Decoded ChatGPT's Ranking Formula", "url": "https://metehan.ai/blog/chatgpt-is-using-reciprocal-rank-fusion-rrf/", "published": "2025-07-22" }, "kind": "mechanism", "quality": 0.91, "canonical_fact_id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#2", "published": "2025-07-22", "age_days": 448, "possibly_outdated": true }, { "id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#3", "fact": "In the article's simulation, ranking fifth for 10 related keywords produced an RRF score of 0.154, versus 0.0164 for ranking first for one keyword.", "evidence": "10 related keywords at position #5: 10 × (1/(60+5)) = 10 × 0.0154 = 0.154\nTotal RRF = 0.154\nThe topic cluster scores almost 10x higher (0.154 vs 0.0164).", "topic": "retrieval-and-ranking", "source": { "title": "Is RRF the Secret to Dominating AI Citations? I Decoded ChatGPT's Ranking Formula", "url": "https://metehan.ai/blog/chatgpt-is-using-reciprocal-rank-fusion-rrf/", "published": "2025-07-22" }, "kind": "finding", "quality": 0.79, "canonical_fact_id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#3", "published": "2025-07-22", "age_days": 448, "possibly_outdated": true }, { "id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#1", "fact": "Reciprocal Rank Fusion combines results from multiple queries using the score 1/(60 + rank position), then adds scores when the same page appears in several result lists.", "evidence": "It's a method to combine search results from multiple queries into one final ranking. The formula is straightforward:\nRRF score = 1/(60 + rank position)", "topic": "retrieval-and-ranking", "source": { "title": "Is RRF the Secret to Dominating AI Citations? I Decoded ChatGPT's Ranking Formula", "url": "https://metehan.ai/blog/chatgpt-is-using-reciprocal-rank-fusion-rrf/", "published": "2025-07-22" }, "kind": "mechanism", "quality": 0.77, "canonical_fact_id": "chatgpt-is-using-reciprocal-rank-fusion-rrf#1", "published": "2025-07-22", "age_days": 448, "possibly_outdated": true }, { "id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#1", "fact": "In Metehan Yesilyurt's test, a new website submitted only through IndexNow had zero indexed pages after three days.", "evidence": "I had added the content on July 8. I only used IndexNow. After 3 days, there were still zero indexed pages.", "topic": "crawlers-and-indexing", "source": { "title": "What Happens When You Launch a Website Without Google or Bing? (AI Citation Test)", "url": "https://metehan.ai/blog/what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test/", "published": "2025-07-18" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#1", "published": "2025-07-18", "age_days": 452, "possibly_outdated": true }, { "id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#3", "fact": "ChatGPT showed three source pages for the test website, with meta descriptions matching Google's results.", "evidence": "ChatGPT (sources): 3 pages shown (Meta Descriptions are same as Google)", "topic": "chatgpt-search", "source": { "title": "What Happens When You Launch a Website Without Google or Bing? (AI Citation Test)", "url": "https://metehan.ai/blog/what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test/", "published": "2025-07-18" }, "kind": "finding", "quality": 0.86, "canonical_fact_id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#3", "published": "2025-07-18", "age_days": 452, "possibly_outdated": true }, { "id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#2", "fact": "By July 18, the test website had three pages indexed by Google and one page indexed by Bing.", "evidence": "And yet… somehow, the site got indexed by Google and Bing.\nCheck my ChatGPT query here: https://chatgpt.com/share/687a0a32-641c-8011-8594-8f7b8e39eeb3\nGoogle: 3 pages\nBing: 1 page", "topic": "crawlers-and-indexing", "source": { "title": "What Happens When You Launch a Website Without Google or Bing? (AI Citation Test)", "url": "https://metehan.ai/blog/what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test/", "published": "2025-07-18" }, "kind": "finding", "quality": 0.84, "canonical_fact_id": "what-happens-when-you-launch-a-website-without-google-or-bing-ai-citation-test#2", "published": "2025-07-18", "age_days": 452, "possibly_outdated": true }, { "id": "temperature-llms#1", "fact": "At temperature 0.0, an LLM selects the highest-probability token at each generation step.", "evidence": "At temperature 0.0, we enter the realm of complete determinism, the model always selects the highest probability token at each step.", "topic": "llm-internals", "source": { "title": "The Temperature Zero Trick: How to Discover What LLMs Really Know About Any Company in 30 Seconds", "url": "https://metehan.ai/blog/temperature-llms/", "published": "2025-07-16" }, "kind": "mechanism", "quality": 0.6, "canonical_fact_id": "temperature-llms#1", "published": "2025-07-16", "age_days": 454, "possibly_outdated": true }, { "id": "llm-optimization-analyzer-screaming-frog#1", "fact": "On a legal search dataset using GPT-4o, batched pointwise scoring improved NDCG@10 from 43.8% to 51.3%, a 7.5 percentage point gain.", "evidence": "Batched PW improved from 43.8% to 51.3% NDCG@10, a significant +7.5 percentage point gain", "topic": "retrieval-and-ranking", "source": { "title": "Is It Possible to Build? Universal LLM Optimization Analyzer with Gemini API for Screaming Frog", "url": "https://metehan.ai/blog/llm-optimization-analyzer-screaming-frog/", "published": "2025-07-09" }, "kind": "finding", "quality": 0.7, "canonical_fact_id": "llm-optimization-analyzer-screaming-frog#1", "published": "2025-07-09", "age_days": 461, "possibly_outdated": true }, { "id": "llm-optimization-analyzer-screaming-frog#2", "fact": "In the C-SEO Bench study, 61% of optimized pages had no citation ranking change, while 26.2% improved and 12.8% declined.", "evidence": "Breakdown from the study:\nNo ranking change: 61%\nPositive change: 26.2%\nNegative change: 12.8%", "topic": "geo-aeo-method", "source": { "title": "Is It Possible to Build? Universal LLM Optimization Analyzer with Gemini API for Screaming Frog", "url": "https://metehan.ai/blog/llm-optimization-analyzer-screaming-frog/", "published": "2025-07-09" }, "kind": "finding", "quality": 0.7, "canonical_fact_id": "llm-optimization-analyzer-screaming-frog#2", "published": "2025-07-09", "age_days": 461, "possibly_outdated": true }, { "id": "citemet-method-part-2-ai-memory-optimization#2", "fact": "Metehan Yesilyurt observed 1.6 times more GPTBot crawl activity on WordPress wp-json endpoints than on regular pages.", "evidence": "GPTBot loves it. I'm seeing 1.6x more crawl activity on wp-json endpoints than regular pages.", "topic": "crawlers-and-indexing", "source": { "title": "CiteMET Method Part 2: The Technical Playbook for AI Memory Optimization", "url": "https://metehan.ai/blog/citemet-method-part-2-ai-memory-optimization/", "published": "2025-07-02" }, "kind": "finding", "quality": 0.9, "canonical_fact_id": "citemet-method-part-2-ai-memory-optimization#2", "published": "2025-07-02", "age_days": 468, "possibly_outdated": true }, { "id": "citemet-method-part-2-ai-memory-optimization#3", "fact": "In Yesilyurt's test, ChatGPT did not execute directives from a fetched page and instead searched its index for another test page.", "evidence": "It's weird, however it didn't work for ChatGPT. ChatGPT looked for search index. Found my another test page that I created before.", "topic": "chatgpt-search", "source": { "title": "CiteMET Method Part 2: The Technical Playbook for AI Memory Optimization", "url": "https://metehan.ai/blog/citemet-method-part-2-ai-memory-optimization/", "published": "2025-07-02" }, "kind": "finding", "quality": 0.8, "canonical_fact_id": "citemet-method-part-2-ai-memory-optimization#3", "published": "2025-07-02", "age_days": 468, "possibly_outdated": true }, { "id": "screaming-frog-muvera-analysis#1", "fact": "MUVERA converts query and document multi-vector sets into fixed-dimensional encodings for single-vector maximum inner product search.", "evidence": "The research describes \"mappings to convert query and document multi-vector sets into FDEs.\"", "topic": "retrieval-and-ranking", "source": { "title": "[MUVERA] How to Run a Multi-Vector Retrieval Analysis with Screaming Frog", "url": "https://metehan.ai/blog/screaming-frog-muvera-analysis/", "published": "2025-07-01" }, "kind": "mechanism", "quality": 0.69, "canonical_fact_id": "screaming-frog-muvera-analysis#1", "published": "2025-07-01", "age_days": 469, "possibly_outdated": true }, { "id": "screaming-frog-muvera-analysis#3", "fact": "Google reported that MUVERA achieved 10% higher recall and a 90% reduction in latency.", "evidence": "Google reported that MUVERA achieves \"10% higher recall with a remarkable 90% reduction in latency.\"", "topic": "retrieval-and-ranking", "source": { "title": "[MUVERA] How to Run a Multi-Vector Retrieval Analysis with Screaming Frog", "url": "https://metehan.ai/blog/screaming-frog-muvera-analysis/", "published": "2025-07-01" }, "kind": "finding", "quality": 0.68, "canonical_fact_id": "screaming-frog-muvera-analysis#3", "published": "2025-07-01", "age_days": 469, "possibly_outdated": true }, { "id": "screaming-frog-muvera-analysis#2", "fact": "Chamfer similarity measures the maximum similarity between each query embedding and the closest document embedding.", "evidence": "Google's research describes Chamfer similarity as measuring \"the maximum similarity between each query embedding and the closest document embedding.\"", "topic": "retrieval-and-ranking", "source": { "title": "[MUVERA] How to Run a Multi-Vector Retrieval Analysis with Screaming Frog", "url": "https://metehan.ai/blog/screaming-frog-muvera-analysis/", "published": "2025-07-01" }, "kind": "mechanism", "quality": 0.65, "canonical_fact_id": "screaming-frog-muvera-analysis#2", "published": "2025-07-01", "age_days": 469, "possibly_outdated": true }, { "id": "query-fan-out-screaming-frog-ai#2", "fact": "A tested e-commerce product page had 12 semantic chunks and covered 3 of 10 predicted queries.", "evidence": "Case Study 2: E-commerce Product Page\nSemantic Chunks Found: 12\nQuery Coverage: 3/10", "topic": "ai-visibility-measurement", "source": { "title": "How to Run Query Fan-Out Analysis in Screaming Frog Using Gemini AI", "url": "https://metehan.ai/blog/query-fan-out-screaming-frog-ai/", "published": "2025-06-26" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "query-fan-out-screaming-frog-ai#2", "published": "2025-06-26", "age_days": 474, "possibly_outdated": true }, { "id": "query-fan-out-screaming-frog-ai#1", "fact": "In an experiment covering more than 2,000 pages, most pages answered only 30% of the queries Gemini predicted Google AI Mode would generate from their content.", "evidence": "After testing it on over 2000 pages, I discovered something that can concern every SEO: most pages only answer 30% of the queries that Google’s AI actually generates from their content.", "topic": "ai-visibility-measurement", "source": { "title": "How to Run Query Fan-Out Analysis in Screaming Frog Using Gemini AI", "url": "https://metehan.ai/blog/query-fan-out-screaming-frog-ai/", "published": "2025-06-26" }, "kind": "finding", "quality": 0.87, "canonical_fact_id": "query-fan-out-screaming-frog-ai#1", "published": "2025-06-26", "age_days": 474, "possibly_outdated": true }, { "id": "google-document-ai-layout-parser#2", "fact": "On a tested app landing page, Google Document AI Layout Parser encountered 121 navigation blocks before reaching the first content block at position 122.", "evidence": "One of the pages had 121 navigation-related blocks before any actual content appeared. The parser had to process all of that before reaching the main content at block 122.", "topic": "crawlers-and-indexing", "source": { "title": "Experiment with Google Document AI Layout Parser: Here's What I Found", "url": "https://metehan.ai/blog/google-document-ai-layout-parser/", "published": "2025-06-22" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "google-document-ai-layout-parser#2", "published": "2025-06-22", "age_days": 478, "possibly_outdated": true }, { "id": "google-document-ai-layout-parser#3", "fact": "In Metehan Yesilyurt's experiment, Google Document AI Layout Parser preserved table and list structures, including cell spans and list types.", "evidence": "The parser perfectly preserves table and list structures, including cell spans and list types.", "topic": "crawlers-and-indexing", "source": { "title": "Experiment with Google Document AI Layout Parser: Here's What I Found", "url": "https://metehan.ai/blog/google-document-ai-layout-parser/", "published": "2025-06-22" }, "kind": "finding", "quality": 0.88, "canonical_fact_id": "google-document-ai-layout-parser#3", "published": "2025-06-22", "age_days": 478, "possibly_outdated": true }, { "id": "google-document-ai-layout-parser#1", "fact": "In Metehan Yesilyurt's test, Google Document AI Layout Parser assigned a unique ID to every DOM element, including empty elements.", "evidence": "The parser assigns a unique ID to every DOM element, including empty ones:", "topic": "crawlers-and-indexing", "source": { "title": "Experiment with Google Document AI Layout Parser: Here's What I Found", "url": "https://metehan.ai/blog/google-document-ai-layout-parser/", "published": "2025-06-22" }, "kind": "finding", "quality": 0.87, "canonical_fact_id": "google-document-ai-layout-parser#1", "published": "2025-06-22", "age_days": 478, "possibly_outdated": true }, { "id": "google-vertex-ai-ranking-api#1", "fact": "Google's SEO Starter Guide ranked second for “what is SEO” and received a 0.996 score from the Vertex AI Ranking API.", "evidence": "SERP #2 → AI Score: 0.996\nGoogle’s own SEO Starter Guide\nComprehensive\nAuthoritative\nAligned with user intent", "topic": "retrieval-and-ranking", "source": { "title": "Experiment with Google Vertex AI Ranking API: Here's What I Found (Free Script)", "url": "https://metehan.ai/blog/google-vertex-ai-ranking-api/", "published": "2025-06-17" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "google-vertex-ai-ranking-api#1", "published": "2025-06-17", "age_days": 483, "possibly_outdated": true }, { "id": "google-vertex-ai-ranking-api#3", "fact": "In Yesilyurt's Vertex AI analysis, pages with two or three passages scoring above 0.900 consistently outperformed keyword-stuffed 2,000-word pages.", "evidence": "Pages with just 2–3 clear passages (scoring >0.900) consistently outperformed keyword-stuffed, bloated 2,000-word pieces.", "topic": "retrieval-and-ranking", "source": { "title": "Experiment with Google Vertex AI Ranking API: Here's What I Found (Free Script)", "url": "https://metehan.ai/blog/google-vertex-ai-ranking-api/", "published": "2025-06-17" }, "kind": "finding", "quality": 0.96, "canonical_fact_id": "google-vertex-ai-ranking-api#3", "published": "2025-06-17", "age_days": 483, "possibly_outdated": true }, { "id": "google-vertex-ai-ranking-api#2", "fact": "In Yesilyurt's analysis, one 300-word page ranked third and had 47 internal links from the homepage.", "evidence": "I found a page ranking #3 with only 300 words of fluff.\nWhy did it rank?\n47 internal links from the homepage", "topic": "retrieval-and-ranking", "source": { "title": "Experiment with Google Vertex AI Ranking API: Here's What I Found (Free Script)", "url": "https://metehan.ai/blog/google-vertex-ai-ranking-api/", "published": "2025-06-17" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "google-vertex-ai-ranking-api#2", "published": "2025-06-17", "age_days": 483, "possibly_outdated": true }, { "id": "semrush-mcp#2", "fact": "The Semrush MCP server adds roughly 1 to 2 seconds of latency per query because requests make an additional network hop between Claude, the server, and Semrush.", "evidence": "The additional network hop (Claude → MCP Server → Semrush → MCP Server → Claude) adds roughly 1-2 seconds to each query.", "topic": "seo-tools-and-automation", "source": { "title": "Semrush MCP Server: Integrating Semrush with AI (Code Inside)", "url": "https://metehan.ai/blog/semrush-mcp/", "published": "2025-06-03" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "semrush-mcp#2", "published": "2025-06-03", "age_days": 497, "possibly_outdated": true }, { "id": "geo-or-seo-the-future-of-seo-is-not-what-you-think#1", "fact": "Google AI Mode expands queries into hundreds of semantically related intents, converts them into high-dimensional vectors, and returns synthesized, context-aware answers.", "evidence": "It fans out the query into hundreds of semantically related intents, converts them into high-dimensional vectors, and returns synthesized, context-aware answers.", "topic": "google-ai-mode-and-overviews", "source": { "title": "GEO or SEO? The Future of SEO is Not What You Think", "url": "https://metehan.ai/blog/geo-or-seo-the-future-of-seo-is-not-what-you-think/", "published": "2025-05-30" }, "kind": "mechanism", "quality": 0.65, "canonical_fact_id": "geo-or-seo-the-future-of-seo-is-not-what-you-think#1", "published": "2025-05-30", "age_days": 501, "possibly_outdated": true }, { "id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#1", "fact": "Germany's Google AI Overview appearances rose from 4 in April to 57,000 in May, a reported increase of 40,981,900%.", "evidence": "Germany 57,000 \\~1.6M +40,981,900% (from just 4 in April)", "topic": "google-ai-mode-and-overviews", "source": { "title": "Google Translate Proxy Pages Are Dominating AI Overviews in Europe, too, And No One Controls It", "url": "https://metehan.ai/blog/google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it/", "published": "2025-05-27" }, "kind": "finding", "quality": 0.88, "canonical_fact_id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#1", "published": "2025-05-27", "age_days": 504, "possibly_outdated": true }, { "id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#2", "fact": "Google AI Overview data reported 109,127 appearances in Spain, 49,336 in the Netherlands, and 35,847 in France.", "evidence": "Netherlands 49,336 \\~1M , \n Spain 109,127 \\~1.2M , \n Germany 57,000 \\~1.6M +40,981,900% (from just 4 in April) \n France 35,847 \\~1M , ", "topic": "google-ai-mode-and-overviews", "source": { "title": "Google Translate Proxy Pages Are Dominating AI Overviews in Europe, too, And No One Controls It", "url": "https://metehan.ai/blog/google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it/", "published": "2025-05-27" }, "kind": "finding", "quality": 0.86, "canonical_fact_id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#2", "published": "2025-05-27", "age_days": 504, "possibly_outdated": true }, { "id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#3", "fact": "Google Translate proxy results are loaded through translate.google.com/translate and rendered under translate.goog instead of being original publisher pages.", "evidence": "These are not original content pages from publishers, they are Google-hosted translated versions of English content, loaded via translate.google.com/translate and rendered under translate.goog.", "topic": "google-ai-mode-and-overviews", "source": { "title": "Google Translate Proxy Pages Are Dominating AI Overviews in Europe, too, And No One Controls It", "url": "https://metehan.ai/blog/google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it/", "published": "2025-05-27" }, "kind": "mechanism", "quality": 0.75, "canonical_fact_id": "google-translate-proxy-pages-are-dominating-ai-overviews-in-europe-too-and-no-one-controls-it#3", "published": "2025-05-27", "age_days": 504, "possibly_outdated": true }, { "id": "google-ai-mode-optimization-query-fan-out#1", "fact": "In Metehan Yesilyurt's website analysis, complex Google AI Mode queries were divided into five to six subtopics.", "evidence": "Observation: Complex queries were divided into 5-6 sub-topics\nFinding: Main topics were covered, but some important sub-topics were missing", "topic": "google-ai-mode-and-overviews", "source": { "title": "Google AI Mode Optimization: I Analyzed My Website with Query Fan-Out Technique - What Did I Learn?", "url": "https://metehan.ai/blog/google-ai-mode-optimization-query-fan-out/", "published": "2025-05-26" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "google-ai-mode-optimization-query-fan-out#1", "published": "2025-05-26", "age_days": 505, "possibly_outdated": true }, { "id": "google-ai-mode-optimization-query-fan-out#2", "fact": "Yesilyurt's analysis classified 83 out of every 100 content pieces as orphan content with no or weak internal linking for relevant Google AI Mode queries.", "evidence": "Observation: 83 out of every 100 pieces of content were grouped as \"orphan content\" with no/weak internal linking for relevant AI Mode queries", "topic": "google-ai-mode-and-overviews", "source": { "title": "Google AI Mode Optimization: I Analyzed My Website with Query Fan-Out Technique - What Did I Learn?", "url": "https://metehan.ai/blog/google-ai-mode-optimization-query-fan-out/", "published": "2025-05-26" }, "kind": "finding", "quality": 0.93, "canonical_fact_id": "google-ai-mode-optimization-query-fan-out#2", "published": "2025-05-26", "age_days": 505, "possibly_outdated": true }, { "id": "google-ai-mode-optimization-query-fan-out#3", "fact": "Yesilyurt's website analysis detected 90 potential semantic clusters and found that the content lacked a hub-and-spoke structure.", "evidence": "Observation: 90 potential semantic clusters were detected\nFinding: Content lacks hub/spoke structure", "topic": "retrieval-and-ranking", "source": { "title": "Google AI Mode Optimization: I Analyzed My Website with Query Fan-Out Technique - What Did I Learn?", "url": "https://metehan.ai/blog/google-ai-mode-optimization-query-fan-out/", "published": "2025-05-26" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "google-ai-mode-optimization-query-fan-out#3", "published": "2025-05-26", "age_days": 505, "possibly_outdated": true }, { "id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#1", "fact": "Google Translate’s /translate subfolder reached 606 million monthly organic visits.", "evidence": "Google Translate’s /translate subfolder has reached 606 million monthly organic visits", "topic": "ai-visibility-measurement", "source": { "title": "Reddit Is Dead, Unseen by SEOs: Google Translate’s Proxy Pages Pull 606M Visits Monthly", "url": "https://metehan.ai/blog/reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly/", "published": "2025-05-20" }, "kind": "finding", "quality": 0.77, "canonical_fact_id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#1", "published": "2025-05-20", "age_days": 511, "possibly_outdated": true }, { "id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#2", "fact": "Google may insert translated authoritative English pages into search results when high-quality non-English content is scarce.", "evidence": "When high-quality content in non-English languages is scarce, Google often inserts translated versions of authoritative English pages directly into the SERP, using translate.google.com/translate links.", "topic": "retrieval-and-ranking", "source": { "title": "Reddit Is Dead, Unseen by SEOs: Google Translate’s Proxy Pages Pull 606M Visits Monthly", "url": "https://metehan.ai/blog/reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly/", "published": "2025-05-20" }, "kind": "mechanism", "quality": 0.73, "canonical_fact_id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#2", "published": "2025-05-20", "age_days": 511, "possibly_outdated": true }, { "id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#3", "fact": "ChatGPT, Perplexity, and Gemini citations sometimes link to Google Translate versions of Reddit rather than original sources.", "evidence": "AI citations in ChatGPT, Perplexity and Gemini sometimes go to Google Translate versions of Reddit instead of original sources.", "topic": "retrieval-and-ranking", "source": { "title": "Reddit Is Dead, Unseen by SEOs: Google Translate’s Proxy Pages Pull 606M Visits Monthly", "url": "https://metehan.ai/blog/reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly/", "published": "2025-05-20" }, "kind": "finding", "quality": 0.67, "canonical_fact_id": "reddit-is-dead-unseen-by-seos-google-translates-proxy-pages-pull-606m-visits-monthly#3", "published": "2025-05-20", "age_days": 511, "possibly_outdated": true }, { "id": "ai-machine-learning-in-search#1", "fact": "Google RankBrain primarily interprets search queries, especially the approximately 15% of daily searches that Google has never seen before.", "evidence": "Query Interpretation: RankBrain's primary function is interpreting search queries, particularly those Google has never seen before (which account for approximately 15% of daily searches).", "topic": "retrieval-and-ranking", "source": { "title": "AI & Machine Learning in Search: The Complete Guide for SEO Professionals", "url": "https://metehan.ai/blog/ai-machine-learning-in-search/", "published": "2025-04-26" }, "kind": "mechanism", "quality": 0.7, "canonical_fact_id": "ai-machine-learning-in-search#1", "published": "2025-04-26", "age_days": 535, "possibly_outdated": true }, { "id": "ai-machine-learning-in-search#2", "fact": "Google BERT analyzes each word in relation to every other word in a sentence to understand context.", "evidence": "Bidirectional Context: Unlike previous models that read text sequentially, BERT analyzes words in relation to all other words in a sentence, enabling true context understanding.", "topic": "llm-internals", "source": { "title": "AI & Machine Learning in Search: The Complete Guide for SEO Professionals", "url": "https://metehan.ai/blog/ai-machine-learning-in-search/", "published": "2025-04-26" }, "kind": "mechanism", "quality": 0.69, "canonical_fact_id": "ai-machine-learning-in-search#2", "published": "2025-04-26", "age_days": 535, "possibly_outdated": true }, { "id": "the-ultimate-hcu-recovery-playbook#2", "fact": "A health and wellness site in Yesilyurt's case study recovered 37% of traffic after Chrome-targeted campaigns for its most valuable commercial content.", "evidence": "A health and wellness site I worked with saw a 37% recovery in traffic after implementing Chrome-targeted traffic campaigns for their most valuable commercial content.", "topic": "retrieval-and-ranking", "source": { "title": "The Ultimate HCU Recovery Playbook: Proven Tactics to Reclaim Your Lost SEO Traffic", "url": "https://metehan.ai/blog/the-ultimate-hcu-recovery-playbook/", "published": "2025-04-25" }, "kind": "finding", "quality": 0.67, "canonical_fact_id": "the-ultimate-hcu-recovery-playbook#2", "published": "2025-04-25", "age_days": 536, "possibly_outdated": true }, { "id": "an-open-source-google-reranker-transforming-google-results-with-ai-magic#3", "fact": "The Google ReRanker combines factor scores into a weighted total score, with configurable weights in the _initialize_weights() method.", "evidence": "The tool combines all factor scores into a weighted total score. Weights are preconfigured but can be adjusted in the _initialize_weights() method.", "topic": "retrieval-and-ranking", "source": { "title": "An Open-Source Google ReRanker: Transforming Google Results with AI Magic", "url": "https://metehan.ai/blog/an-open-source-google-reranker-transforming-google-results-with-ai-magic/", "published": "2024-11-19" }, "kind": "mechanism", "quality": 0.7, "canonical_fact_id": "an-open-source-google-reranker-transforming-google-results-with-ai-magic#3", "published": "2024-11-19", "age_days": 693, "possibly_outdated": true }, { "id": "an-open-source-google-reranker-transforming-google-results-with-ai-magic#2", "fact": "The Google ReRanker's get_search_results() function uses the requests library and BeautifulSoup to retrieve and parse search results.", "evidence": "get_search_results() uses the requests library to make a search query and parses results using BeautifulSoup. It extracts the titles, URLs, and snippets for further analysis.", "topic": "crawlers-and-indexing", "source": { "title": "An Open-Source Google ReRanker: Transforming Google Results with AI Magic", "url": "https://metehan.ai/blog/an-open-source-google-reranker-transforming-google-results-with-ai-magic/", "published": "2024-11-19" }, "kind": "mechanism", "quality": 0.68, "canonical_fact_id": "an-open-source-google-reranker-transforming-google-results-with-ai-magic#2", "published": "2024-11-19", "age_days": 693, "possibly_outdated": true }, { "id": "how-to-rank-1-in-searchgpt-a-step-by-step-guide#2", "fact": "SearchGPT matches queries against indexed documents using TF-IDF or BM25 relevance scoring.", "evidence": "Document Matching: Each query is matched against indexed documents using TF-IDF (Term Frequency-Inverse Document Frequency) or BM25 (Best Matching) for relevance scoring.", "topic": "retrieval-and-ranking", "source": { "title": "How I Got to Ranking #1 in SearchGPT: A Step-by-Step Guide", "url": "https://metehan.ai/blog/how-to-rank-1-in-searchgpt-a-step-by-step-guide/", "published": "2024-11-13" }, "kind": "mechanism", "quality": 0.73, "canonical_fact_id": "how-to-rank-1-in-searchgpt-a-step-by-step-guide#2", "published": "2024-11-13", "age_days": 699, "possibly_outdated": true }, { "id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#2", "fact": "A newly created appsamurai.com page submitted through IndexNow appeared in SearchGPT after 48 hours with one citation pointing to the correct URL.", "evidence": "It's indexed after 48 hours, there is only one single citation result that points to a correct URL. Bing index records for this page on appsamurai.com, only took around 2 minutes.", "topic": "chatgpt-search", "source": { "title": "How to Submit & Index Your Page on SearchGPT? (An Experiement & Results)", "url": "https://metehan.ai/blog/how-to-submit-index-your-page-on-searchgpt-an-experiement-results/", "published": "2024-11-10" }, "kind": "finding", "quality": 0.95, "canonical_fact_id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#2", "published": "2024-11-10", "age_days": 702, "possibly_outdated": true }, { "id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#3", "fact": "The new metehan.ai page remained unindexed on SearchGPT and Bing after three days.", "evidence": "Unfortunately, there were no results for metehan.ai, and my page remained unindexed on SearchGPT & Bing after three days.", "topic": "crawlers-and-indexing", "source": { "title": "How to Submit & Index Your Page on SearchGPT? (An Experiement & Results)", "url": "https://metehan.ai/blog/how-to-submit-index-your-page-on-searchgpt-an-experiement-results/", "published": "2024-11-10" }, "kind": "finding", "quality": 0.91, "canonical_fact_id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#3", "published": "2024-11-10", "age_days": 702, "possibly_outdated": true }, { "id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#1", "fact": "SearchGPT used GPTBot, rather than the expected OAIBot, when requesting a page title during the experiment.", "evidence": "The response and logs were interesting on my server: SearchGPT used GPTBot as indicated in my server logs, rather than the expected OAIBot.", "topic": "crawlers-and-indexing", "source": { "title": "How to Submit & Index Your Page on SearchGPT? (An Experiement & Results)", "url": "https://metehan.ai/blog/how-to-submit-index-your-page-on-searchgpt-an-experiement-results/", "published": "2024-11-10" }, "kind": "finding", "quality": 0.9, "canonical_fact_id": "how-to-submit-index-your-page-on-searchgpt-an-experiement-results#1", "published": "2024-11-10", "age_days": 702, "possibly_outdated": true } ] }