GEO Research Digests

Ahrefs' Schema AI Citations Study, Digested

Schema AI citations study, digested: Ahrefs tracked 1,885 pages adding schema against matched controls. What it found, what it can't show, and what to do next.

AI Crawlers JavaScript Study: Vercel's Findings Explained

AI crawlers JavaScript study from Vercel: ChatGPT and Claude bots fetch JS files but don't run them. Figures, caveats, and the check to run on your site.

AI Overview Citations Top 10 Study: Ahrefs, Digested

AI Overview citations top 10 study: Ahrefs found 37.9% of cited URLs appeared in the top 10 in March 2026. What it shows, what it misses, and what to do.

AI Search Citation Concentration Study, Digested

AI search citation concentration study digested: top 20 news sources took 67.3% of OpenAI citations. What it shows, what it doesn't, and how to benchmark.

Anthropic Crawler Documentation: ClaudeBot and More

Anthropic crawler documentation, digested: what ClaudeBot, Claude-SearchBot and Claude-User do, how to set each in robots.txt, and how to test the file is reachable.

Bing data-nosnippet Support: What It Does for AI

Bing data-nosnippet support, digested: marked content stays indexed and rankable but is kept out of snippets and AI summaries. Learn the risks and a quick audit.

C-SEO Bench Digest: Does Conversational SEO Work?

C-SEO Bench tested conversational SEO rewrites across LLMs. Most were ineffective or harmful. Here is what the study measured, what it missed, and what to do.

Claude Web Search Tool Documentation: What It Reveals

Claude web search tool documentation, digested: what the API returns (url, title, page_age, cited_text), what it costs, and what it can't tell you.

Document Chunking Strategies Study: Who Controls Chunks

Document chunking strategies study (arXiv 2603.06976) explained: what 36 methods tested, who sets chunks, and why it doesn't support chunking your pages.

FActScore: How Factual Precision Is Measured

FActScore explained: how arXiv 2305.14251 splits AI answers into atomic facts, scores the supported share, and what brand teams can borrow for accuracy checks.

Generative Engine Optimization Paper (arXiv 2311.09735)

Generative engine optimization paper digested: what arXiv 2311.09735 tested, what the 30 to 40% band means, and why it says nothing about AI search reach.

GEO Critical Survey (arXiv 2607.14035), Digested

GEO critical survey digested: what arXiv 2607.14035 found across 45 studies, what it can't show, and how to separate lab effects from getting found.

Google AI Features and Your Website: The Doc, Digested

Google AI features and your website, digested: what Google's doc says about AI Overviews eligibility, what it leaves out, and a five-step check you can run.

Google AI Optimization Guide: What It Says, Digested

Google AI optimization guide, digested: what the July 2026 document says about llms.txt, markup, page length and writing for AI, and which tasks you can drop.

Google Common Crawlers Documentation and Google-Extended

Google common crawlers documentation, digested: what Google-Extended and Google-CloudVertexBot control, and why no robots.txt token blocks AI Overviews.

Google Guidance on AI-Generated Content, Digested

Google guidance on AI-generated content: appropriate AI use is allowed, scaled content abuse is not. What the policy says and how to use AI safely.

Google Helpful Content Guidance and the E-E-A-T Myth

Google helpful content guidance explained: what people-first content means, why E-E-A-T isn't a ranking factor, and why any E-E-A-T score is invented.

How ChatGPT Search Works: OpenAI Help Article Explained

How ChatGPT search works, per OpenAI's help article: eligibility needs OAI-SearchBot and open firewalls, prompts get rewritten, and memory shapes queries.

How Do LLMs Cite Sources? The Attribution Mechanism Study

How do LLMs cite sources? A mechanistic study on Llama-3.1-8B shows citation is separable from source use. What it found and how to read AI citation counts.

IndexNow Documentation: What the Spec Promises for AI

IndexNow documentation, digested: the spec never mentions AI, a 200 response confirms receipt only, and Google is not among the seven participating engines.

LLM Search Agent Manipulation Study (arXiv 2606.16821)

LLM search agent manipulation study digested: attack success from 0.0% to 31.4% on a mock search API, what it does not show, and how to defend your brand.

Meta Web Crawlers Documentation, Digested

Meta web crawlers documentation digested: what Meta-WebIndexer, Meta-ExternalAgent and Meta-ExternalFetcher do, and what to check in robots.txt and your firewall.

OpenAI Crawlers Documentation: GPTBot, OAI-SearchBot

OpenAI crawlers documentation, digested: what GPTBot, OAI-SearchBot and ChatGPT-User each control, two common robots.txt errors, and a per-bot decision.

Otterly 2,000 AI Blogs Experiment: Method and Limits

Otterly 2000 AI blogs experiment digested: how Google deindexed both sites, why Copilot kept citing them, what the test cannot prove, and what to check.

Perplexity Crawlers Documentation, Digested

Perplexity crawlers documentation explained: what PerplexityBot and Perplexity-User do, why robots.txt alone is not enough, and what to check at your firewall.

Perplexity Predictability and AI Citation: arXiv 2509.14436

Perplexity predictability AI citation study (arXiv 2509.14436): what it found, why it is not 'write simply', and why it should not drive your edits.

Perplexity Sonar Migration to the Agent API: What Changed

Perplexity Sonar migration to the Agent API: web search is opt-in, so unmigrated code can return ungrounded answers without warning. What to check and fix.

SAFE Long-Form Factuality and F1@K, Digested

SAFE long-form factuality (arXiv 2403.18802) explained: how F1@K pairs precision with recall against K, and why brand teams should track error rate and worst error.

Scaled Content Abuse Policy: What Google Targets

Google's scaled content abuse policy, quoted and explained: what counts as volume without value, how mass AI sites were silently deindexed, and what to check.

What Gets Cited: The 252,000-Trial GEO Factorial Study

What gets cited GEO study (arXiv 2605.25517): topical relevance and list position drove citation, formatting edits barely moved it. Design, limits, takeaways.