Original research on how AI searches, what it cites, and which brands it recommends. Read the findings, inspect the method, use the data.
Latest studyStudy 013 ·
We asked ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews 200 questions that call for a number, then followed every statistic they quoted back to its source. Four in ten reached a study that says the same thing.
40%of the statistics AI quotes lead back to a study that says the same thing. The rest changed, can’t be seen or lead nowhere.
Read the studyThe results people ask us about most, each drawn from its study’s own data.
| Day | Still cited |
|---|---|
| First seen | 100.0% |
| Day 1 | 35.1% |
| Day 2 | 33.9% |
| Day 3 | 32.9% |
| Day 5 | 30.7% |
| Day 7 | 28.8% |
| Day 10 | 25.8% |
| Day 14 | 22.8% |
73.5%of 108,650 cited URLs appeared on a single day only. Fewer than half are still cited the next day.
Study 009The half-life of AI citations9.4×Person schema is that much more common on AI-cited pages than on the web.
Study 008The anatomy of an AI citation13.3%of AI-cited domains have the file. Neither citation test shows what adding it would cause.
Study 005The llms.txt effectThirteen studies, newest first. Each has its own sample, method and limits.
13 studies
We asked ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews 200 questions that call for a number, then followed every statistic they quoted back to its source. Four in ten reached a study that says the same thing.
We decoded 3,476 structured Search Console queries that revealed buyer personas, competitor sets, markets and evaluation rules used in AI visibility tracking.
An exploratory benchmark separating AI citation from recommendation and measuring how recommendation rates change around a brand page's first observed citation.
Live randomized experiment measuring whether Markdown changes AI crawler coverage, retrieval demand, and transfer efficiency.
Longitudinal citation persistence and visibility volatility benchmark behind the Citation Decay study.
Benchmark of the content, schema, authorship, and structural traits shared by AI-cited pages.
Benchmark of websites embedding hidden prompts that attempt to influence AI chatbot citation and memory behavior.
Page-type citation efficiency benchmark cross-referencing AI citations with AI crawler visits.
llms.txt adoption and citation-impact benchmark across AI-cited domains.
Cross-model agreement benchmark showing how often AI systems recommend the same brands for the same prompts.
AI crawler behavior benchmark covering traffic share, entry pages, session depth, and revisit patterns.
Prompt-to-search-query rewrite benchmark showing how AI systems transform user questions before retrieving sources.
Public AI Citation Index covering 80.8M citation appearances, domain concentration, source categories, prompt intent, and explicit coverage limits.
Public snapshots, source code and a maintained resource index. Every dataset keeps its own licence and coverage limits.
Monthly public snapshots of the top 500 brands in global AI visibility rankings.
Open the repository Resource index · CC0Papers, tools, datasets and crawler guidance for generative engine optimisation.
Explore the index Public citation index80.8M citation appearances, with definitions, downloads and explicit coverage limits.
Inspect the indexBring your expertise and audience. We can help turn a useful question about AI search into a sourced study.
Not sure where to start? Take a guided tour.
AI Site Grader
AI Query Scanner
llms.txt Generator