AI engines
How each AI engine finds and cites sources
One reference page per surface we track: what it is, what its vendor actually documents about attribution, what nobody outside the vendor knows, and what you can do about it.
Each page keeps three things apart that are usually mixed together — documented vendor behaviour with a link to the document it was read from, our own measurement of that surface, and the parts that are genuinely undocumented. Where a page says we do not know, that is the finding, not a gap we filled in.
What OpenAI documents about how ChatGPT retrieves and attributes sources, what stays undocumented, and what GetCited.me measures on it.
Read the pageWhat Google documents about Gemini's search grounding and citations, what stays undocumented, and what GetCited.me measures on it.
Read the pageWhat Perplexity documents about the crawlers behind its links, what stays undocumented, and what GetCited.me measures on it.
Read the pageWhat Anthropic documents about when Claude searches and how it cites, what stays undocumented, and what GetCited.me measures on it.
Read the pageGoogle's own guidance says there are no AI-specific requirements. What that means, what is not documented, and what GetCited.me measures on the surface.
Read the pageHow to read these pages
Every statement about a vendor's behaviour links to that vendor's own documentation or to the relevant standard. We publish no statistics about citation share, because we have no primary source for one. These systems change without notice, so each page carries the date its sources were last read.