ChatGPT citation statistics

Headline numbers

What ChatGPT cites

Reported analysis of a large citation corpus finds ChatGPT's top citations dominated by encyclopedic and reference-style content. Wikipedia sits foremost among them, at a rate substantially higher than the other major engines tracked. This figure is Partial : a vendor reported it, from a stated 680-million citation corpus that isn't independently public. See most-cited domains in AI search for the full cross-platform comparison.

Ranking overlap

At roughly 12%, ranking on Google correlates only weakly with being cited by ChatGPT. That's a sharp contrast to AI Overviews, where the same study found overlap closer to 38%. If you're optimizing specifically for ChatGPT citation, classic Google ranking is a much less reliable lever than it is for Google's own AI surface.

Referral traffic & conversion

ChatGPT referral traffic is reported to convert at meaningfully higher rates than average organic traffic, across several vendor studies. But the specific multiple varies widely by source and definition. See the provenance audit's conversion-spread section for why we won't restate a single number here without qualification.

What's more reliably reported: ChatGPT's absolute referral volume stays a small fraction of total site traffic for most sites, even where conversion quality is higher.

How ChatGPT Search actually retrieves

ChatGPT Search performs live retrieval at query time, rather than answering purely from training data. It has historically drawn in part on a Bing-derived index, alongside OpenAI's own crawl infrastructure. The exact current architecture isn't fully disclosed publicly, and has likely evolved since ChatGPT Search's initial launch.

GPTBot vs. OAI-SearchBot

This is the single most consequential technical distinction for ChatGPT visibility. GPTBot crawls to improve future model training. It has no bearing on today's citations. OAI-SearchBot performs the live fetch that actually produces a citation in a ChatGPT Search answer.

A site that blocks OAI-SearchBot while allowing GPTBot removes itself from ChatGPT Search citations, while still contributing to future model training. That's likely the opposite of most site owners' intent. Full detail in the dedicated crawler guide.

A brief timeline

ChatGPT's web-access capability began as a limited browsing feature. It evolved from there into a dedicated Search product, with its own retrieval infrastructure and citation UI. This lineage matters when reading older statistics or screenshots circulating online.

A citation pattern documented under an earlier browsing feature may not describe current ChatGPT Search behavior. This page's figures describe the current-generation product as of its stated dates, not the feature's entire history.

How ChatGPT differs from Perplexity, in one table

The two get compared often. Here's a direct side-by-side, for anyone deciding where to focus effort:

See the dedicated Perplexity statistics page for the full detail behind the right-hand column.

What this means for your own content

If ChatGPT is a priority surface, encyclopedic clarity matters more than it does elsewhere. Write like a reference entry, not a sales page. State the plain fact first, in neutral language, then add supporting detail underneath it.

Don't rely on Google rank as your main signal here. A page can rank poorly on Google and still get cited by ChatGPT, and the reverse happens too. Check your server logs for OAI-SearchBot specifically, using the crawler guide, rather than assuming Search Console data tells the same story.

Verification status

The ranking-overlap figure (§3) traces to Ahrefs with a disclosed sample — Traceable . The source-skew figure (§2) is vendor-reported from a non-public corpus — Partial . Conversion claims are treated as Broken chain per the provenance audit until a reconciled, like-for-like figure exists.