Which Sites Do AI Answers Cite Most?

What did the study measure?

Using its own Brand Radar tool, Ahrefs examined about 78.6 million AI responses in June 2025: 76.7 million AI Overviews, 957 thousand ChatGPT and 953.5 thousand Perplexity responses. For each engine the study gives a top-10 list of the most-mentioned domains and reports them as "mention share".

A vendor ran this study; that is, it publishes data from its own tool. We label it "Ahrefs study (vendor research), June 2025" and read it as a data point, not an independent verification.

The numbers

Label: Ahrefs study (vendor research), June 2025, single month, top-10 lists only. "Not in top 10" means the domain does not appear in the top-10 list the study gives for that engine; it does not mean zero mentions.

Limits of the numbers

• One month, one vendor. This is June 2025 data; engine behavior changes over time.
• Top 10 only. Domains outside the list do not show up in this table.
• Different studies give different results. Studies by other vendors such as Profound, Peec AI and BrightEdge report very different shares because their counting methods differ. We could not verify those studies' figures, so we do not give them here.
• No data for Turkey. We found no study of which sites are cited for Turkish-language queries. This table does not show that it applies to Turkey as is.

What does it mean?

The safe common ground across studies is this: third-party platforms such as Wikipedia, Reddit, YouTube and news sites are cited heavily in AI answers, and their overlap with Google’s classic organic top 10 is partial and varies by engine. So to appear as a source in one engine, ranking well on your own site may not be enough; the places where others write about you can be raw material for those answers.

Read this inference carefully too: a platform being cited often does not mean that being on it will carry you into the answer. The right question is not "are we on Wikipedia?" but "which sources do the engines rely on for the questions our buyers ask, and are we present in them?"

What to do

1. Build a source-gap map. Run the questions buyers ask in your sector many times, list which domains are cited, and note where competitors are mentioned and you are not.
2. Close the gaps with real publications. Expert articles, placements in trade publications, accurate presence on reference or comparison pages. That is the subject of our digital PR and citations service.
3. Be realistic about Wikipedia and Wikidata. A Wikipedia page is not something to be "bought"; it requires significant coverage in reliable independent sources and avoids promotional editing. We do not sell pages; where the entity qualifies we offer a readiness assessment and a policy-compliant Wikidata entry.
4. Measure. Run the same question set at regular intervals and watch which sources change; trust the trend, not a single answer.

The AI visibility work that Zen Media, part of our group, publishes on its own site follows the same logic: it measures whether distributed content is used as a source in answers. Details and labels are on the Results page.