of pages ChatGPT cites on “best X” queries are listicles
Ahrefs — 750 commercial prompts, 26,000+ cited sources, December 2025
AI assistants do not survey a profession before recommending someone. They retrieve documents — and overwhelmingly, documents of one particular shape. On commercial-investigation queries such as “best X” or “X vs Y”, a June 2026 analysis by AIVO found listicles present in 100% of cited answers. Ask an assistant who the best implant surgeon in Warsaw is, or which agency to hire, and what determines the answer is not professional quality. It is the document layer these systems read.
This brief summarises what the published data says about how AI systems choose whom to recommend, and what that implies for professionals who are excellent at their work but poorly represented in text.
Authority metrics do not predict citation
The most counterintuitive finding in the citation literature comes from Surfer, which analysed roughly five million citation URLs across four AI search surfaces. The authority metrics of a website — the numbers an entire SEO industry was built on — show a rank correlation with AI citation that is statistically indistinguishable from zero.
For established professionals this cuts both ways. The bad news: a distinguished career, by itself, produces no citations. The good news: a famous website is not the price of entry. The right documents simply have to exist.
What being cited is worth
The reshaping of search is measurable. Seer Interactive, tracking 3,119 search terms across 42 organisations, found that the presence of an AI Overview cuts organic click-through roughly from 1.62% to 0.61% — a decline of more than half.
An AI Overview cuts organic click-through by more than half
Average organic CTR on tracked search terms, %
Seer Interactive — 3,119 search terms across 42 organisations, September 2025
The same dataset contains the inversion that matters. Brands cited inside the AI answer saw organic click-through roughly 35% higher than uncited brands. Seer is careful to note this is a correlation rather than proven causation. Still, the direction is consistent: traffic is not disappearing uniformly, it is being redistributed toward the entities the system names.
Freshness is a selection criterion
Recency behaves less like a ranking tiebreaker and more like a filter. Ahrefs found that ChatGPT cites pages roughly 458 days newer than the baseline Google organic result for the same query, and that 79% of cited “best of” lists had been updated within the prior year. A definitive article that has not been touched in three years is not competing on merit — it is largely out of the running.
The mechanism, in plain terms
| Stage | What the system does | What it needs from you |
|---|---|---|
| Retrieval | Pulls documents matching the question | Documents that exist, in retrievable form |
| Selection | Prefers structured, specific, current sources | Profiles, data pieces, Q&A, tables, dated updates |
| Attribution | Names entities it can identify confidently | A consistent, verifiable identity across sources |
Nothing in this pipeline measures how good a professional actually is. Every stage measures how well they are documented. One further caveat is worth stating plainly: the source graphs of different assistants barely overlap. Ahrefs found that only 14% of top sources are shared across ChatGPT, Perplexity and Google’s AI Overviews. Being visible to one system says almost nothing about the others.
Methodology
- Sources
- All figures are drawn from third-party published studies, each verified against its primary source at the time of publication: Ahrefs (ChatGPT best-lists research, 750 prompts / 26,000+ sources, December 2025; freshness and source-overlap analyses), AIVO (commercial-intent citation analysis, n=138 queries, June 2026), Surfer (~5M citation URLs, updated July 2026), Seer Interactive (3,119 terms across 42 organisations, September 2025).
- Scope limits
- The Ahrefs listicle figure covers top-of-funnel commercial prompts, not all ChatGPT queries. The AIVO 100% figure applies to commercial-investigation intent specifically; the same study reports 71% for local intent, 50% transactional and 13% informational. The Seer citation advantage is a correlation, not a demonstrated causal effect.
- Updates
- This brief will be revised as new citation data is published. Corrections are handled under our corrections policy.
Developed from a structured interview with the author by the Datum Review desk. The author reviewed and approved the final text.
Disclosure: The author is the founder of Datum Review.
