of answers citing their page never named them
The citation and brand mention did not always travel together. A citation-only metric can therefore miss an important distinction.
The second problem
Cited, and still not recommended.
There is a kind of losing that looks exactly like winning. Your page gets cited. The assistant quotes your research, borrows your framing, uses your numbers — and then recommends somebody else.
Ahrefs published 34 self-promotional pages across five domains and analyzed 9,886 answers from ChatGPT, Gemini, Perplexity and Copilot between February 7 and May 31, 2026. In the Evolve conference condition, 43% of answers that cited a conference-promoting page did not mention Ahrefs Evolve. That is an observed result from this specific experiment; it does not by itself show why a different brand was or was not recommended.
Diagnostic split
One answer can produce two different observations.
Generated answerinspect the exact answer state
Observation ACitation
Was a page used or linked as a source?
Observation BRecommendation
Which brand or option did the answer actually name?
of an established brand’s new mentions came from third-party pages
In the established-brand condition reported by Ahrefs, 6% came from the brand’s own content. This is a study result, not a universal weighting rule.
of eligible days a cited page actually appeared
About one day in three in this experiment. The result shows volatility in this sample and cautions against treating one observation as stable.
mention rate on the narrow query — against 15.8% on the broad one
The narrow and broad query conditions produced very different mention rates in this experiment; that supports testing query specificity rather than assuming one universal effect.
Ahrefs, Self-Promotional Content and AI Search, 2026. 34 pages, 5 domains, 9,886 answers, four engines, February–May 2026.
The mechanism
Why a quote and a recommendation come apart.
A citation and a recommendation are different observable outcomes. The Ahrefs experiment demonstrates that a promoted page can be cited without the promoted brand being mentioned. It does not prove a single retrieval mechanism or universal cause. To diagnose a real site, inspect the cited passage, the brands named in the answer and the other sources present before assigning a cause.
Third-party evidence can matter, but the study is not a universal weighting formula. In Ahrefs’ established-brand condition, 94% of new mentions came from third-party pages and 6% from the brand’s own content. That makes third-party representation worth measuring for similar questions, while leaving the exact causal weight unproven.
The tell
- Citation share rising while pipeline is flat. A useful diagnostic signal. The next step is to inspect whether citation and recommendation actually diverge in the answer set being measured.
- You appear in answers about the category but not in answers about the choice. If one class of questions names the brand and another does not, record the difference before assuming the cause.
- Your comparison pages get cited most. Inspect the exact cited passages and adjacent brand mentions rather than assuming a comparison page caused the recommendation outcome.
The fix
Five moves, in order.
None of them is “write more.” Three of them are edits to pages you already have.
- 01
Separate the two numbers
Score every prompt twice — cited, and recommended. Some tracking approaches emphasise citation without separately recording recommendation. Where the two measures diverge, you have identified a gap to investigate rather than a diagnosis by itself.
- 02
Read the cited passage, not the page
Open the exact paragraph the assistant quoted. Compare the cited passage with the surrounding page and the other sources in the answer. Record what is actually present before changing the copy.
- 03
Collapse the distance
If the page separates a core category claim from the evidence or qualification needed to understand it, make the passage clearer and self-contained. Do not manufacture superlatives or unsupported claims merely to make a sentence more quotable.
- 04
Audit competitor adjacency
Find pages where your brand and competitors are discussed together and compare them with the answers being observed. Treat adjacency as something to test, not a proven causal rule.
- 05
Move the claim off your own site
In Ahrefs’ established-brand condition, 94% of new mentions came from third-party pages and 6% from the brand’s own content. Use that as a reason to audit third-party representation, not as a guaranteed formula for recommendations.
Questions
Cited but not recommended, answered.
Can an AI answer cite my page without naming my brand?
Yes. In Ahrefs’ 2026 self-promotional experiment, 43% of answers that cited an Evolve conference-promoting page did not mention Evolve. The experiment covered 34 pages on five domains and 9,886 answers across ChatGPT, Gemini, Perplexity and Copilot from February 7 through May 31, 2026.
How do I know if this is happening to me?
For a fixed set of questions, record citation and brand recommendation or mention as separate observations. Where they diverge, inspect the exact cited passage, the other sources used and the brands named before deciding what to change.
Does adding more content fix it?
There is no universal evidence that simply adding more pages resolves a citation-versus-recommendation gap. Diagnose the answer set and source passages first; then make targeted content or corroboration changes that can be measured.
How volatile were citations in the Ahrefs experiment?
The promoted pages appeared on roughly one-third of eligible days between their first and last appearance in the experiment. That illustrates volatility in this sample; it should not be converted into a universal probability or mandatory measurement cadence.
BackgroundGEO vs SEOThe nine decisions that differ, and why ranking is weak evidence of retrievability.
ReferenceRecommendation rate, definedTwelve terms including the two numbers this page is about.
PracticalWhat this work costsHow scope, measurement and delivery requirements affect a proposal.
Want to know whether this is happening to you?
Use the heuristic readiness scorecard, or send a brief if you want the current automated checks considered against your site. Any follow-up scope and timing are confirmed from the evidence and access available.