What did Ahrefs measure, and what did it find?
Ahrefs: in a study published on 2025-05-26, Louise Linehan (data pulled by Xibeijia Guan) looked at which factors of 75,000 brands correlate with how often they are mentioned in Google AI Overviews. About 26% of the brands were never mentioned in an AI Overview; the analysis used the other 74% or so. It kept only domains with a DR above 40, took each domain's highest-volume keyword with at least 800 searches a month, and used Spearman correlation.
| Factor | Correlation |
|---|---|
| Branded web mentions (other pages naming your brand) | 0.664 |
| Branded anchors (link text is the brand name) | 0.527 |
| Branded search volume | 0.392 |
| Domain Rating (DR) | 0.326 |
| Referring domains | 0.295 |
| Branded traffic | 0.274 |
| Backlinks | 0.218 |
Numbers are from the table in the Ahrefs article, opened and checked on 2026-11-05 (the article also lists ad traffic, URL rating, number of pages and more, not shown here). Reading: a brand's web mentions track its AI Overview mentions far more closely than its backlink count does (0.664 against 0.218); backlinks, referring domains and DR are still positively correlated, not zero.
What did Seer Interactive measure?
Seer Interactive: Christina Blake and Nick Haigler published on 2025-01-07. They took 300,000+ finance and SaaS keywords, narrowed about 600,000 People Also Ask questions to 10,000, asked them through OpenAI's GPT-4o API, and looked at what correlates with the brands that get mentioned.
| Factor | What they found |
|---|---|
| Google page-one ranking | Correlation of about 0.65 with the brands the model mentions |
| Bing ranking | About 0.5 to 0.6, less influence than Google |
| Backlinks | Weak or neutral; the article gives no number |
| Images, video and other content formats | Weak or neutral; the article gives no number |
The authors say so themselves: this is phase one, it only finds correlations and does not prove cause; PAA questions stand in for real queries; models often name a product rather than a brand, which makes matching mentions back to domains hard; the article does not report significance.
Taken together, can we say brand mentions matter more than backlinks?
What we can say: in these two samples, “mentioned elsewhere” and “ranks well in search” relate more strongly to being mentioned by AI than the backlink count does. What we cannot say: that getting more mentions will make AI mention you. Three reasons, all our inference:
- Correlation is not cause: big brands are mentioned everywhere and also by AI, so brand size may push both up. Ahrefs itself stresses that correlation is not causation.
- Sample bias: Ahrefs kept only domains with a DR above 40, Seer only tested finance and SaaS; small sites and other industries may differ.
- The studies measure different things: Ahrefs looks at Google AI Overviews, Seer at GPT-4o API answers, so they cannot simply be merged into one conclusion.
Note too that mentions and backlinks are not exclusive: being mentioned by another site with a link gives both signals at once. In Seer, backlinks were weak and Google ranking strong; ranking is influenced by backlinks, but Seer did not test that link, so it is our inference.
What can we see in our own evaluation?
On 2026-10-26 we evaluated GammaSlide: the AI read 28 cited pages, and 0 of them mention GammaSlide. This does not compare mentions with backlinks. It only shows that in this evaluation none of the pages the AI cited mentioned this product, so it is no surprise the answers recommended others. It is one site and one evaluation, not other industries.
What does this change about what you do?
1. Do not build backlinks for their own sake
In both studies backlinks are not the strongest signal, and there is no evidence that buying links raises AI recommendations.
2. First give others a reason to mention you
Real reviews, rankings, forum discussions, and your data being cited. How such mentions come about: see What signals does AI actually look at when recommending brands? and Why does AI like to cite Reddit?.
3. Keep mentions real
Using several accounts, buying reviews, or posting the same text in bulk breaks platform rules, and no study shows it works.
4. Keep doing the basics
Let AI crawlers in, put the body text in the HTML, state prices and features clearly. These are separate from mentions and backlinks, and all are needed.
5. Re-test with the same questions after a few weeks
Watch how often you appear in answers and whether the cited pages change; do not judge from a single answer.
FAQ
Do brand mentions really matter more than backlinks for AI search?
In Ahrefs' sample, a brand's web mentions correlate 0.664 with AI Overview visibility and backlinks 0.218, so mentions are stronger. But this is correlation, not cause, and the sample only contains domains with a DR above 40.
Are backlinks useless for AI search?
No. In Ahrefs, backlinks, referring domains and DR are all positively correlated, just weaker than brand mentions. Seer says backlinks are weak or neutral for mentions in ChatGPT-style answers, but Google ranking correlates more strongly, and ranking is influenced by backlinks; that last link is our inference.
What counts as a brand mention?
In the Ahrefs study it means other pages that contain your brand name, with or without a link. Links whose text is the brand name are branded anchors, a separate factor with a correlation of 0.527.
Why not cite other studies?
We only write about studies whose original we opened and whose author and date we could check. Other figures retold in third-party articles we could not match to an original, so we leave them out.
Can you pay for mentions?
We have seen no study showing that it works. Buying reviews, posting in bulk and using several accounts all break platform rules and damage your credibility once spotted.
Sources
- Ahrefs · An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) （Louise Linehan, published 2025-05-26, checked 2026-11）
- Seer Interactive · STUDY: What Drives Brand Mentions in AI Answers? （Christina Blake and Nick Haigler, published 2025-01-07, checked 2026-11）
- kagentx evaluations: the data on the case study pages; each example gives its date.