Brand mentions vs backlinks for AI search

For years, SEO programmes treated the backlink as the main off-site currency: a crawlable vote, preferably do-follow, preferably from a strong domain. Answer engines have reopened an older idea in a new form — that being named, even without a link, might also matter.

The question is no longer whether mentions and links both exist. It is what the available evidence actually says about each, where that evidence is silent, and how to spend effort without pretending the studies are a ranking guarantee.

Last updated: 12 November 2026.

A brand mention is a named reference to a company or product in published content: a tool list, an example in an article, a roundup, a transcript, a discussion. It may or may not include a hyperlink. On offsiteposts.com, a mention is counted once per published appearance, not each time the name is repeated inside the same article.

A backlink is a hyperlink from another site to yours. If it is do-follow, it can pass ranking credit in the classic search sense, subject to how search engines treat it. If it is marked rel="sponsored" or rel="nofollow", Google’s documentation treats that as the recommended way to flag paid or untrusted links.

These are not opposites. A linked mention can be both a mention and a backlink. A named reference with no link is a mention only. A bare URL in a comments section is a link with very little linguistic context.

Ahrefs analysed 75,000 brands and measured Spearman correlations between a set of factors and how often those brands appeared in Google AI Overviews. The headline comparison is stark, and it is still only a correlation:

Branded web mentions: 0.664. Branded anchors (the brand name used as clickable link text): 0.527. Branded search volume: 0.392. Domain Rating: 0.326. Number of referring domains: 0.295. Number of backlinks: 0.218.

Ahrefs notes that the three strongest correlations were all off-site factors, that web mentions correlated much more strongly than backlinks in this dataset, and — repeatedly — that correlation is not causation. All of the relationships sat in the moderate-to-very-weak range on the Spearman scale. About 26% of brands had zero AI Overview mentions. Brands in the top quartile for web mentions averaged 169 AI Overview mentions, more than ten times the next quartile (which averaged 14). The bottom two quartiles barely registered. Source: An Analysis of AI Overview Brand Visibility Factors, https://ahrefs.com/blog/ai-overview-brand-correlation/

Read that carefully. It does not say “mentions cause citations.” It says that, among the factors Ahrefs chose to measure, the brand-name-on-the-web factor moved together with AI Overview visibility more tightly than raw link counts did.

A later Ahrefs study widened the lens to ChatGPT, Google AI Mode and AI Overviews. In that analysis, YouTube mentions showed the strongest correlation with AI visibility (around 0.737), slightly ahead of YouTube mention impressions, and branded web mentions remained high (about 0.66 to 0.71). ChatGPT showed weaker correlations with some classic authority metrics than AI Mode did. The three assistants still tended to mention many of the same brands. The same causation warning applies. Source: Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews, https://ahrefs.com/blog/ai-brand-visibility-correlations/

A narrower Ahrefs comparison of the top 50 domains in Brand Radar found a strong mentions-to-visibility correlation for Google AI Overviews (0.65), a weak one for Perplexity (0.30), and a very weak one for ChatGPT (0.15). If you only remember the 0.664 figure, you will overfit Google Overviews and misread ChatGPT. Source: Google Seems More Biased Towards Big Brands Than ChatGPT and Perplexity, https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/

Slate Research’s 1 October 2026 study of 100 brand-name searches found that 93 brands showed an AI Overview for their own name, and that 71% of 652 unique citations were pages the brand does not own. Only 29% pointed to a domain the brand owns. Source: AI Overviews on brand searches, https://slatehq.com/data-studies/ai-overviews-brand-searches

That finding is about citations on branded searches, not about “best software” queries. It still undercuts the idea that a perfect homepage is enough. When people search a brand, the overview often points elsewhere — Wikipedia, publishers, reviews, videos, social profiles, and other third-party pages. Ahrefs’ October 2026 write-up also found AI Overviews appearing on a large majority of US desktop branded searches in its sample (82.91% on 29 September, up from 61.3% in July). Source: AI Overviews Currently Appear for 83% of Branded Searches, https://ahrefs.com/blog/ai-overviews-on-branded-searches/

If an overview is going to talk about you, it needs sources. Many of those sources will not be you.

It would be a mistake to read the 0.218 backlink correlation as “stop caring about links.”

First, that number is about AI Overview brand visibility in one Ahrefs study, not about classic Google rankings. Ahrefs has separately argued that links still correlate with rankings even if they “matter less” than they once did; the company’s January 2025 analysis of the top 1 million US keywords is the piece we already cite on this site for that point. Source: Links matter less, but still matter, https://ahrefs.com/blog/links-matter-less-but-still-matter

Second, branded anchors — a link signal that also contains the brand name — correlated much more strongly (0.527) than raw backlink count in the AI Overview study. A nameless link and a named, contextual citation are not the same object.

Third, Domain Rating and referring domains still showed positive, if weaker, correlations. They were not zero.

Fourth, Google Search is still a system in which links are a known signal, and Google’s spam policies still treat paid links that pass ranking credit as link spam. Google recommends marking paid or sponsored links with rel="sponsored" or rel="nofollow". Do-follow paid links can conflict with those guidelines and carry risk. That policy has not been retired because AI Overviews exist. Source: Google spam policies, link spam, https://developers.google.com/search/docs/essentials/spam-policies#link-spam

A serious programme can care about mentions for AI visibility and still treat paid do-follow links as a separate, disclosed, risk-aware decision made with an SEO lead.

We do not have an official Google explanation that “mentions are weighted at X.” A cautious interpretation, consistent with how these systems are described, is more mechanical than mystical.

Answer engines retrieve text and then generate an answer. A passage that names a product in a relevant sentence is something a model can quote, paraphrase or place in a list. A link with no surrounding language about the brand gives the model less to say. Query fan-out in Google’s AI features, and query rewriting in ChatGPT search, both send the system looking for pages that answer sub-questions. Pages that discuss brands by name in those subtopics are simply more usable as sources.

That is an interpretation, not a leaked ranking factor. It also explains why low-quality, repetitive, unsupportable mentions are a poor bet: they are less likely to be useful passages, and they can create a public record you will not want retrieved.

Do not drop links because a correlation favoured mentions. The studies do not say links have no relationship with AI visibility, and they say almost nothing, in these particular papers, about the return on your next twenty links versus your next twenty mentions.

Do not buy “mention volume” as a vanity metric. The Ahrefs quartile gap is striking, but the brands at the top of the mentions distribution are often already famous. You cannot infer that buying a handful of sponsored name-drops will move you into that quartile.

Do care about named, accurate, third-party text. Being described correctly on sites your buyers already read is useful for humans, and it is consistent with the off-site pattern in the research. If the placement is paid, label it as sponsored. If the claim is not supportable, do not publish it.

Treat YouTube and other transcripts as part of the mention picture, not a side project. The later Ahrefs work put YouTube mentions at the top of the correlation table. That is evidence to take video and transcripts seriously, not a brief to buy fake shout-outs.

Measure mentions and citations separately from rankings and traffic. A new named article can show up in an AI answer without moving a classic keyword, and the reverse is also possible. Keep the scorecards apart.

offsiteposts.com sells one-time packages of third-party posts, named brand mentions and do-follow links. Mentions are counted once per published appearance. Links, when included, are do-follow; we state Google’s paid-link policy plainly and do not claim blanket compliance or ranking gains. AI-assistant mentions, traffic and rankings are not guaranteed.

That mix exists because a post can carry a mention, a mention can exist without a link, and a link is a different deliverable with a different risk profile. The research above is one reason mentions are in the package at all. It is not a case study, and it is not a promise.

If you want the ChatGPT-specific retrieval details, link to /blog/how-to-get-your-brand-mentioned-in-chatgpt. If you want Google’s official AI Overview eligibility rules, link to /blog/how-google-ai-overviews-choose-sources.