Brand mentions are the new backlinks for AI search

TL;DR: Ahrefs analyzed 75,000 brands and found that unlinked brand mentions correlate with AI Overview citations at 0.664, while backlinks correlate at just 0.218, a gap of roughly 3 to 1. Backlinks still matter for traditional Google rankings, but for getting cited by ChatGPT, AI Mode, and AI Overviews, the volume and spread of your brand name across the web now outweighs how many sites link to you.

---

Table of contents

• The study that changed the conversation
• Why mentions outperform links for AI citations
• What counts as a brand mention
• Brand mentions vs backlinks: a side-by-side comparison
• Where mentions carry the most weight
• How AI models actually use brand mentions
• Building a mention strategy: the decision flow
• How to earn unlinked mentions on purpose
• How to track brand mentions
• Do backlinks still matter at all
• Common mistakes teams make chasing mentions
• A 90-day mention-building plan
• Frequently asked questions
• Key takeaways

The study that changed the conversation

On May 26, 2025, Ahrefs published an analysis of 75,000 brands, filtered to domains with a Domain Rating above 40 and a top keyword search volume of at least 800 a month (Ahrefs). Authors Louise Linehan and Xibeijia Guan ran a Spearman correlation between a long list of brand and site metrics and how often each brand showed up in Google AI Overviews.

The result reordered the priority list for anyone doing AI search work. Branded web mentions correlated at 0.664. Branded anchor text came in at 0.527. Branded search volume sat at 0.392. Backlink count, the metric an entire industry built its budgets around, landed at 0.218 (Ahrefs).

Ahrefs followed up on December 12, 2025 with a wider study across ChatGPT, AI Mode, and AI Overviews, again using Ahrefs Brand Radar data (Ahrefs). Branded web mentions held steady, correlating at 0.656 to 0.709 depending on the surface. YouTube mentions came out on top across all three, at roughly 0.737. Ryan Law, Ahrefs’ Director of Content Marketing, put it plainly: “The strongest correlation… was branded web mentions. If your brand is mentioned in a ton of different places on the web, that correlates very highly with your brand being mentioned in lots of AI conversations as well.” Ahrefs CMO Tim Soulo summarized the shift in one line: “Citations [are] the new backlinks for the AI search era.”

The magnitude of the gap between top and bottom performers is the part worth sitting with. Brands in the top 25% for web mentions averaged 169 AI Overview mentions. Brands in the 50th to 75th percentile averaged 14, and the bottom half averaged 0 to 3 (Ahrefs). Twenty-six percent of the brands in the study had zero AI Overview mentions at all. That is not a small optimization gap. That is the difference between existing in AI answers and not existing in them.

Why mentions outperform links for AI citations

A backlink is a navigation instruction. It tells a crawler and a human reader “go here for more.” A brand mention, especially an unlinked one, is a different kind of signal: it tells a language model that independent sources talk about this entity, in this context, without being asked to.

Large language models do not crawl the web in real time the way Googlebot does. They are trained on, and in retrieval-augmented systems supplemented by, huge volumes of text where entities appear and reappear across unrelated sources. When your brand name shows up in a product roundup on one site, a Reddit thread on another, a YouTube transcript on a third, and a trade publication on a fourth, the model is accumulating evidence that this entity is real, discussed, and associated with a specific topic. That is closer to how citation networks work in academic research than how PageRank works in search.

Reddit, in particular, is heavily over-represented as a source in ChatGPT and Perplexity answers, according to multiple independent visibility trackers cited in coverage of the Ahrefs data (CiteFlow). A single Reddit thread that mentions your product by name, with no link at all, can carry more weight in an AI answer than a backlink from a mid-authority blog.

This does not mean links stopped mattering for Google’s classic organic rankings. Domain Rating still correlated at 0.326 and referring domains at 0.295 in the same Ahrefs study, both higher than raw backlink count. Link-based authority signals have not disappeared. They have been demoted, relative to mentions, specifically for the AI citation layer sitting on top of search.

What counts as a brand mention

A brand mention is any instance of your company, product, or founder name appearing in text, audio, or video content that a language model can ingest, whether or not it links back to you. That includes:

• Your brand name in a comparison article, even buried in a table with nine competitors
• A Reddit or Quora answer where a user names your product by name
• A YouTube video transcript that says your product name out loud, with no link in the description
• A trade publication naming you as an example in an industry roundup
• A press mention, a podcast transcript, an analyst report, a Wikipedia reference
• Your name in a G2, Capterra, or Trustpilot review, positive or negative

The common thread is that none of these require a hyperlink to count. Ahrefs’ own methodology measured “branded web mentions” separately from “branded anchor text,” and the unlinked mention metric was the stronger predictor of the two (Ahrefs).

Wikipedia mentions show this pattern at real scale. Profound’s analysis of roughly 730,000 ChatGPT conversations from October through December 2025 found Wikipedia referenced in about one of every six conversations that included citations, accounting for 5% of all citations logged, ahead of named news outlets like Reuters (Profound). A single sourced line about your company on a Wikipedia page carries the kind of independent, third-party weight this whole strategy is built on.

Brand mentions vs backlinks: a side-by-side comparison

Factor	Backlinks	Unlinked brand mentions
Correlation with AI Overview citation (Ahrefs, 75K brands)	0.218	0.664
Requires the other site’s cooperation to link	Yes, often gated by outreach	No, a name-drop is enough
Helps classic Google organic rankings	✓ Yes, still a core ranking factor	✗ Indirect at best, not a documented Google ranking signal
Helps AI Overview / ChatGPT / Perplexity citations	Weak signal (0.218 to 0.295 range)	✓ Strongest single-site signal measured (0.656 to 0.737)
Scalable through PR and community engagement	✗ Slower, requires editorial buy-in per link	✓ Achievable at volume through reviews, forums, comparisons
Trackable with standard tools	✓ Ahrefs, Semrush backlink reports	✓ Google Alerts, Ahrefs Brand Radar, mention-tracking tools
Risk of manipulation flags if done at scale	✓ Yes, link schemes are a known penalty vector	✗ Harder to game, since mentions come from independent third parties
Cost per unit at scale	Higher, outreach-intensive	Lower, often earned through PR, reviews, and community presence

Read this table as a shift in where to spend time, not a reason to stop building links. A durable AI-visibility strategy needs both, but if your budget is capped, the data says mentions now carry more weight for the fastest-growing traffic source in search.

Where mentions carry the most weight

The December 2025 Ahrefs follow-up ranked mention sources by correlation strength, and the order surprised a lot of SEOs who had spent a decade optimizing for link-worthy content instead of mention-worthy presence.

YouTube led at roughly 0.737 across ChatGPT, AI Mode, and AI Overviews alike (Ahrefs). That is a stronger signal than web mentions overall. A video transcript that says your product name, reviewed favorably or even neutrally, appears to carry disproportionate weight, likely because YouTube’s scale and transcript indexing give language models an unusually clean, high-volume source of spoken brand references.

Reddit, Quora, review sites, and established trade publications rounded out the list of high-value mention surfaces, per analysis of the same dataset (CiteFlow). None of these are traditional link-building targets. Most SEO teams do not have a Reddit strategy, a YouTube strategy, or a review-site strategy sitting next to their backlink outreach plan. That gap is now the opportunity.

Profound’s study of roughly 700,000 social-platform citations inside ChatGPT responses, covering October through December 2025, found that 99% of Reddit citations point to individual discussion threads rather than subreddit pages or user profiles, and 85% of YouTube citations point to a specific video instead of a channel page (Profound). X, despite sitting at the center of real-time news and commentary, barely registered in the same dataset. Presence on one platform does not transfer to another, which is why a mention strategy has to name specific platforms instead of treating “social” as one line item.

How AI models actually use brand mentions

The mechanism is entity association, not link equity. When ChatGPT, Perplexity, or Google’s AI Mode generate an answer, they are pulling from a mix of training data and, for retrieval-augmented systems, live search results. In both cases, the model is trying to answer a question like “what companies solve this problem” or “what do people say about this product.”

If your brand appears consistently, across independent sources, in the context of a specific problem or category, the model has repeated evidence to draw on when it decides which names to surface. A single glowing case study on your own site does not do this. Ten independent, unprompted mentions across Reddit threads, comparison articles, and review platforms do.

Much of that surfacing happens in content nobody explicitly requested. Profound’s June 2026 analysis of 50,000 prompts across seven industries found that 47% of AI response content is unsolicited editorial content, comparisons, opinions, and recommendations the user never asked for (Profound). That editorial layer is exactly where a competing brand name gets named or skipped. The model is filling space with judgment calls, and it can only render a judgment about names it already has evidence for.

This is why the Ahrefs authors were careful to caveat their own findings: “Correlation isn’t causation,” as the December study put it (Ahrefs). Brands with strong mention profiles tend to also be brands with strong products, active communities, and real market share. You cannot manufacture mentions your way into a market position you do not otherwise have. What you can do is make sure the position you do have gets reflected in the places language models draw from.

flowchart TD
    A["Someone searches a question<br/>in ChatGPT / Perplexity / AI Overviews"] --> B{"Does the model have<br/>training data or retrieved<br/>results mentioning your brand?"}
    B -->|No mentions found| C["Your brand is never surfaced,<br/>regardless of your backlink profile"]
    B -->|Few, isolated mentions| D["Low-confidence inclusion,<br/>often dropped in favor of<br/>more-mentioned competitors"]
    B -->|Repeated mentions across<br/>independent sources| E["High-confidence citation,<br/>brand named directly in the answer"]
    E --> F["Reader sees your brand name<br/>with no click required"]
    D --> G["Reader sees a competitor instead"]
    C --> G

Building a mention strategy: the decision flow

Most teams do not need a new department to act on this. They need to redirect a slice of the effort currently spent chasing backlinks toward earning mentions in places language models actually read.

Start by auditing where your brand already appears, unlinked, using the tracking methods in the next section. Then decide, category by category, whether you are underrepresented relative to competitors. If a competitor shows up in twenty Reddit threads about your product category and you show up in two, that gap is your priority list, not a hypothetical backlink target from a site that has never heard of you.

How to earn unlinked mentions on purpose

None of this requires manipulation. It requires being present, by name, in the conversations that were already happening.

Answer real questions where they are asked. Reddit and Quora threads about your product category are where a large share of AI-cited mentions originate. Answer honestly, name your product only when it is relevant, and expect no link. The mention is the point.

Get reviewed on the platforms people already trust. G2, Capterra, Trustpilot, and category-specific review sites feed both human buyers and AI answer engines. A steady stream of real reviews, not incentivized five-star spam, builds the kind of repeated, independent mention pattern the Ahrefs data points to.

Pursue inclusion in comparison and roundup content, not just guest posts. A guest post with a link is one mention. Getting named as an option in ten independent “best X for Y” articles, even without a link in half of them, is ten mentions across ten domains.

Show up on YouTube, even indirectly. You do not need your own channel. Being mentioned by name in a product review, a tutorial, or an industry commentary video puts your brand in the transcript data that correlated strongest with AI citations in the December 2025 study.

Pitch trade press and analysts on substance, not press releases. A named mention in a trade publication, even a brief one, sits in the same tier of trusted, independent sources the models weight heavily.

Do not neglect your own site’s clarity. None of this replaces having a clear, crawlable page that states what your product does in plain language. Mentions get your name into the conversation; your own content has to hold up once someone or something looks you up.

How to track brand mentions

You cannot manage what you do not measure, and mention tracking looks different from backlink tracking because most mentions have no URL pointing back to you to crawl.

Google Alerts remains a free, blunt-instrument starting point for catching new web mentions of your exact brand name, though it misses a large share of what gets discussed on platforms like Reddit and in video transcripts. Ahrefs’ Brand Radar, the tool behind both studies cited above, tracks branded mentions specifically across AI answer engines, which is the closest thing to measuring the exact signal this piece covers. Dedicated mention-monitoring tools built for PR and social listening also cover Reddit, forums, and review sites more thoroughly than a generic web alert.

Whatever tool you use, track three things over time: total mention volume, the diversity of domains mentioning you, and the ratio of linked to unlinked mentions. A rising unlinked-mention count on a widening set of independent domains is the pattern the Ahrefs correlation data says to chase.

Do backlinks still matter at all

Yes, for a narrower job than they used to have. Backlinks remain a documented, direct input into Google’s classic organic ranking algorithm, and Domain Rating (0.326) and referring domains (0.295) still outcorrelated raw backlink count in the Ahrefs AI study, suggesting link-based domain authority carries some residual signal even for AI citation (Ahrefs). If you rely on organic Google traffic for revenue today, do not stop building links.

What changes is the marginal allocation of new effort. If your next dollar or hour can go toward one more outreach-driven backlink or toward a genuine, name-earning presence in ten Reddit threads, three review platforms, and one YouTube mention, the correlation data says the second path now pulls more weight for the fastest-growing share of search traffic.

Common mistakes teams make chasing mentions

Buying fake reviews or astroturfing forum threads is the fastest way to burn the exact trust signal this strategy depends on. Independent, organic mentions are valuable precisely because they are independent. Manufactured ones are detectable by both platform moderators and, increasingly, by the models themselves, which are trained to discount coordinated inauthentic patterns.

Treating mention-building as a one-time sprint instead of an ongoing presence is another common failure. The brands with 169 AI Overview mentions in the top quartile of the Ahrefs study did not get there with a single PR push. They accumulated mentions continuously, across dozens of independent domains, over years.

Ignoring negative mentions is a third mistake. A model weighing repeated, independent mentions does not filter for sentiment the way a human reader instinctively does. Addressing criticism publicly, on the same platforms where it appears, turns a liability into another data point that your brand is actively discussed and responsive.

A 90-day mention-building plan

Days 1 through 30: run a mention audit against your top three competitors using Ahrefs Brand Radar or an equivalent tool, and identify the platforms where you are most underrepresented. Set up Google Alerts and a dedicated mention tracker for your brand name and common misspellings.

Days 31 through 60: assign someone to answer relevant Reddit and Quora threads honestly, twice a week, in your product category. Reach out to three review platforms your buyers actually use and request verified listings, not incentivized reviews. Pitch two trade publications with a substantive angle, not a press release.

Days 61 through 90: track whether your unlinked mention count moved, and on how many distinct domains. Cross-check against whether your brand starts appearing, even occasionally, in AI Overview or ChatGPT answers for your category’s core queries. That is the metric that actually validates the effort, more than mention count alone.

Frequently asked questions

What is a brand mention in SEO and AI search?

A brand mention is any instance of your company or product name appearing in content, whether or not it links back to your site. It can be a Reddit comment, a YouTube transcript, a review, or a line in a trade article. AI answer engines treat these as entity signals, evidence that independent sources discuss your brand, which is distinct from the navigational role a backlink plays.

Do unlinked mentions actually help Google rankings, or only AI citations?

The Ahrefs data measured correlation with AI Overview citations specifically, not classic organic rankings. Google has not confirmed unlinked mentions as a direct ranking factor for traditional search. Treat mentions as an AI-citation strategy first, and a possible secondary signal for organic visibility, not a proven replacement for links in classic SEO.

How strong is the correlation between brand mentions and AI citations?

Ahrefs found branded web mentions correlated with AI Overview citation at 0.664 across 75,000 brands, compared to 0.218 for backlink count (Ahrefs). A December 2025 follow-up found the correlation held between 0.656 and 0.709 across ChatGPT, AI Mode, and AI Overviews, with YouTube mentions at roughly 0.737.

Should I stop building backlinks entirely?

No. Backlinks remain a documented factor in Google’s classic organic ranking algorithm, and Domain Rating and referring domains still outperformed raw mention volume for some signals in the same Ahrefs dataset. Redirect a share of new effort toward mentions, but keep link building running for the organic traffic you already depend on.

What platforms produce the most valuable AI-citation mentions?

YouTube led the December 2025 Ahrefs study at roughly 0.737 correlation across all three AI surfaces tested. Reddit, Quora, review platforms, and established trade publications followed as strong, independent mention sources, according to analysis of the same dataset (CiteFlow).

How do I track unlinked brand mentions?

Google Alerts is a free starting point for catching web mentions of your exact brand name, though it misses much of what happens on Reddit and in video transcripts. Ahrefs Brand Radar tracks branded mentions specifically inside AI answer engine outputs. Dedicated PR and social-listening mention tools cover forums and review sites more completely than a generic alert.

Can I buy or fake brand mentions?

You can, but it defeats the purpose. The value of a mention comes from its independence: a real third party discussing your brand unprompted. Astroturfed forum posts and incentivized reviews are detectable by platform moderators and by models trained to discount coordinated inauthentic patterns, and getting caught costs you trust on the exact platforms you were trying to win.

Is a mention on Reddit more valuable than a mention on my own blog?

For AI citation purposes, yes, generally. A mention on your own domain is not independent evidence to a model deciding whether to trust and cite your brand. A mention on Reddit, in a thread you did not create, from a user with no relationship to your company, is the kind of independent signal the Ahrefs correlation data measured.

Does the sentiment of a mention matter, positive versus negative?

The Ahrefs correlation studies measured mention volume and distribution, not sentiment specifically. A model accumulating evidence that your brand is discussed does not automatically filter for tone, which means negative mentions still contribute to entity recognition. That makes responding to criticism publicly a mention-building action, not just a reputation task.

How many mentions does a brand need to show up in AI Overviews?

Ahrefs found brands in the top 25% for web mentions averaged 169 AI Overview mentions, versus 14 for the 50th to 75th percentile and 0 to 3 for the bottom half (Ahrefs). There is no fixed threshold, but the gap between quartiles suggests consistent, high-volume presence matters more than a handful of isolated mentions.

What is Ahrefs Brand Radar and how does it relate to this data?

Brand Radar is the Ahrefs tool that tracks how often a brand appears across AI answer engine outputs, including ChatGPT, AI Mode, and AI Overviews. It was the data source behind both the May 2025 and December 2025 correlation studies referenced throughout this article (Ahrefs).

Does branded anchor text still matter if links matter less?

Branded anchor text correlated at 0.527 in the original Ahrefs study, second only to unlinked web mentions and ahead of backlink count. So when you do earn a link, using your brand name as the anchor text, rather than a generic phrase, appears to carry more weight for AI citation than link volume alone.

Why did backlinks correlate so much lower than mentions in this study?

The Ahrefs authors did not give a definitive causal explanation, only correlation data, and explicitly cautioned that correlation is not causation. The likely mechanism is that language models build entity trust from repeated appearance across independent sources, a pattern backlinks do not directly capture since a link can exist without a model ever parsing the linking text as an endorsement.

Should small brands with no PR budget bother with mention building?

Yes. Unlike backlink outreach, which often depends on domain authority you do not yet have, mentions can be earned through direct participation: answering questions on Reddit and Quora, requesting genuine reviews, and pitching a real story to trade press. None of those require a media budget, only consistent presence.

How is this different from traditional online reputation management?

Reputation management historically focused on suppressing negative content and boosting owned-channel visibility in classic search results. Mention building for AI citation is about volume and independence of third-party references, regardless of whether they are glowing, neutral, or critical, because the model is measuring entity presence, not sentiment score.

Does this apply to local businesses too?

The Ahrefs study filtered for brands with Domain Rating above 40, which skews toward larger, more established sites. Local businesses likely see a similar directional effect from review mentions and local press, but the specific correlation numbers cited above have not been validated at the local-business scale.

What is the fastest way to start if I have limited time?

Run a one-week audit of where your brand currently appears, unlinked, versus your top two competitors, using a free tool like Google Alerts as a starting point. Whichever platform shows the biggest gap, whether that is Reddit, review sites, or YouTube, put your next few hours of effort there first.

Does a mention need to use my exact brand name to count?

Based on the Ahrefs methodology, which measured branded search volume and branded web mentions using the exact brand term, exact-name matches are what the correlation data captured. Close variants, misspellings, or product nicknames may still register with language models but were not directly measured in the cited studies.

Will this replace content marketing and SEO entirely?

No. Mentions accumulate because a brand has something worth discussing, usually a real product, real content, or real community engagement behind it. Mention building is an addition to a content and SEO program, directing part of that effort toward third-party, independent visibility rather than only owned-channel optimization.

Where can I read the original Ahrefs research myself?

The original May 2025 study, “An Analysis of AI Overview Brand Visibility Factors,” and the December 2025 follow-up, “Brand visibility in ChatGPT, AI Mode, and AI Overviews,” are both published on the Ahrefs blog and include the full methodology and correlation tables referenced throughout this article (Ahrefs, Ahrefs).

Key takeaways

Ahrefs measured a roughly 3-to-1 correlation advantage for unlinked brand mentions over backlinks when predicting AI Overview citations, across a 75,000-brand study. YouTube mentions carried the single strongest signal in the December 2025 follow-up, ahead of general web mentions, with Reddit, review platforms, and trade press close behind. Backlinks have not stopped mattering for classic Google rankings, but the marginal hour or dollar for AI visibility now buys more from earning an honest mention on a platform you do not own than from one more outreach email chasing a link.

If you are deciding where to spend the next quarter of effort, the data points in one direction: go where the conversation about your category is already happening, and make sure your name is in it. For teams building out the rest of an AI-search strategy, related reads on this blog cover why you might rank on Google but stay invisible in ChatGPT, the 40-point AI visibility audit, how to use Reddit for AI citations without getting banned, and share of answer as the metric replacing share of voice.