Brand Mentions Beat Backlinks in Ahrefs’ Own Data
Brand mentions vs backlinks AI visibility, settled by Ahrefs' own 75,000-brand study. See the ranked signals and what they cannot prove.
By David Jubé · Jul 11, 2026 · 13 min read
Ahrefs’ 75,000-brand study ranked nine signals against Google AI Overview visibility, and unlinked brand mentions finished first while backlink count finished seventh.
Which signal deserves the next hour, brand mentions or backlinks? On brand mentions vs backlinks AI visibility, the largest published test says mentions.
Backlinks did not stop working, and the study’s own authors are careful about how far a rank correlation travels. Read the nine signals in order, each with its limit attached, and you get a defensible order of operations instead of a slogan.
Key takeaways
- Across 75,000 brands, Ahrefs measured branded web mentions at 0.664 and backlink count at 0.218 against Google AI Overview visibility, on the Spearman rank scale.
- Ahrefs’ three strongest signals are all off-site: branded web mentions, branded anchors and branded search volume. Every link metric it tested ranks below all three.
- Seer Interactive’s ChatGPT test ran in the same direction, reporting the impact of backlinks as “weak or even neutral” while page-one Google rankings tracked at roughly 0.65.
- Ahrefs states in the study that correlation is not causation, and that every factor it measured landed between moderate and very weak. Squaring 0.664 gives about 0.44.
- Backlinks remain a documented Google ranking factor. What this data revises is the order of effort for AI visibility, not the value of a link.
Brand Mentions vs Backlinks AI Visibility: Nine Signals, Ranked
Ahrefs ran a Spearman rank correlation between nine brand-level signals and how often each of 75,000 brands was named inside Google AI Overviews. Unlinked web mentions of the brand name came out on top at 0.664, and raw backlink count came seventh at 0.218.
One paragraph carries the whole answer. What follows is the mechanism, the second engine that repeats it, and the four things the number cannot support.
Here is the full order, with every figure taken from the published 75,000-brand correlation study.
Where Backlinks Land in the Order
Read the families rather than the individual rows and the shape gets clearer.
The three strongest signals are all things other people write about you. The three weakest are things you buy. Link metrics sit in the middle, not at the bottom, which is worth saying plainly.
- Off-site text leads: mentions at 0.664 and branded anchors at 0.527.
- Demand follows: branded search volume at 0.392.
- Link metrics cluster from 0.326 down to 0.218.
- Paid signals close the table at 0.216 and 0.215.
The gap that matters is not backlinks against zero. It is 0.664 against 0.218, with three measures of off-site text sitting above every measure of the link graph, in a dataset built by the company whose product is the link graph.
Ahrefs also reported a distribution effect underneath the correlations. Brands in the top quartile for web mentions averaged 169 AI Overview mentions, the quartile below averaged 14, and the bottom half averaged between zero and three.
Twenty-six percent of the sampled brands were never mentioned at all, which tells you something about the shape of the distribution before you read a single coefficient.
What a Spearman Number Can Carry, and What It Cannot
Spearman rank correlation asks a narrower question than the decimal suggests. It does not ask whether more mentions produce more AI visibility; it asks whether brands ranked higher on mentions also tend to rank higher on AI Overview presence, on a scale running from minus one to one.
Ahrefs is explicit about the ceiling on its own numbers. The study states that correlation is not causation, and that all the factors studied “revealed moderate to very weak correlations on the Spearman scale.”
That single line is the honest reading, and it is the one to carry forward. The strongest thing in the table is still only moderate.
What 0.664 Explains, and What It Leaves Open
Squaring the coefficient gives a rough sense of how much of the rank variation is accounted for.
- Branded web mentions: 0.664 squared is about 0.44, so a little under half the rank variation moves with mentions.
- Backlink count: 0.218 squared is about 0.05, so roughly a twentieth of it moves with links.
- Everything else: the majority of the variation is unexplained by either signal, and the study does not claim otherwise.
It helps to know what a genuine non-signal looks like at this scale. Otterly’s study of 1,028,959 cited URLs across six AI platforms found URL length correlating with citation count at minus 0.025, and path depth at 0.002.
Those are the numbers of a variable doing nothing at all. A coefficient of 0.218 is small, but it is not that, and the distinction matters when you decide what to stop doing.
Second Engine, Same Direction: Seer Interactive on ChatGPT
One vendor’s number can be an artefact of one vendor’s index. A second test, on a different engine, with a different method, is what turns a finding into a pattern.
Seer Interactive pulled more than 300,000 finance and SaaS keywords, extracted close to 600,000 People Also Ask questions from Google and Bing, narrowed those to 10,000, then ran them through OpenAI’s GPT-4o to see which brands the model named.
Its published read was blunt: “We expected backlinks to play a big role, but their impact was weak or even neutral.” That result sits alongside everything else known about how answer engines choose their sources, which is that ranking still feeds the pipeline even when links do not.
- Page-one Google rankings correlated with LLM mentions at roughly 0.65 in Seer’s data.
- Bing rankings mattered less, at roughly 0.5 to 0.6.
- Ahrefs, summarising Seer’s charts inside its own study, puts the ChatGPT figures at 0.25 for domain rank and 0.10 for backlinks.
Different engine, different sampling frame, same direction of travel.
A third dataset comes at brands from another angle. Moz generated 50,000 query fan-out prompts across 1,000 subtopics and found 12.8% of them carried a brand mention, with just two fan-out types, entity and comparison, accounting for 97% of those mentions.
Moz’s conclusion is a caution as much as a finding: brand mentions “are heavily biased by the types of prompts we choose to track.”
Hold onto that, because it means your measured AI visibility depends partly on which prompts you chose to measure. If you are picking a tracker, the tools worth comparing and what each one actually counts is the practical version of the same problem.
Backlinks Did Not Stop Working, They Stopped Being the Lever Here
This is where the finding gets misquoted, so the boundary is worth stating precisely.
Ahrefs’ own ranking-factors guide still lists backlinks first among the seven factors it says Google uses, citing Google’s Andrey Lipattsev confirming in 2016 that they are one of Google’s strongest ranking factors. Nothing in the AI Overview study retracts that.
Classic organic ranking and AI Overview presence are different targets, and a signal can be strong for one and mid-table for the other.
The mechanism is not mysterious.
- A backlink is a routing instruction: it tells a crawler that one page vouches for another, and it carries traffic.
- A mention is text: a language model trained on web text learns a brand the way it learns anything else, from how often the name appears and in what company.
- Both are earned the same way: the outreach that produces one frequently produces the other, which is why this is a question of emphasis rather than replacement.
Backlinko draws the same line between linked and unlinked mentions, noting that unlinked mentions build brand context for search engines and AI systems without passing traffic. The two do different jobs rather than competing for one.
Where Link Building Still Earns Its Place
If you are still building the link side from nothing, earning authority without a link budget covers that work on its own terms, and none of it is wasted here. Moz’s beginner’s guide to link building remains the reference for doing it without shortcuts.
There is also a floor beneath both. A brand that cannot be crawled and indexed appears in neither channel, and Botify makes the case that indexation is where visibility in traditional and AI search both start. Fix that once, then argue about the ranking of the signals above it.
Your Mention Footprint Is Measurable This Week Pull your brand’s web mention count and your referring domain count for the same period, then look at which of the two has actually moved. If mentions are flat while links climb, your AI visibility effort is going into the weaker of the two signals. Book a free diagnosis
Four Things This Ranking Cannot Tell You
This is the fourth claim in the series to go through the same test, and it is the only one so far where the evidence argues for spending more rather than less. That makes the limits more important, not less.
The Reverse-Causality Problem
A brand that is already large accumulates mentions and AI visibility together, as two symptoms of being large.
The correlation cannot separate the brand that became visible because it was mentioned from the brand that got mentioned because it was already visible. Ahrefs’ own quartile data leans that way, since visibility concentrates at the top while the bottom half barely registers.
The Sample Is Not Your Business
Ahrefs filtered to domains with a Domain Rating above 40 whose top keyword carried at least 800 monthly searches. That is a deliberate filter for brands large enough to measure, so the finding is best supported for established brands rather than for a company nobody has written about yet.
The other two limits are narrower and easier to miss.
- The primary dataset is Google AI Overviews only.
- Seer’s work corroborates the direction on ChatGPT, but it is a different test rather than a replication.
- Nothing here establishes a figure for Perplexity, Gemini or Copilot.
There is a fourth limit that sits outside the data entirely. AI visibility is not the whole of search demand, and Nielsen Norman Group’s study of how people choose between chatbots and search found users reaching for AI to explore and synthesise, then going back to traditional search when accuracy and trust mattered.
Even where mentions do land, they are not the same thing as a recommendation, which is the distinction the closing article in this series is built on.
Where the Next Hour of Effort Goes
Given a gap this wide, replicated in direction on a second engine, I would put the next unit of AI visibility effort into earning mentions rather than into acquiring another link.
The work looks like digital PR rather than link outreach. Pitching a data point a journalist can use, showing up in the comparison pieces that Moz found carry almost all the brand mentions, being the name a writer reaches for when covering your category, whether or not the reference carries a hyperlink.
Mentions You Can Actually Earn
None of that is new work, which is the point. Siege Media’s nineteen worked digital PR campaigns are the same craft the industry has run for a decade, now pointed at a signal that was previously treated as a consolation prize.
- Publish a number: original data is the cheapest reason for someone else to write your name.
- Get into the comparisons: entity and comparison prompts carried 97% of brand mentions in Moz’s set, so that is where presence pays.
- Answer journalist queries: pitch the expertise rather than the product, and take the mention without the link if that is what is on offer.
- Ask for the brand in the anchor: branded anchors ranked second at 0.527, so “click here” wastes a link you already earned.
- Track mentions, not only referring domains: the signal you are trying to move has to be the number you watch.
Measuring What You Decided to Move
Measurement has to move with the tactic. iPullRank’s Mike King frames it as a category change rather than a metric change: “AI Search channels are more branding channels so you have to think about performance differently”. Moz sells a Brand Authority score on the same premise, that brand strength is a measurable quantity rather than a mood.
Attribution stays hard either way, which is a separate problem covered in measuring AI traffic when there is no referrer.
Two Caveats Before You Move a Budget
Two caveats sit on top of it. SparkToro’s write-up of Similarweb’s downstream study found direct visits and branded search rising faster for AI-recommended brands than for unmentioned ones across finance, travel and beauty, then asked the obvious question back: would those users have found the brand anyway, and would the result hold for a small unknown brand?
Nobody has answered that yet.
Start with monitoring, because you cannot earn what you cannot see. SparkToro’s argument for treating brand monitoring as more than reputation management is the right frame, and HubSpot’s vendor playbook for tracking answer-engine visibility is a serviceable checklist once you have decided what to count.
The deeper move underneath all of it is becoming a well-documented, consistently named entity rather than chasing any single tactic, which is how a brand becomes a known thing to a model. Mentions are how that recognition gets built, and showing up in an AI Overview is where you find out whether it worked.
Frequently Asked Questions
Do brand mentions really correlate more strongly with AI visibility than backlinks?
Yes, in the largest published test of the question. Ahrefs measured unlinked branded web mentions at 0.664 against Google AI Overview visibility across 75,000 brands, and raw backlink count at 0.218 on the same Spearman scale. Seer Interactive’s ChatGPT test ran in the same direction.
How was the 75,000-brand correlation study built?
Ahrefs ran Spearman rank correlations between nine brand-level signals and how often each brand appeared inside Google AI Overviews. It filtered to domains rated above 40 whose top keyword carried at least 800 monthly searches, then analysed millions of AI Overview responses. It is a correlation study, not a controlled experiment.
Does a correlation of 0.664 prove that mentions cause AI visibility?
No, and Ahrefs says so inside the study. Correlation is not causation, and a large brand accumulates mentions and AI visibility together as two symptoms of being large. Squaring the coefficient gives roughly 0.44, so most of the rank variation is driven by something the study did not isolate.
Do backlinks still matter for search rankings?
Yes. Ahrefs’ own ranking-factors guide still lists backlinks first among the factors it says Google uses, citing a 2016 confirmation from Google. The AI Overview finding revises where the marginal hour of AI visibility work should go, and it does not retract the value of a link for classic organic ranking.
How do you actually earn brand mentions?
Publish something citable, get named in the comparison and roundup pieces where Moz found brand mentions concentrate, and answer journalist queries in your category. Ask for your brand name in anchor text when a link does happen, since branded anchors ranked second in the Ahrefs study at 0.527.
Which numbers should you track for brand mentions and AI visibility?
Track mention volume rather than referring domains alone, using a monitoring tool or a brand-strength score. Expect the number to depend partly on which prompts you chose, since Moz found two fan-out types carried 97% of brand mentions across its 50,000-prompt set. Measure the trend.
Continue Reading
More From This Series
- What the AEO Evidence Actually Shows Right Now
- Schema Markup and AI Citation: What a Test Found
- llms.txt Has Adoption but No Evidence Behind It
- Every AI Overview Statistic Contradicts the Next
- Being Cited by AI Is Not the Same as Being Chosen
More from TDM Insights
- SEO and AI Visibility Tracking Tools, Compared
- Backlinks for Startups: Earn Authority, No Budget
- Entity SEO: How to Become a Known Thing to AI
- How to Show Up in Google AI Overviews
- How AI Answer Engines Choose Their Sources
- How to Measure AI Traffic With No Referrer
Explore TDM Insights Topics