How to Use Brand Mentions to Improve SEO and AI Search Visibility

Key Takeaways

• Ahrefs studied 75,000 brands and found branded web mentions correlate with AI Overview visibility at 0.664. Backlinks correlate at 0.218. That gap is the whole argument.
• We ran the play on ourselves. Stan Ventures went from 0% to 1.2% to 4.7% AI share of voice across two tracked months, tracked in Rankscale across ChatGPT, Gemini, Claude, and Perplexity.
• Mentions went from 14 to 126. Citations went from 7 to 60. Average rank inside answers improved from #4.5 to #3.8.
• Not one of those 60 citations points at stanventures.com. Every single one comes through somebody else’s comparison page. That’s the mechanism, and it’s also the risk.
• 26% of the brands in the Ahrefs study had zero AI Overview mentions. If your client sits in the bottom half of web mentions, they’re effectively invisible.
• An unlinked mention on a site AI systems already cite beats a linked mention on a site they ignore.

A client’s keyword positions climb into the top three. Their organic traffic falls anyway.

That exact discrepancy came up twice in our partner strategy calls this quarter. One of those accounts got flagged as critical enough to trigger a full site analysis, because the rankings report said the campaign was working and the traffic report said it wasn’t.

Both reports were right. The AI Overview sitting above those improved rankings answered the question, named a set of brands, and the client wasn’t among them.

Be precise about what changed, because the usual framing is wrong. Google didn’t stop counting links. It added a layer on top that reads the open web as language rather than as a graph of votes, and that layer decides which brands get named in the summary.

This is the conversation SEO agencies are walking into on renewal calls right now. And the honest answer is uncomfortable. The thing that got the other three brands into that answer box wasn’t their backlink profile. It was how often the rest of the web talks about them.

We didn’t want to write this from theory, so we ran it on ourselves first and tracked every number. That data is further down, standings and all.

Definition

What Is a Brand Mention for SEO?

A brand mention is any public reference to a company, product, domain, or founder, whether or not it carries a link.

Your clients collect them whether you run a campaign or not.

That covers a Reddit comment naming your client as the tool that finally worked. It covers a roundup post listing them fifth out of twelve. It covers a podcast host saying the name out loud in a transcript that gets indexed.

What a search engine does with that reference is pattern recognition. It sees your client’s name turn up near a set of topic words often enough, across enough independent sources, and it starts treating the two as connected. SEOs have called this co-citation for years, and the underlying idea hasn’t changed.

What changed is the weighting. A system generating a sentence about the best providers in your client’s category has to have read sentences about providers in that category first. It reaches for the names it has seen most often in that exact context.

So when you audit a mention, the link is your second question. Your first is whether the source is one that search engines and language models already treat as credible.

Correlation Data

The Numbers Nobody Puts in Front of Clients

Ahrefs ran a correlation study across 75,000 brands to see which factors line up with brand visibility in AI Overviews. They used Spearman coefficients, so higher numbers mean a stronger relationship. Here’s the full ranking, and it’s worth screenshotting before your next pitch.

Read the top three rows again. Every one is an off-site signal that has nothing to do with your client’s own website, and every one outranks the link metrics below it. Web mentions correlate roughly three times more strongly than backlinks do.

Ahrefs is careful to say correlation is not causation, and you should repeat that caveat rather than let the chart oversell for you. But the ordering is hard to argue with, and the full study on AI Overview brand visibility holds up to scrutiny.

The practical read: a client with a strong link profile and a thin mention footprint will keep ranking and keep losing share of the answer. Those two things are measured separately now, and only one of them shows up in the rank report you send every month.

The distribution matters more than the coefficient. Ahrefs split the 75,000 brands into quartiles by web mention volume, and the spread between them is not a gentle slope.

Our Own Data

We Ran This on Stan Ventures First

Correlation studies are easy to nod along to and hard to act on. So before we sold this to a single agency partner, we pointed it at our own brand and tracked what happened.

The setup: a fixed prompt set around white label link building and agency fulfillment, monitored in Rankscale across ChatGPT, Gemini, Claude, and Perplexity. Same queries every month. No changing the test to flatter the result.

Here are the two months side by side.

Look at the last row before you get excited about the others. Sentiment dropped almost nine points.

That’s what happens when your sample goes from 14 mentions to 126. At 14 mentions you’re mostly being described by people who chose to write about you. At 126 you’re being described by the whole category, including comparison posts that rank you fourth and move on. A softer average across nine times the volume is a trade we’ll take, but we report it either way, because a mention count that only goes up is a report nobody believes.

One more piece of context that matters. The total citation pool across the tracked prompt set went from about 5,200 to roughly 19,500 between Month 1 and Month 2, and the number of brands the engines named at all rose from 7 to 10. The category got noisier and more crowded at the same time we were climbing. The 4.7% is share taken, not share handed over.

Category Movement

Who Actually Gained Ground

A growth chart with nothing to compare it against is the oldest trick in agency reporting. So here’s every tracked brand in our category, same prompt set, same two months.

Five of the eight went backwards.

We added more than twice the ground of the next brand on the list, and we did it in a month when most of the field lost share.

Those losses aren’t because the other brands stopped working. Look back at the citation pool numbers. It nearly quadrupled between the two months, from roughly 5,200 to 19,500, while the count of brands the engines will name at all went from 7 to 10. When the denominator grows that fast, standing still reads as a decline. Only the brands actively adding cited placements gained.

That’s the whole thesis of this article sitting in one table. Nobody on that list published their way to a gain. The brands that moved up got named on more of the pages AI engines already read.

One caveat we’d rather say than have you catch. This is a movement table, not a size table. Several brands on it still hold more total share than we do, built over years of coverage we haven’t matched yet. We’re the fastest riser in the category, not the biggest name in it. Both things are true, and the same distinction belongs in the reports you send your clients.

Linked vs Unlinked

Unlinked Mentions Are Not a Consolation Prize

For twenty years the standard advice on unlinked brand mentions was to chase down the author and ask for the hyperlink. That advice made sense when PageRank was the only thing being counted. It’s now roughly half the story.

Language models read text, not link graphs. When a model builds its picture of who matters in a category, it’s reading the words around a brand name, how often that name appears near the topic, and what else appears in the same sentence. A hyperlink adds nothing to that calculation.

So an unlinked mention inside a well-read comparison post can carry more weight in an AI answer than a linked mention buried on a domain nothing cites. That flips the old priority order. Vet the domain first, worry about the link second.

There’s one caveat worth holding onto. Branded anchor text still correlates at 0.527, which is second on the whole table. The best outcome is a linked mention on a cited domain, because you collect both signals from one placement.

Practically, that means the unlinked mention audit is still worth running, just for a different reason than before. You’re not hunting for link equity. You’re building an inventory of every place a client is already being discussed, so you can tell which of those places AI systems actually read.

Signal Comparison

Brand Mentions vs Backlinks: What Each One Buys You

You’ve probably had this presented to you as a budget fight. Drop that framing.

The two signals do different jobs, and the useful question is which job your client needs done this quarter.

If your client is scrapping for position four on a commercial term, they need links. If their category is being answered by AI before anyone scrolls to the organic results, they need mentions. Most of your retainers need both, weighted differently by quarter.

The mistake we see most often is pitching mentions as a link building replacement. It sells badly and it underperforms, because branded anchor text sits at 0.527 on that same table. Linked mentions carry both signals at once.

If you want the mechanics of how link equity and anchor distribution work together, our guide to backlink profile analysis covers the audit side in more detail.

Citation Surface

Where AI Systems Actually Pull Citations From

Here’s the part almost every guide on this topic skips, and it changes how you spend the budget.

Ahrefs analyzed the 100 most cited domains in ChatGPT. The top of that list is Reddit, Wikipedia, Amazon, Forbes, and Business Insider. In a separate study of ChatGPT’s top 1,000 most-cited pages, they found that roughly two thirds of those citations are off-limits to marketers, meaning there’s no placement for you to buy and no editor for you to pitch.

So most of the citation surface is closed to you.

Your strategy has to live in the part that’s open.

Understand this before you set a budget. Agencies routinely burn the whole allocation placing clients in outlets that will never be quoted, then report the coverage as a win because it looked impressive on a slide. We’ve been handed a few of those reports to clean up.

That open share is mostly comparison content, category roundups, listicles, review posts, and community threads. It’s unglamorous, it’s where buyers actually make decisions, and it’s where you can place a client deliberately.

Our own citation data says the same thing, and it says it bluntly. These are the pages the engines were pulling from when they named brands in our category, across both months.

Read that last row twice, because it’s the whole lesson.

Every one of our 60 citations came through a page somebody else owns. The engines never quoted stanventures.com. They quoted the roundups that mention Stan Ventures. Our own content contributed nothing directly to that 4.7%, and we publish a lot of it.

That’s a rented footprint. It works, and we’d run it again, but a competitor who buys their way onto the same five pages closes the gap in a quarter. Which is exactly what we did to the brands above us.

The single biggest driver in Month 2 was one comparison page going from 41 pulls to 168. We didn’t place a second time on it. The page itself gained authority and started getting cited more, and every brand named on it, including us, rode that up. Pick pages that are climbing, not just pages that are big.

The Process

The Five Steps That Took Us From 1.2% to 4.7%

No secret sauce here. Five steps, run in order, over about eight weeks.

The order matters more than any individual step. Most programs fail at step one and then blame steps three through five.

Publisher Vetting

The Vetting Standard That Separates Signal From Noise

A mention on a site nobody cites is a mention nobody sees.

This is where most brand mention programs quietly fail, and where yours will too if you skip the screen.

In our own brand mention services delivery, we screen candidate publishers against a citation threshold before a single pitch goes out. The working benchmark is 40 or more existing citations across AI platforms. If a domain isn’t already being pulled into AI answers, a mention placed there has no path into one.

The threshold exists because citation history is the closest thing to a leading indicator available right now. A domain that has already been pulled into forty AI answers has proved models retrieve from it, and the mention you place there inherits that retrieval path.

Our own run is the argument for it. The pages that moved us were already appearing in the citation data before we placed on them. That single filter changes the economics of the whole program, and it gives you something concrete to put in front of a client who wants to know why one placement costs more than another. Here’s the screen we run, in order.

Reporting

How to Actually Measure Brand Mentions

This is where your retainer gets lost.

The tactic works, the reporting doesn’t exist, and the client cancels because nothing on the dashboard changed.

Across the strategy calls our team ran this quarter, a pattern kept repeating. Agencies had bought AI visibility tooling, activated a fraction of the prompt tracking they were paying for, and had no methodology for tying any of it to revenue. One partner had 5,000 monthly prompt checks available and had used a few hundred.

None of this is a tooling problem. Every platform in this space now ships prompt tracking, sentiment scoring, and competitor benchmarking. What’s missing is your decision about which four numbers go on the monthly report and stay there for twelve months.

So the gap is the framework rather than the software. These are the four we report, and they’re the four in our own table above.

Branded query lift is the fifth one worth watching. When someone sees a brand named in an AI answer and doesn’t click, the mention still landed, and it shows up in Search Console as a rise in queries containing the brand name. Set your reporting cadence before the campaign starts, well before the client thinks to ask. Our breakdown of zero-click SEO measurement covers how to isolate branded lift when the click never happens.

Failure Patterns

What Goes Wrong

Four patterns account for most of the brand mention programs that stall out in the first two quarters.

Agency Scoping

Building This Into an Agency Offer

You don’t need a separate brand mention product to sell this. If you already operate as an SEO reseller under your own brand, this slots straight into that stack.

The agencies handling it well are folding citation criteria into the publisher vetting they already run for link placements.

Same blogger outreach team. Same domain screen, with the AI citation count added as a gate. Same monthly report, with share of voice and average answer rank added as two new rows. That’s the whole build.

That’s a scoping change rather than a new service line. If you’re weighing whether to build the capacity in-house or hand delivery to a white label link building partner, our breakdown of whether white label is worth it for small agencies runs the math, since the cost structure is the same.

Pricing it is simpler than you’d assume. If your placement work is already scoped per link, adding the citation screen changes which domains qualify rather than how many hours you bill, so your retainer math holds.

Your window here is narrower than it looks. We moved 3.5 points in eight weeks from a standing start, which tells you the category is still soft enough to climb. It also tells you your competitors can do the same thing to you. Brands that establish a footprint early get pulled into answers that then reinforce the footprint, and if your client is sitting in the bottom two quartiles today, they’re starting from zero against competitors who have been compounding for a year.

For SEO Agencies

Put Your Clients in the Answer Box

We ran this play on our own brand before we sold it. Walk through the citation screen, the publisher list, and the reporting framework with a strategist who does this every day for 150+ agency partners.

Book a Strategy Call

Correlation and citation-surface data from Ahrefs. Stan Ventures AI visibility figures measured in Rankscale across ChatGPT, Gemini, Claude, and Perplexity, using a fixed category prompt set over two consecutive 30-day periods, shown as Month 1 and Month 2.