How Do You Build Branded Search Demand That AI Notices?

Why Is Branded Search Suddenly an SEO Priority?

Because the strongest signals in AI search are not on your website. Ahrefs studied 75,000 brands in December 2025 and found that branded web mentions correlated with AI visibility far more strongly than backlinks. The things people say about you off your site now shape whether an AI names you at all.

This is a real shift in where the work sits. For fifteen years, SEO teams optimised pages and chased links. Both still matter. But the evidence now points at something older and harder: whether people know your name and say it out loud in public.

We build sites for B2B companies, so we watch this closely. A beautifully built site that nobody mentions is invisible to an answer engine. Here is the framework we use to think about it, and the honest limits of what anyone can measure.

What Does the Data Actually Say About Brand Mentions?

Ahrefs analysed 75,000 brands and millions of AI responses across ChatGPT, Google AI Mode and AI Overviews, published on 12 December 2025. Using Spearman correlation, branded web mentions scored 0.664 for ChatGPT, 0.709 for AI Mode and 0.656 for AI Overviews. Those were among the highest figures in the study.

The pattern held across related signals. Branded anchors scored 0.511 for ChatGPT, 0.628 for AI Mode and 0.527 for AI Overviews. Branded search volume, meaning how many people type your name into a search box, scored 0.352, 0.466 and 0.392 across the same three surfaces.

Ahrefs is careful about what this means, and so are we. The study states plainly that correlation is not causation. These numbers do not prove that buying mentions lifts AI visibility. They show that visible brands and mentioned brands are the same brands.

Why Do Backlinks Matter Less Than They Used To?

They do not matter less for ranking. They matter less for this particular job. In the same Ahrefs study, number of backlinks and URL rating showed only weak correlation with AI visibility, while Domain Rating scored between 0.266 and 0.326 depending on the surface. Mentions beat links by a wide margin.

That makes sense once you think about how these systems read the web. A link is a machine readable vote. A mention is a sentence a human wrote about you in context. Language models are built to understand sentences. They do not need the link to understand that a page is talking about your company.

The practical consequence is uncomfortable for a lot of SEO programmes. A link building campaign that produces links without conversation is optimising the weaker signal. We cover the mechanics of this in our piece on unlinked brand mentions in AI search.

What Counts as a Brand Mention?

Any public text where your company name appears in a real sentence. A review on G2, a comparison thread on Reddit, a LinkedIn post from a customer, a podcast transcript, a conference agenda, a job listing, a press quote, a YouTube video description. None of these need to link to you to count as a mention.

Ahrefs also found that YouTube mentions carried the highest correlation of any factor in the study, at around 0.737. We would not read that as a mandate to start a YouTube channel. We would read it as evidence that mentions in places with heavy crawl coverage and strong transcripts carry real weight.

The useful reframe is this. Stop asking where you can get a link. Start asking where your buyers already discuss the problem you solve, and whether your name comes up there.

How Do You Create Mentions Without Faking Them?

Slowly, and by being worth mentioning. Google's own guide to optimizing for generative AI features says creating inauthentic mentions provides minimal benefit. That is Google telling you directly that a mention farm is wasted money. The mentions that count are the ones someone chose to write.

The tactics that work here are unglamorous. Publish original data nobody else has, because a number with your name attached gets quoted. Take a clear position on a contested question, because people cite the argument they are agreeing or disagreeing with. Show up on other people's podcasts and panels. Make it easy for customers to talk about you publicly.

Product surface area helps too. Free tools, public benchmarks, open templates and public changelogs all give people a concrete reason to name you. We have seen this pattern hold across the B2B sites we work on: the companies that get talked about are the ones that shipped something quotable.

One more thing that costs nothing. Be consistent about your own name. If your company is written three different ways across your site, your docs and your social profiles, you are splitting your own signal. Entity consistency is basic hygiene, and we explain why in our guide to entity SEO.

Can You Measure Your Position in AI Answers?

No, and this is the most important thing to understand before you buy a tool. SparkToro published research on 28 January 2026 in which 600 volunteers ran 12 prompts through ChatGPT, Claude and Google AI a combined 2,961 times. The same question rarely produced the same answer.

The headline number is blunt. SparkToro reports there is less than a 1 in 100 chance that ChatGPT or Google's AI will give you the same list of brands in any two responses. Getting the same list in the same order is worse again, at roughly 1 in 1,000 runs.

SparkToro's conclusion is worth quoting because it saves people money. The research states that measuring your brand's presence in AI answers with precision is a fool's errand, and that any tool giving a ranking position in AI is, in their words, full of baloney.

What Should You Measure Instead?

Frequency, not position. SparkToro's own recommendation is that visibility as a percentage across dozens to hundreds of prompts, each run multiple times, is a reasonable metric. You are estimating how often you appear, not where you rank. That is a survey, not a scoreboard.

Pair that with two first party numbers. Branded search volume in Google Search Console tells you whether more people are typing your name. The Generative AI performance report in Search Console shows impressions inside Google's AI features, though Google's documentation is clear that it carries impressions only and no clicks or position data.

Then add the count that matters most and is easiest to forget: how many new public mentions of your name appeared this quarter, and where. Tools like Ahrefs Brand Radar exist for this, but a tracked spreadsheet works fine when you are starting. We go deeper on the reporting side in our guide to measuring AI search traffic.

How Long Does Branded Demand Take to Build?

Longer than any campaign calendar. Nobody can honestly give you a number here, and we will not invent one. What we can say is that branded demand behaves like a stock rather than a flow. It accumulates from many small public moments and it does not respond to a single quarter of effort.

This is why the metric is useful and why it is unpopular. It resists the quick win. A team that needs a result in six weeks will always choose paid search over building a reason to be mentioned, and they will be right in the short term and wrong over three years.

The honest framing for a board is that branded demand is compounding infrastructure, in the same category as documentation quality or site performance. You fund it because the alternative is renting attention forever.

What Does This Framework Look Like in Practice?

Four moves, run continuously. First, audit where your name currently appears and where it does not. Second, pick the two or three public places your buyers actually gather. Third, ship something quotable into those places every month. Fourth, track mention count and branded search volume as your leading indicators.

The audit is the step most teams skip, and it is the cheapest. Search your own brand name and read what comes back. Read the review sites. Read the threads. You will usually find that your name appears in far fewer conversations than your team assumes, and that the ones it does appear in are years old.

The shipping step needs a real owner. In our experience this fails whenever it is assigned to everyone, because quotable work is nobody's default job. One person with a monthly output target beats a committee with a strategy document.

Where Should a B2B Team Start This Quarter?

With a baseline and one bet. Record your current branded search volume, your current impressions in the Generative AI performance report, and a rough count of public mentions. Then pick a single quotable asset you can actually ship in ninety days and put one name against it.

Resist the urge to buy a visibility tool first. The SparkToro research makes clear that precision here is not available at any price, and a dashboard will not create the mentions. The spend belongs in the thing worth mentioning, not the measurement of whether it worked.

If you want help deciding what that quotable asset should be, or you want a site that makes it easy for people to cite you, we are happy to talk it through. You can find us at phoenix.studio.