Brand mentions in AI answers: how to measure them and how to earn them

In Ahrefs' study of 75,000 brands, the factor most closely tied to how often a brand appeared in Google's AI Overviews was branded web mentions, with a Spearman correlation of 0.664. Domain Rating came in at 0.326. Raw backlink count, the thing most link building budgets are still built around, scored 0.218 (Ahrefs, 2025).

The gap at the extremes is larger. Brands in the top quarter for web mentions averaged 169 AI Overview mentions. The bottom half averaged between zero and three. The author is careful to say this is correlation, and big brands get mentioned more for reasons that also get them into AI answers. We agree. But if you are deciding where to spend the next pound, it is a hard chart to ignore.

Mentions, citations and entities are three different things

Most AI visibility reporting blurs these together. They behave differently and you fix them differently.

These can split. A competitor's comparison page can be the cited source for an answer that names you as the expensive option. Your own docs can be cited while a rival is the one recommended. So report on all four.

Unlinked mentions count for more than most SEOs assume

A model reads text. A sentence in a trade publication that says your software "is what most of the accountancy firms we spoke to use for payroll" adds to the association between your brand and that task whether or not the publisher linked it, and whatever rel value the link carries.

That changes how you value placements. A nofollow or rel="sponsored" link on a well-read, relevant publication may do little for PageRank, yet the paragraph around it is exactly the kind of text that shapes how a model describes you. A followed link in an article nobody would quote does neither job well.

There is a limit to this, and it is an uncomfortable one for anyone selling placements, us included. Muck Rack's analysis of more than 25 million links in AI responses found about 84% of citations came from earned media and about 27% from journalism. Paid and advertorial content made up 0.3% (Muck Rack, What Is AI Reading?, May 2026). Sponsored articles rarely become the cited URL. Where they help is in adding accurate, relevant mentions to publications that engines read and do cite for their editorial content. That is the job to buy them for. Don't expect the sponsored page itself to be the footnote.

Measuring share of voice without fooling yourself

AI answers are not stable. In a study run by Rand Fishkin of SparkToro and Patrick O'Donnell of Gumshoe, about 600 volunteers ran 12 identical prompts through ChatGPT, Claude and Google's AI nearly 3,000 times. The chance of getting the same list of brands twice was under 1 in 100. The same list in the same order was closer to 1 in 1,000 (Search Engine Land, January 2026).

So "we rank third in ChatGPT" is a meaningless sentence. What does hold up is frequency: across many prompts and many runs, how often do you appear? The study's authors recommend exactly that.

Build a prompt set that matches how buyers ask

Start with 40 to 100 prompts grouped by intent: category discovery ("best payroll software for UK small businesses"), comparisons ("Brand A vs Brand B"), problems ("how do I run payroll for two employees"), and brand questions ("is Brand A any good"). Use the phrasing your sales team hears. Run each prompt several times per engine, on a schedule, and keep the raw answers.

The metrics worth tracking

How often to measure, and what counts as a change

Weekly runs are enough for most brands. Daily runs mostly measure the engines' randomness. With a prompt set of 50 and a handful of runs each, a move of two or three points in visibility from one month to the next is probably noise. We'd want to see a shift hold for six to eight weeks, across more than one engine, before crediting it to anything you did. Keep a dated log of placements, launches and coverage so you can line changes up against it later.

For sentiment, keep a log of the attributes engines attach to you. "Good for enterprise", "expensive", "limited integrations". Those phrases come from somewhere, usually reviews and comparison pages, and they tell you which sources to work on.

Why citations still matter for clicks

Pew Research tracked 900 US adults' real Google searches in March 2025. Users clicked a result on 8% of visits to pages with an AI summary and 15% of visits without one, and clicked a link inside the summary on about 1% (Pew Research Center). Seer Interactive's data across 42 client organisations found organic click-through on AI Overview queries was 0.70% when the brand was cited and 0.52% when it was not, roughly 35% higher (Seer Interactive, September 2025). Seer is clear it can't prove the citation causes the lift. Fewer clicks overall, then, and a better share of what's left if you are cited.

Where AI engines get their sources

Google says there are no additional requirements for AI Overviews or AI Mode: a page needs to be indexed and eligible to show with a snippet. Those features may use query fan-out, running related searches across subtopics, which means sources are found through ordinary search on narrower questions.

ChatGPT search is different plumbing. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers". Check your own robots.txt, and check it on publications you plan to pay for. A publisher that blocks OAI-SearchBot can still be useful for Google, but it won't be a ChatGPT source.

The mix of cited domains also moves fast. Semrush tracked more than 230,000 prompts over 13 weeks in 2025 and saw Reddit go from being cited in close to 60% of ChatGPT responses in early August to about 10% by mid-September (Semrush, November 2025). Any plan that depends on one source type is fragile. We would not build a strategy around Reddit, or Wikipedia, or any single publisher.

Recency matters too. In Muck Rack's May 2026 data, 57% of journalism citations with a known date had been published in the previous 12 months, and citations peaked in an article's first month. A great mention from 2022 is fading. Steady coverage beats one big moment.

How to earn mentions in the sources models use

The work is closer to PR than to traditional link building, though the two overlap more than people think. This is the order we'd do it in.

1. Run a citation audit

   Export every URL cited across your prompt set and group by domain. The ten to thirty publications that recur in your niche are your target list. Note which ones mention competitors and not you.

2. Fix your entity description

   Use the same one-line description of what you do on your homepage, About page, LinkedIn, Crunchbase, Google Business Profile, app store listings and press kit. Models stitch descriptions together; contradictions produce vague answers.

3. Get into the comparison and "best of" content

   Answers to category prompts lean heavily on listicles and comparisons. Reach the editors who maintain them with something concrete: a free account for testing, pricing they can verify, a customer they can call.

4. Give journalists something to quote

   Original data, a named expert with an opinion, fast answers to reporter requests. Muck Rack found industry-trend queries cite journalism in 46% of responses, more than twice the rate of how-to queries.

5. Place well-written articles on cited publications

   Where a publication cited in your niche accepts contributed or sponsored content, use it to put accurate, specific mentions in front of readers. Label it the way the publisher and the law require. The formats guide covers which type fits which goal.

6. Work on the negatives

   If engines call you expensive or hard to set up, find the review threads and comparison pages saying so. Answer them, fix the product, or publish the evidence that the claim is out of date.

Press releases deserve their own note, since they're the default tool many teams reach for here. Their share of AI citations has swung a lot; see press releases, SEO and AI visibility before budgeting for wire distribution.

Doing this in Rankticker

The AIO Performance tab in each project tracks the metrics in the table above: AI visibility %, share of voice, brand mentions, citation share, a visibility trend, visibility by engine (ChatGPT, Google AI Overviews, Perplexity, Gemini), answer sentiment and the prompts tracked. Details are on the AI visibility page and in the product tour.

Below the metrics, AIO opportunities lists publications verified as sources AI models cite in your niche, and you can order from them directly. In the marketplace those publications carry an AI badge. If you'd rather not pick one by one, set up a campaign with the objective "improve AI visibility" and approve the recommendations it makes. For the wider strategy, read link building for AI search.