Which Mentions Drive AI Visibility: Ahrefs' 260M Prompts Study

Reddit mentions correlate most strongly with AI visibility in Ahrefs' 260M-prompt study, at about 0.75, ahead of general webpage mentions (0.70), news articles (0.67), YouTube (0.64) and TikTok (0.60). Those are the mentions that correlate with AI visibility, but Ahrefs says the data shows correlation, not cause.

Ahrefs published the study on 3 November 2026. This post reads it from the Ahrefs, November 2026 original, adds a worked example from its own numbers, and sets out which mention types a brand can realistically earn and how to track them.

Key takeaways

• All seven mention types Ahrefs tested correlated strongly with AI mentions (0.59 to 0.75). None should be written off.
• Reddit threads that rank in Google were the closest match, at about 0.75. Test Reddit first, openly and within each community's rules.
• Press coverage (0.67) did not beat ordinary webpage mentions (0.70). Breadth across many sites looks as useful as a few big-name hits.
• Returns diminish: 10x more webpage mentions went with only about 3.7x more AI mentions. Judge progress over months, not weeks.
• ChatGPT tracked off-site mentions least closely, so measure each engine separately instead of trusting one blended number.
• Correlation is not cause. Confirm any change on your own fixed prompt set before you move budget.

What does the Ahrefs 260M-prompt study say about AI mentions?

Ahrefs picked 10,000 US brands and compared how often AI names each one with how often it is mentioned on sites it does not control. For the AI side it used public answers in its Brand Radar database to 260M prompts across AI Overviews, AI Mode and ChatGPT, leaving out any prompt that named the brand. It measured each link with Spearman correlation, where 1 means two rankings line up exactly (Ahrefs, November 2026).

That design matters. Because branded prompts were removed, the study asks a buyer-style question: when someone asks a category question without naming you, do brands with more off-site mentions get named more often?

Which types of brand mentions help AI visibility most?

Ahrefs ranked seven mention types by their combined correlation with unbranded AI mentions, and gave the multiplier that goes with a 10x increase in each. The table below puts both side by side, taken from the same article (as checked on 2026-11-05).

The spread is narrow. The gap between the strongest and weakest factor is small, which is why Ahrefs says none of the seven should be dismissed. TikTok scores lowest on correlation yet has the second-highest multiplier, so a high rank on one column does not settle the question.

Three findings that change plans

Press does not beat the wider web. News article mentions (0.67) correlated a little less than webpage mentions (0.70). Ahrefs reads this as a case for a wide spread of coverage as well as big-name placements.

Traffic of the mentioning page did not add weight. Webpage mentions and total traffic of webpage mentions both scored 0.70. Ahrefs concludes that a mention on a quiet niche page seems to count about as much as one on a busy page.

ChatGPT is the loosest fit. ChatGPT tracked off-site mentions less closely than AI Mode or AI Overviews on 6 of 7 factors, though the factors ranked in roughly the same order on all three.

How many mentions does it take to move AI visibility?

More than most teams budget for. Ahrefs found that a brand with 10x more webpage mentions typically has about 3.7x more AI mentions. By platform the multiplier was 3.5x on AI Overviews, 3.0x on AI Mode and 2.2x on ChatGPT.

Ahrefs' own illustration: a brand on 1,000 pages shows about 73 AI mentions, and one on 10,000 pages about 273. That is 9,000 extra pages for roughly 200 extra AI mentions.

A worked example from the published numbers

Ahrefs reports that the median brand had 8 Reddit threads naming it and 29% had none. Take a brand at that median. Applying the 5.5x-per-10x figure, getting from 8 to 80 ranking threads would go with roughly 5.5 times its AI mentions. That is our arithmetic on Ahrefs' multiplier, not a forecast, and the relationship is a typical pattern across 10,000 brands, not a promise for yours.

The practical reading: 80 genuine threads is a large, sustained programme. Ahrefs itself says to track progress over months rather than weeks.

Is this just big brands being big?

Partly. Ahrefs tested it by splitting brands into bands of monthly brand searches. The correlation between webpage mentions and AI mentions fell inside each band but did not vanish.

Read that table carefully before you quote the headline numbers. Inside a band the figures are 0.34 to 0.54, well below 0.70. Brand size explains part of the headline link. For a small brand, the fairer expectation is the band figure for your size, not the all-brand figure.

"We can't say mentions cause AI visibility" (Ahrefs, November 2026)

Ahrefs adds that Reddit tracked AI visibility most closely and that brand size seems to explain only part of the link, so it would test Reddit and a wide spread of everyday coverage first.

Why are the engines' results different?

Single-platform views hide a lot of brands. In the study, 46% of brands were invisible in ChatGPT, 23% in AI Mode and 18% in AI Overviews. Across all three combined, only 14% received zero AI mentions (Ahrefs, November 2026).

That is why combined correlations look higher than single-engine ones: fewer brands sit at zero, so differences are easier to see. For a brand team it means a blended score can look healthy while one engine ignores you. See why ChatGPT, Gemini, Claude and Perplexity disagree about your brand and the data on engines naming different top brands.

How does this compare with Ahrefs' earlier 75,000-brand study?

The earlier study, published in December 2025, looked at 75,000 brands and found YouTube mentions had the strongest correlation, about 0.737, with branded web mentions at 0.66 to 0.71 (Ahrefs, December 2025). The new study puts YouTube at 0.64 and Reddit first.

Do not read that as a reversal. The two studies use different brand samples (75,000 versus 10,000 US brands), different prompt sets and different ways of counting a mention, and the new one removes branded prompts. The stable pattern is that every off-site mention measure moved with AI visibility. The order within the top group is not stable enough to build a strategy on one study.

Which mention types can a brand realistically earn?

Sort the seven types by how much of the work you control. This is our reading of the study, not Ahrefs' advice.

Keep your own site crawlable while you build these. OpenAI's crawler page says it recommends allowing OAI-SearchBot in robots.txt so a site can appear in ChatGPT search results (OpenAI, 2026). You can check yours with the free AI Crawler Checker. For the earned-media method in detail, see Earned Media for AI Answers and Community and Creator Signals Without Spam.

How do you track whether mentions move your AI visibility?

Count the mentions, then count the answers. Ahrefs' own tracking route uses its Brand Radar, Mentions Explorer and Firehose tools. Whatever you use, the loop is the same.

1. Fix 20 to 30 unbranded buyer prompts. Do not change the wording mid-test. The AI Prompt Finder can suggest questions buyers ask about your category.
2. Record a baseline per engine: is your brand named, how high, and which sources are cited. Run each prompt several times, since single answers vary (see AI citation stability).
3. Log your off-site mentions by type each month: Reddit threads, webpages, domains, news, video.
4. Compare share of voice against up to three rivals with the free Competitor AI Share of Voice report.
5. Review after 90 days, per engine. Only call a change real if it exceeds run-to-run noise.

For the full method, read how to track your brand across AI engines and the AI Visibility Measurement Standard.

Check one more thing: mention versus citation

A mention names your brand. A citation lists your URL. Ahrefs' academy lesson says only about 28% of AI mentions include a link (Ahrefs Academy). Track both, because a Reddit thread can be the cited source while your brand name is the mention. See citation vs brand mention.

Where Noma fits

Noma by Nurdd is an AEO and AI visibility platform. It asks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, the questions your buyers ask, every day. It shows whether your brand is named, how high, which competitors are named instead and which pages each answer draws on, so you can see whether new mentions move your share of voice. It does not count off-site mentions itself.

See it on your own brand: book a demo or start the 3-day free trial. For a quick first look, run the free brand audit, which shows how AI engines describe your brand today.

What to check this week

• Count the Reddit threads that name you and your top three rivals. Ahrefs' median was 8, with 29% of brands at zero.
• List the 10 sites that mention your closest competitor most and you least.
• Confirm your robots.txt allows search bots, not only training choices.
• Save a baseline of 20 unbranded prompts across engines before you change anything.

How we researched this

We read the Ahrefs study in full from its primary page, including its method and tables, and compared it with Ahrefs' December 2025 study and its academy lesson on AI visibility. We read OpenAI's crawler documentation for the robots.txt point. Figures are quoted as Ahrefs states them. Facts and figures were checked on 2026-11-05. We ran no study of our own, and the worked example is arithmetic on Ahrefs' published multiplier.

Next step

Pick the mention type you can start on this month, fix your prompt set, and measure before and after. To see how 13 AI engines treat your brand today, book a demo or start the free trial. A card is taken at signup so billing can start on day 3; cancel before then and nothing is charged.