Which Types of Mentions Correlate the Strongest With AI Visibility? (260M AI Prompts Analyzed)

But which kind of mention?

In our previous research, we analyzed different brand factors alongside AI visibility—including web mentions, anchors, search volume, number of referring domains and so on.

This time, we zeroed in on brand mention types.

We compared seven of them, from Reddit threads to TikTok videos, with how often the same brand appears in AI answers.

In total, we studied 10,000 brands and 260M (non-branded) AI prompts.

Use the results as a starting point for deciding where to focus your PR and brand-building efforts.

As always, a huge thanks to our data scientist Xibeijia Guan for pulling the data together.

Every mention type we studied correlated strongly with AI brand mentions across ChatGPT, AI Mode and AI Overviews—but none more so than Reddit.

The scores run from 0.59 to 0.75, where 1 would mean the brands rank in exactly the same order on both measures (off-site mentions and AI mentions).

The gap between the strongest and weakest factor here is small, meaning none of the seven mention types should be written off.

But, with a correlation of ~0.75, Reddit mentions (within the threads that rank in Google) track AI visibility more closely than any other factor.

That’s partly by design: Reddit licenses its content to both Google and OpenAI, giving them direct access to its posts and comments to train their models and surface in AI answers.

Across Brand Radar, our database of ~500M AI prompts, Reddit is also currently the second most cited source across ChatGPT, AI Mode and AI Overviews—just behind Wikipedia.

Surprisingly, brand mentions in the press don’t seem to count for more.

“News article mentions” correlated a little less strongly with AI visibility than mentions across the web at large (0.67 combined versus 0.70).

In other words, press coverage still shows a strong positive correlation, but it doesn’t appear to carry extra weight than mentions on general webpages.

If you’re planning earned media, aim for a wide spread of mentions across many sites as well as big-name coverage.

Getting mentioned on a high-traffic page doesn’t appear to win a brand any more AI visibility than getting mentioned on a quiet one.

When we looked at combined correlations across all three AI platforms, “Webpage mentions” and “Total traffic of webpage mentions” both scored 0.70.

This means pitching small, niche sites may get you just as much AI visibility as pitching large publications, because a mention on a low-traffic page seems to count about as much.

ChatGPT tracks off-site mentions slightly less closely than AI Mode or AI Overviews, so mentions are a slightly weaker guide to visibility there.

You can see this in the heat map above, where ChatGPT’s correlations are mostly yellow (0.60 to 0.65)—meaning they’re slightly weaker—while AI Mode and AI Overviews are mostly orange and green.

Still, the factors rank in roughly the same order on all three platforms.

Web mentions and AI mentions rise together, so we measured how many extra AI mentions a brand typically gets as its webpage coverage grows.

For example, if one brand is mentioned on 1,000 pages and another on 10,000, the second would typically show up in AI answers about 3.7 times as often.

This pattern applies to every platform, though it’s strongest on AI Overviews (3.5x) and weakest on ChatGPT (2.2x).

AI mentions don’t grow one-for-one with coverage.

If a page receives 10x more webpage mentions, it only earns 3.7x more AI mentions.

In the table above, going from 1,000 to 10,000 webpage mentions takes 9,000 new pages, but only adds around 200 AI mentions.

These are correlations, so we can’t say for certain that more mentions cause more AI visibility.

But if they do, you shouldn’t expect a handful of new mentions to help your brand show up in AI answers.

If you want a visible change in AI visibility, you probably need a large, sustained increase in the number of sites talking about you, and you should track your progress over months rather than weeks.

Bear in mind, this multiplier isn’t the same for every mention type.

A brand with 10x more Reddit mentions, for example, tends to get about 5.5x more AI mentions.

The typical (median) brand in our study had just 8 Reddit threads mentioning it, and 29% had none at all.

Check how many Reddit threads mention you in Brand Radar.

If there are 8 or more, you are ahead of half the brands we studied.

When we last published this study, one of the most common questions we got was: are these correlations just showing us the shape of a big brand?

In other words, were the strong links between web mentions and AI mentions just the effect of a handful of large, well-known brands?

If so, web mentions wouldn’t be driving AI visibility at all.

They’d simply be a sign of a brand that has built so much interest and authority, that it has subsequently built the strongest associations in AI.

We wanted to put that theory to the test.

In the chart below, the trend line shows how many AI mentions a brand typically gets for its number of webpage mentions, so sitting above it means getting more AI mentions than your web coverage would predict.

The red dots (the most-searched brands) mostly sit above the trend line, while the green dots (the least-searched) mostly sit below it.

But brand size doesn’t tell the whole story.

To separate the two, we compared brands within search volume bands, so that brand size was roughly held constant.

In every band, brands with more webpage mentions still tended to earn more AI mentions.

We broke the search-volume bands down even further, into 50 groups of brands with near-identical search demand, and the pattern held in every one.

Admittedly, the relationship was weaker within each band than across all 10,000 brands, which tells us brand size does account for part of the pattern.

But it doesn’t account for all of it.

So, webpage mentions still appear to predict AI visibility independently of how well known a brand already is.

You can track every factor in this study with Ahrefs.

Start with Reddit mentions, which correlated most strongly with AI visibility.

Brand Radar shows how often your brand appears in AI answers across ChatGPT, AI Mode, AI Overviews and other platforms.

Head to the “Cited pages” report, and filter out prompts that include your brand name to see your unbranded visibility.

Then add a “Your brand: Mentioned” filter, and a Reddit domain filter to see which threads your brand gets discussed in.

Mentions Explorer helps you find every page that mentions your brand.

Set up a report to monitor your brand and competitors, then head to the “All mentions” tab.

To focus on your mentions alone, switch from the “All brands” tab to your brand.

From there you’ll see quality scores, mention sentiment, and your full list of web mentions.

Then switch to the “Overview”, click into your competitor’s report, and scroll down to the “Websites” tab.

This will show you which sites write about your competitor most often.

Pitch those publishers a story or exclusive access to proprietary data, to build your pool of mentions.

Firehose watches the web for new pages that mention your brand and sends them to you as they’re found.

Use it to spot new coverage early, so you can join discussions about your brand or follow up with publishers while it’s fresh.

Alternatively, track your competition and intercept those opportunities.

We picked 10,000 US brands and compared how often AI names them with how often they’re mentioned on sites they don’t control.

For AI mentions, we used public answers in Brand Radar to 260M prompts across AI Overviews, AI Mode and ChatGPT, leaving out any prompt that named the brand.

Brand Radar now tracks brands as entities, so Apple the company is counted separately from apple the fruit.

We compared those AI mentions with seven types of off-site mention.

Off-site mentions came from Content Explorer, which matches brand names as exact phrases rather than as entities. To limit false matches, we only included brand names of at least five characters and excluded common words and personal names.

Content Explorer counts also exclude the brand’s own website, and only include pages with some estimated organic traffic.

We measured each relationship using the Spearman correlation, which shows how closely two rankings line up, from 0 (no relationship) to 1 (identical order).

For every mention measure we studied, brands with more mentions also tended to show up more in AI answers.

Given that press coverage and high-traffic pages didn’t stand out, we suspect “mentions are the new links” is more about how widely you’re talked about than who’s doing the talking.

We can’t say mentions cause AI visibility, but Reddit tracked it most closely and brand size only appears to explain part of the link, so we’d make Reddit and a wide spread of everyday coverage the first places to test.