Blog
Link records at the domain level may not be good predictors of mentioning AI search
Jun
Entering this study, I suspected that link indicators at the domain level would not be a good mention of mention. I expected Google AI to show more correlation because they can exploit some classic search signals, while other systems may not. In the case of classic search, usually indicators at page level are more predicted by rankings than those at the domain level.
I looked at the 50 best sites listed at the Brand Ahrefs radar for Google AI, Chatgpt and embarrassment reviews. This is an AI ~ 76.7 m review, horses of ChatgPT by 957 thousand. And 953.5k hints of embarrassment for the month of 2025.
I compared the site, he mentions their rank Ahrefs (AR) in Ahrefs. AHREFS RANK (AR) evaluates all websites in our database so that according to the size and quality of their following domains.
Sidenote.
The brand’s radar is not only another LLM visibility monitor, we follow a immense number of queries in all these systems and you can ask for every product, service or brand and compare competitors. It’s more like a site explorer than to track ranks. In addition, we have a network visibility indicator so that you can see how you talk about the Internet and the search requirements indexes to see how popular you are in searching.
Here’s what the brand’s radar looks like.
Let’s start.
Embarrassment shows a moderate correlation Ahrefs, between the mention of the most frequently cited domains and their domains, Google AI is browsing a very needy correlationAND Chatgpt without correlation. I expected Google to have a stronger relationship than embarrassment, so the results surprised me.
This correlation would probably change with a larger set of data except the only 50. Top 50 is very well connected and cited websites.
I will give the usual correlation not equal to the waiver of causal fitness.
Here Spearman rank correlations Memory Share vs. Ahrefs ranking in the top 50 domains in each system:
| You have an assistant | Spearman RHO | Value P | Correlation |
|---|---|---|---|
| Google AI review | -0.12 | 0.47 | Very tender |
| Chatgpt | 0.01 | 0.95 | Lack of relationship |
| Embarrassment | -0.34 | 0.095 | Faint |
Google AI Overindex review on UGC pages, such as YouTube, Reddit and Quora, as well as encyclopedic content such as Wikipedia. We have already seen it in several other studies.
These sites have additional weight in mention than provided for by their link profile.
Google Ai Discussing underwear on social networking sites such as LinkedIn, Instagram and Facebook. Another pattern we saw in other studies.
We also see a similar story from chatgpt, in which they are overweight in such things as Wikipedia and information services such as Reuters, just like before.
Tempting also shows the same prejudices we saw. YouTube and Wikipedia are overweight, social media sites are underweight.
Final thoughts
I’m not sure if this is enough to get a good measure. In fact, we saw higher correlations in Google AI reviews compared to DR, when we studied 75,000. Sites than we saw in correlations for normal Google rankings in 1 million keywords, 0.326 vs 0.131.
We conduct a larger test outside 50, and we will look at more page level indicators for more information.
