Starts Nov 23Live Cohort 3 of Vibe Coding & AI Agents (Nov 23, 18:30 MSK): 3 weeks, 9 live hands-on sessionsCohort 3 (Nov 23) →
ReleaseInsight · On RadarHabr AI→ original

Ahrefs: A Brand With 10x the Web Mentions Gets Only 3.7x More in AI Answers

Brands with ten times as many mentions on third-party sites appear only about 3.7 times as often in AI answers, not ten times. Ahrefs matched 10,000 brands against answers to 260 million queries in ChatGPT and Google. This is a comparison between different companies, not an experiment with purchased publications, so the 3.7 figure cannot be treated as a forecast for your brand. Treat the placements as hypotheses and test them with a small pilot.

AI-processed from Habr AI; edited by Hamidun News
Ahrefs: A Brand With 10x the Web Mentions Gets Only 3.7x More in AI Answers
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #1The tightest link is with Reddit threads surfaced in Google results: a coefficient of about 0.75
Key metric #2Regular external pages scored 0.70, and news articles 0.67
Key metric #3Alice AI and GigaChat were not included in the sample; ChatGPT is more weakly linked than Google

Contractors selling "AI optimization" promise: order a batch of publications, and ChatGPT and Google will start naming your brand in their answers. Ahrefs' study from November 3, 2026 is a rare case where this topic has large-scale data behind it. The headline figure reads as follows. Take two companies: the first is written about on 1,000 external pages, the second on 100. In AI answers, the first will be named not 10 times as often, but about 3.7 times as often. It sounds appealing, but this describes the difference between the two companies, not a promise that publications will work.

What exactly was counted

Ahrefs matched data on 10,000 brands against answers to 260 million queries. They looked at ChatGPT and two Google systems: AI Overviews and AI Mode. Queries where the user named the company themselves were excluded. Only cases where the AI mentioned the brand without a prompt were counted.

A mention is counted when the name appears in an answer at least once. Three occurrences in a single answer still count as one mention in the Mentions metric from Brand Radar. Citation is different: there, a domain is credited if at least one of its pages is among the answer's sources. The brand name in the answer text and a link to the site are separate metrics. The 3.7 coefficient refers specifically to mentions. Moreover, a mention does not mean the AI recommends that company.

Ahrefs searched for web mentions through Content Explorer using exact name matching. The brand's own site was excluded. Only pages that, by Ahrefs' estimate, receive visitors from regular search were included in the count. To remove false matches, they kept names five characters or longer and excluded common words and personal names.

Which platforms are more strongly linked

The link was measured with Spearman correlation. It shows how consistently brands fall into the same order on two metrics: whoever has more mentions on sites usually also has more in AI. A value of 0.75 does not mean a 75% chance of appearing in an answer, nor a 75% increase in recommendations.

  • External web pages: correlation of 0.70 across the three systems combined.
  • News articles: 0.67.
  • Reddit threads visible in Google search: about 0.75. This is the strongest link in the study. A thread is a discussion with participants' replies.

The number of pages mentioning a brand and their total traffic produced the same correlation, 0.70. But it would be wrong to conclude that "page traffic doesn't matter": pages without search traffic were not included in the count at all. The comparison ran on an already filtered set. The gap between 0.70 and 0.67 also does not prove that an ordinary publication pays off better than a media article.

What marketers and product managers should do

The authors state directly that they cannot establish causal influence. It is possible that well-known brands are simply mentioned more often, by both people and models. The companies were grouped by similar search demand to roughly account for size, but this is not an experiment where one company was "boosted" with publications and another was not.

The practical takeaways are down to earth:

  1. If a contractor proposes adding the brand to an article that is already published, ask for Ahrefs' estimate of search visitors for that specific page. That lets you check whether it resembles the pages in the sample.
  2. If they propose a new publication, you cannot verify it in advance: search traffic is only estimated after publication. A promise of "more materials" does not tell you whether those pages will resemble the ones Ahrefs counted.
  3. Reddit and presence on regular websites are reasonable first hypotheses to test, but not a guaranteed channel. The result cannot be carried over to any social media post.
  4. Do not transfer the numbers to your own market. Alice AI and GigaChat were not part of the study. Even among the systems studied, the link varied: on six of seven factors, ChatGPT was more weakly tied to external mentions than the two Google systems.

The 3.7 coefficient cannot be used as a forecast for a single company. A sensible plan is a small test: several placements across different types of platforms, measuring mentions in the AI services you care about before and after, and only then budgeting for scale. Ahrefs' data helps you choose what to test, but the decision to buy placements should rest on the results of your own placements.

Story

Key Questions Answered

What exactly was counted?

Ahrefs matched data on 10,000 brands against answers to 260 million queries. They looked at ChatGPT and two Google systems: AI Overviews and AI Mode. Queries where the user named the company themselves were excluded. Only cases where the AI mentioned the brand without a prompt were counted. A mention is counted when the name appears in an answer at least once. Three occurrences in a single answer still count as one mention in the Mentions metric from Brand Radar.

Which platforms are more strongly linked?

The link was measured with Spearman correlation. It shows how consistently brands fall into the same order on two metrics: whoever has more mentions on sites usually also has more in AI. A value of 0.75 does not mean a 75% chance of appearing in an answer, nor a 75% increase in recommendations. External web pages: correlation of 0.70 across the three systems combined. News articles: 0.67. Reddit threads visible in Google search: about 0.75. This is the strongest link in the study.

What marketers and product managers should do?

The authors state directly that they cannot establish causal influence. It is possible that well-known brands are simply mentioned more often, by both people and models. The companies were grouped by similar search demand to roughly account for size, but this is not an experiment where one company was "boosted" with publications and another was not.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…