Claim: seeding brand mentions across the web boosts AI visibility

The claim

“Get your brand mentioned on as many blogs, forums and video channels as possible and AI search will recommend you.”

Why the verdict

Google agrees that its generative AI features “can show what’s being said about products and services across the web, including in blogs, videos, and forum discussions. However, seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem. Our core ranking systems focus on high-quality content while other systems block spam; our generative AI features depend on both.”

Officially documented · Google Search Central

Google’s definition of spam includes “attempting to manipulate generative AI responses in Google Search.”

Officially documented · Google Search Central

The distinction is authenticity. Genuine discussion of a product is part of what AI features can draw on. Manufactured mentions are what Google says its systems try to discount or block.

What answers cite

Re-tested 2026-10 (consumer ChatGPT, logged out, US, English; via DataForSEO): differs from this claim. 19% of 267 citations were third-party sites (±4 pts, 30 answers).

In DATA17’s August 2026 sample of one AI assistant (24 prompts asked three times each, 72 answers, 297 citations), 80.5% of cited sources were third-party sites and 19.5% company websites. The most cited sources were structured category lists (“best X tools” pages), and the top four were all such list sites. Traditional media ranked far lower. Across the 72 answers, only 18 of the category’s 171 brands were mentioned.

Observed in practice · observed 2026-08 · DATA17 AI Research Lab · 解数智能 (unpublished report, not publicly available; readers can't check it)

Our re-test (above) did not reproduce the 80.5%. We asked consumer ChatGPT 10 fixed questions about background-removal tools, three times each. Vendors’ own sites (product pages, help centres, app store listings) made up about four-fifths of citations. The other finding held: nine in ten third-party citations were list or comparison pages. How much an answer leans on third parties depends on the assistant and the category. Measure your own (Measuring AI visibility by sampling prompts).

What others publish about a product still matters, especially list and comparison pages. That is a reason to earn real coverage on review and comparison sites, not to manufacture it. Paid or fake list placements are a target of Google’s spam policies. In China, the 2026 campaign against AI data poisoning targets them too (GEO regulation in China).

Social buzz does not carry over

DataStory compared brands’ social media share with their share in AI answers (9 consumer categories, six Chinese AI platforms on web and app (12 endpoints), each question asked 3 times; social data July 2025 to June 2026). Selling points where a brand led on social media kept “only about a fifth” of that share in AI answers. The average over 8 categories was 34.5% on social media against 5.6% in AI answers. In 8 of the 9 categories, the leading social media brands were strong on social but weak in AI. (Quotes translated from Chinese.)

Observed in practice · observed 2026-08 · 数说故事 (unpublished report, not publicly available; readers can't check it)

The report’s explanation is that AI answers lean towards sources that can be verified and traced. This is the authors’ inference. The study did not test it.

Hypothesis · 数说故事 (unpublished report, not publicly available; readers can't check it)

With three asks per question, the category averages rest on hundreds of answers each. Single-brand or single-keyword figures in the report fall well short of the samples recommended in Measuring AI visibility by sampling prompts.

What the verdict does not say

• That reputation doesn’t matter. Real coverage and discussion are part of the web that AI features summarize.
• Anything measured. Neither Google nor independent research documents how much authentic mentions influence citations.
• Anything about AI products other than Google Search, whose policies on this are not documented in these terms.