AI brand mentions: why off-site signals decide

AI Brand Mentions: Why Off-Site Signals Decide Your Visibility

AI platforms almost certainly know your brand. Getting them to mention it to a customer at the moment of purchase is a completely different problem. A Q2 2026 study by Victorious (175 brands, 5 verticals) found that AI correctly recognises 96% of the brands tested, but names them in only 11% of purchase-intent queries. The other 89% simply do not exist in AI responses.

Key takeaways

• The strongest correlations with AI mentions are referring domains (0.49) and third-party web mentions (0.45), not your own website content.
• 2,000 indexed third-party mentions is the critical benchmark: below it, brands receive an AI citation in only 3% of cases.
• AI uses different sources depending on buyer journey stage: directories and comparison sites dominate category research, while educational content drives problem awareness.
• Cited URLs and clicked URLs are almost entirely separate populations, so AI visibility and business impact must be measured independently.
• Google AI Overviews opt-out is a risky move without clean data, which won't be available until December 2026 at the earliest.

Why your own website is not enough

Of the 49,391 citations analysed in the Victorious study, just 4 pointed to a brand's own domain. That is not a bug, it is by design. AI platforms like ChatGPT, Claude and Google AI Overviews are built to cite independent sources, not company websites. Semrush confirms this with concrete benchmarks: Patagonia has over 87,000 AI mentions thanks to 103,000+ third-party mentions on Reddit, OutdoorGearLab and REI; Shopify has 45,000+ AI mentions driven by coverage on G2, Capterra and Reddit.

Two trust factors matter: source independence and source authority. Your own blog satisfies neither.

Where AI visibility actually comes from

The buyer journey has two distinct stages, and each requires a different approach.

Category research (consideration stage): This is where AI cites directories and comparison sites. The brand mention rate here is 12 times higher than in the problem awareness stage. If you are not listed on G2, Capterra, TrustRadius or in the roundup articles that AI already cites, you will not be named when someone asks "best software for X" or "alternatives to Y". ChatGPT, for example, cited Fit Small Business for accounting software in 6 separate prompts.

Problem awareness (early stage): AI draws on educational content here, but rarely names the brand directly (mention rate: 0.10%). Publishing this content still makes sense because it builds authority and future citation potential.

Original research is a cross-stage tactic that works for both. The Houzz 2026 study earned 180 backlinks from 81 domains and was cited in 180 AI prompts. Publishing your own research quarterly raises both domain authority and the probability of AI citations.

How to measure AI visibility properly

Similarweb data highlights a critical nuance: 65% of URLs that ChatGPT cites sit 2-3 folders deep in a site's architecture, but 58.8% of referral traffic lands on the homepage. Cited URLs and clicked URLs are almost entirely unrelated populations. Merging them into one dashboard leaves you blind to both problems.

Proper AI performance reporting uses two separate KPIs:

1. Citation rate + folder depth - measures whether AI trusts your content as a source.
2. Referral landing pages + downstream conversions - measures what happens after a human clicks.

A range of tools now covers AI visibility monitoring at different price points. OtterlyAI Lite starts at $29/month, AthenaHQ has a free tier, and the Semrush AI Visibility Toolkit at $99/month offers the best value for agencies, including a white-label add-on for $20. For enterprise clients, Semrush Enterprise AIO tracks 11 AI platforms with a database of 289 million prompts. If you manage multiple clients, the agency solutions page covers approaches to AI monitoring at scale.

The opt-out trap in Google AI Overviews

Google has introduced a Search Console setting to exclude websites from AI Overviews, AI Mode and generative Discover features, pushed by the UK's CMA regulator but available globally. The problem: in 15.5% of US and 17.46% of UK trending news queries, the Top Stories carousel appears inside AI Overviews. Opting out can therefore remove both placements at once.

Analytics cannot yet separate traffic from AI Overviews versus AI Mode. The CMA has requested click data and CTR from Google starting December 2026. Page-level controls arrive in March 2027. Until then, this is a property-level toggle with very limited visibility into its consequences.

The practical recommendation: before any opt-out decision, check the generative AI performance report in Search Console and understand how many impressions AI features are currently delivering. For news and content publishers, opting out without this data is a gamble.

What this means for your strategy

AI search is not replacing Google, it is layering on top of it. Similarweb data shows that 95% of ChatGPT's 494 million users still use Google in parallel. The AI category is growing fast (9.5 billion monthly visits, +70% year-on-year), but abandoning classic SEO would be a mistake.

Three priorities stand out:

• Get past the 2,000 indexed external mention benchmark by diversifying sources (referring domains, review platforms, directories).
• Publish original research quarterly and target placement in the roundup lists that AI already cites.
• Set up separate measurement for citation rate and business impact before optimising anything.

Charlie's AI specialist agents can help audit your current AI visibility and identify the specific sources where your brand is missing.