Brand Trust Signals in AI Search: What to Audit

Five recent analyses point the same way: who Google and AI assistants show is increasingly a matter of selection, not ranking, and selection leans on brand trust signals. Those signals live on your site and far beyond it, and anyone tracking AI visibility with one number, in one language, in one channel is reading the picture wrong.

Key takeaways

• Your website is still the foundation of trust: valid structured data, reviews and content depth correlate with much better results.
• AI visibility is two problems, mentions and citations, and each needs a different tactic.
• Brand recommendations change with the query language and the platform, so an average across everything misleads.
• Google is building a loyalty ecosystem where brands with their own audience win.
• Position is losing value as a metric. What matters is whether you are in the set the system generates.

Why brand trust signals now decide visibility

The Wix report summarized by Neil Patel draws on data from 300 million users. Search still drives 42% of traffic, direct 40% and social 13%. LLM traffic is under 1% but grew fourfold year over year. Sites with valid structured data show double the revenue and nine times the traffic, pages with 50+ reviews far higher median revenue, and sites with 50+ blog posts seven times the sessions.

A caution: these are correlations. Successful sites probably do many things well. I still treat them as a working hypothesis, because they line up with what Search Engine Journal reports: 86% of Google clicks go to a page the user did not type in directly, and Google is pushing down sites without a wider presence. HouseFresh lost over 90% of its traffic within six months.

Step 1: make your site a readable, credible source

Start at home, where you have full control.

• Validate structured data in Google's Rich Results Test and fix errors on products, articles and organization.
• Run a systematic review program. Even 1-5 reviews make a visible difference for ecommerce.
• Put credentials, case studies and contact details on the homepage, which is what both people and machines look for when verifying you.
• Build content depth. A 50+ article target only makes sense if the pieces are expert, not just numerous.
• Consider LLMs.txt, clean page structure and metadata. AI agents now crawl hundreds of thousands of sites every month.

Step 2: separate mentions from citations, platform by platform

Martech, citing the Semrush AI Visibility Index, warns that a single score hides two different things. Across 22 categories and four platforms (ChatGPT, Gemini, AI Mode, AI Overviews), fewer than half of the top 100 mentioned brands also appeared in the top 100 citations. Ford ranked in the top 100 for mentions on all four US platforms in July, yet sat around 748th to 1320th for citations. The reverse also exists: Investing.com was 7th in citations and 50th in mentions.

In practice:

1. For each platform, record separately whether you are mentioned and whether you are cited.
2. If mentions are weak, work on brand awareness and authority off your site.
3. If citations are weak, improve your content and earn links from the categories that platform tends to cite.
4. Measure monthly. A one-off snapshot cannot support a budget decision.

Step 3: test every language you sell in

The SEJ test showed that the same question in a different language returns a different winning brand. For London-Tenerife flights, Claude recommended British Airways in 52.5% of Persian answers but 0% of Russian ones. easyJet was the mirror image (72.5% in Russian, 12.5% in Persian). Consumer electronics and B2B SaaS were more stable.

If you sell in several markets, English alone is not enough, and for travel it is especially risky. Even where one brand wins across languages, watch who comes second, because that is where real share is decided.

Step 4: build an audience Google cannot take away

According to SEJ, Google is building a loyalty ecosystem (preferred sources, follows, profiles) where brands and creators with their own audience win. Audiences also differ: 52.3% of searches by 18-24 year olds end without a click and nearly half of younger users prefer video or audio, while 55-64 year olds click ads seven times as often.

• Add KPIs beyond organic: share of direct and branded traffic, third-party mentions, engagement.
• Drop last-click and measure incremental uplift over weeks. SEO always looks like the hero in the last click.
• For younger audiences invest in video and creators, for older ones in long-form text and PPC.

Step 5: be distinct, and stop worshipping position

Google researchers published a theoretical proof that a language model can rank without a separate index. The experiments ran on Mistral-7B and small datasets, so this is a foundation, not a deployment. The direction is still clear: page 2 fades, visibility becomes closer to binary, and from the outside you cannot tell "ranked badly" from "not generated".

That argues for content that is clearly different from competitors (not a third variation of the same text), for expecting new pages to have a harder time entering such a system, and for watching how ad inventory shifts as blue links shrink.

All of this works as routine, not a one-off project. Charlie brings these checks into one place, and the AI specialists help with schema audits, review programs and per-language monitoring. Agencies running many clients can look at the agency setup.