Which AI trust signals actually get your brand cited?

What the scorecard says about AI citations

NP Digital scored seven trust signals across six AI platforms. Third-party citations won on every one, at 4.5 to 4.8 out of 5.

The study interviewed 100 marketers and asked them to rate each signal from 1 to 5. The score blends how often a signal drove a citation with how prominent the brand looked in the answer.

Four things stand out. Third-party citations led everywhere. Expert authors came second at 4.0 to 4.6. Backlinks swung hardest, from 1.9 on ChatGPT to 3.9 on Google AI Overviews. Structured data bottomed out at 1.4 on ChatGPT.

Neil Patel called third-party citations the closest thing to a universal trust signal. That is a strong claim from one dataset, so we checked it against everyone else who has measured the same question.

How to grade evidence about AI search visibility

Grade every claim by how it was produced. A controlled experiment beats a correlation study, and both beat a survey of opinions.

The NP Digital grid is a survey. It measures what 100 experienced marketers believe after watching their own campaigns. That is worth reading. It is still belief rather than a measurement of engine behavior.

The rule cuts both ways. Plenty of confident figures circulate here with no method behind them. Claims that author credentials lift citations by 60%, or that 96% of citations carry strong E-E-A-T signals, trace back to vendor blogs citing each other. We could not find a primary study at the end of any of those chains.

Brand mentions vs backlinks across 75,000 brands

Ahrefs measured 75,000 brands and found branded web mentions correlate with AI visibility at 0.664. Backlinks reached 0.218, about a third as strong.

Ahrefs filtered for domains with a Domain Rating above 40, then scanned millions of AI Overview responses to count brand mentions. A December 2025 expansion widened it to ChatGPT and Google AI Mode, and added YouTube mentions, which scored highest at roughly 0.737.

This is the strongest corroboration of Patel's headline finding. Two teams, completely different methods, same ranking. One asked marketers what they had seen. The other counted what engines actually did.

Ahrefs states the limit plainly, and so do we. Correlation is not causation. Brands that get mentioned everywhere also tend to be bigger and better known.

Why third party citations dominate what engines quote

Muck Rack analysed more than 25 million cited links across ChatGPT, Claude and Gemini. Earned media supplied 84% of them. Paid and advertorial content supplied 0.3%. Across three editions since July 2025, that share held between 82% and 89%.

One study went further and tested cause. Stacker and Scrunch took eight articles and ran them in two conditions: hosted only on the brand's domain, and distributed across hundreds of third-party news sites. They measured 944 prompt and platform combinations across five AI platforms.

Same articles. Different addresses. Four times the citation rate.

One caveat before you quote a number. Meltwater analysed 5.35 million citations and put earned and news media at 39.5%, because it counts corporate sites separately. Both figures are defensible. Read the definition first.

Does schema markup help AI citations? A controlled test

Ahrefs tracked 1,885 pages that added JSON-LD schema and matched them against 4,000 control pages. Citations barely moved.

This is one of the few true experiments in the field. Louise Linehan and Xibeijia Guan found pages that added JSON-LD between August 2025 and March 2026. They matched each against control pages with similar prior citation levels, then measured 30 days before and after. The result confirms the low structured data scores in the NP Digital grid, by a completely different route.

Two limits matter. Every page in the sample already had more than 100 AI Overview citations, so the test only covers pages engines already knew. And Microsoft's Fabrice Canel confirmed in March 2025 that schema helps its models understand content for Copilot, while Google says no special markup is needed at all.

Our read: schema is hygiene. Ship it for entity clarity. Do not budget it as a citation lever.

Reddit AI citations explain the widest platform gap

NP Digital scored community engagement at 4.0 on Perplexity and 1.5 on Claude. Citation data explains that gap exactly.

Profound studied 680 million citations. Reddit was the single most-cited domain on Perplexity and on Google AI Overviews. ChatGPT's most-cited domain was Wikipedia. Measured as a share of each platform's top sources, Reddit reaches 46.7% on Perplexity and 21.0% on AI Overviews.

These weights also move fast. Semrush tracked 230,000 prompts over 13 weeks. ChatGPT cited Reddit in close to 60% of responses in early August 2025, then around 10% by mid-September.

One number is worth keeping. Only about 11% of domains are cited by both ChatGPT and Perplexity. A source that dominates one engine can be invisible on another.

What generative engine optimization research proves

Two peer-reviewed papers reach opposite conclusions, and the second one should change how you plan.

Aggarwal and colleagues published GEO at KDD 2024. They built a benchmark of 10,000 queries and tested nine content changes. Cite Sources, Quotation Addition and Statistics Addition each gained 30% to 40% on their main visibility metric. Keyword stuffing lowered citation rates.

Then Puerto and colleagues published C-SEO Bench at NeurIPS 2025. They tested the same kind of methods across six domains and 1,921 queries, and added something the first study did not model: several competitors adopting a tactic at once.

Most methods were largely ineffective, and several hurt ranking. As adoption rises, gains converge toward zero. Early movers win. The field catches up.

That is the deepest reason a published scorecard cannot run your programme. Any list of what works is a list of what worked before everyone read it.

An AI visibility plan ranked by evidence

Fund the signals with behavioral evidence behind them. Treat the rest as hygiene and spend accordingly.

Fund first. Earn third-party mentions and coverage. Two large independent datasets agree, and one controlled test took the same article from 7.6% to 34%.

Fund by platform. Community presence pays on Perplexity and pays little on Claude. Check which engines cite you before staffing a Reddit programme.

Ship because it is correct. Named authors, clear entity pages and clean schema are cheap and useful. Stop measuring them as growth levers.

Then measure your own brand. Only 11% of domains are cited by both ChatGPT and Perplexity, so an average across six engines describes none of them. Track citations per engine and per prompt, weekly, and tie them to CRM and billing. That is what citedby does across 10 answer engines.

AI trust signals: frequently asked questions