Developing
A vendor study. Yext sells listings and reputation tools, which are what its recommendations favour.
In 60 seconds
- Yext's study, published June 15, 2026, analysed 17.2 million AI citations across ChatGPT, Perplexity, Gemini and Claude.
- It found verified, structured, directly distributed data made up 54.53% of distinct citation sources; websites produced 4.31 citations per URL.
- ChatGPT cited hospitality websites 38.08% of the time; Claude cited user-generated content at 2 to 4 times the rate of other engines.
- Yext recommends treating directory listings as visibility infrastructure.
Impact Layer
- What changed
- Large-scale data on where AI engines get their citations, by engine and industry.
- Why it matters
- It suggests brand-owned pages are only part of what answer engines cite, so listings, reviews and structured data matter.
- Who is affected
- SEO, local and brand teams, particularly in retail, hospitality and food and beverage.
- Martech stack area
- AI, SEO & Search
- Stack impact
- Audit your listings and reviews alongside your own pages, and test your key questions in each engine.
- Business impact
- Moderate Yext sells listings and reputation management, which its recommendations favour. The causal link to Claude's design is described by Yext as observed, not proven.
- Related technologies
- Related vendors
- What to watch
- Independent replication, and whether engines change source patterns as products evolve.
Yext published a study on June 15, 2026 analysing 17.2 million AI citations across ChatGPT, Perplexity, Gemini and Claude.
What it found, per Yext
- Across engines: “Verified, structured, directly distributed data was 54.53% of distinct citation sources.” Listings dominated, while websites generated only 4.31 citations per URL.
- Gemini relies heavily on Google’s search index and favours official websites and established sources with strong SEO foundations.
- Perplexity shows consistent patterns across industries, drawing on official sites and directories.
- ChatGPT varies by industry: hospitality websites received citations 38.08% of the time, about double other models’ rates.
- Claude cited user-generated content at two to four times the rate of competitors, and almost ten times the rate of Gemini in food and beverage. Yext notes that linking this to Claude’s design is an observation, not proof.
Yext’s recommendations
Map content strategy to each model’s behaviour; treat directory listings as visibility infrastructure; include reputation management; structure website content for clarity and verification.
Our read
The study is useful because it shows that engines differ: what helps in Gemini is not what helps in ChatGPT or Claude. It is also a vendor study from a company that sells listings and reputation tools, so its conclusions point at its own products. The safe use is as a checklist: confirm your business data is accurate wherever engines might read it, and test your own key questions in each engine instead of assuming one pattern.
Sources
- Yext: How ChatGPT, Perplexity, Gemini and Claude actually decide what to cite (opens in a new tab)
Source of the dataset, findings, quotation and recommendations.