Yext has published research suggesting that the choice of generative AI platform influences which sources appear in answers. Major models show distinct citation patterns rather than drawing from a shared pool of information.
The study analysed 17.2 million AI citations across OpenAI, Google's Gemini, Anthropic's Claude and Perplexity. Yext described the differences as separate "information personality" profiles that shape responses for users researching, comparing and seeking recommendations.
According to the analysis, each platform tends to favour a particular mix of sources. Depending on whether a model leans toward official brand information, customer reviews, news coverage, forums or structured reference material, the resulting answer can change materially.
Four sourcing styles
Gemini showed the strongest preference for content that brands control or influence. In the dataset, 93% of Gemini citations came from sources organisations manage directly or indirectly, including 51% from first-party websites and 42% from third-party business listings. In local search results, 3% of citations came from news outlets or online forums.
The pattern resembles conventional search behaviour that ranks structured, verifiable information highly. It also suggests Gemini is less likely than other models to draw on independent reporting or consumer discussion when compiling a response, based on the citation mix.
Claude relied more heavily on user-generated material. Reviews accounted for 15% of its citations-two to four times higher than Gemini, OpenAI and Perplexity. Websites and listings still made up most of Claude's citations at 81%, but customer sentiment and broader online discussion appeared more often in its citation base.
Perplexity showed the most consistent citation behaviour across industries. Brand-owned websites represented between 37% and 50% of citations in most sectors examined. The study also characterised Perplexity as favouring structured, "answer-ready" sources. Its sourcing patterns varied less by topic than other models, producing a steadier citation profile across categories.
OpenAI showed the largest swings in citation patterns between industries. In Food and Beverage, first-party websites represented 28% of citations; for non-profit organisations and religious institutions, the figure was 44%. Hospitality stood out for heavier use of official sources: OpenAI cited hotel websites 38% of the time in that segment, compared with 17% for Perplexity and 22% for both Claude and Gemini.
Owned content rises
The research also points to a broader shift in what AI platforms cite. Yext compared the results with an earlier analysis based on 6.8 million citations. In the newer study, citations from brand-controlled sources, including websites and business listings, rose to 90% from 86%. Over the same period, citations from reviews fell to 5.5% from 8%.
The figures suggest owned digital assets are playing a growing role in AI-generated responses. The trend could influence how organisations prioritise updates to websites and business listings, particularly if AI systems treat structured information as easier to cite.
Fragmented visibility
Yext argued that differences between platforms make AI visibility harder to standardise. A brand that appears in answers from one model may not appear in another if each platform draws on different source types. Yext framed this as a shift away from a single optimisation approach toward a model-by-model assessment of where information is published and how it is structured.
Model-specific citation behaviour can also shape competitive comparisons. If one model relies more on reviews, organisations with stronger ratings and broader review coverage may be surfaced more often. If another prioritises official sources and listings, brands with accurate, complete profiles across directories and first-party properties may benefit.
The study linked these patterns to differences in user expectations and search behaviour. Some users prefer official, authoritative information, while others place more weight on public sentiment or traceable citations. The data suggests the AI platform chosen can determine which perspective is emphasised in the response.
Yext's findings come as AI tools increasingly sit alongside traditional search and discovery channels. Citation trends matter for publishers, marketplaces and brands because they affect which sites and documents gain visibility, and which sources different models treat as credible references.
Sam Davis, VP of Global Solutions Engineering at Yext, said: "The more we analyse AI citations, the clearer it becomes that visibility strategies must be model-specific. The sources that make a brand visible in Gemini are not the same that will drive visibility in Claude, so we can't look at AI search as a 'one and done' strategy. Success isn't about general ranking, it's about being cited in the right places."