What We've Learned About How AI Engines Choose Brands
Patterns from measuring thousands of queries across ChatGPT, Google AI, Perplexity, and Claude. What actually drives AI recommendation rates, and what doesn't.
Why the #1 SEO result is often not the #1 AI recommendation
AI engines don't mirror search rankings. The signals that earn a brand a recommendation are structurally different from the signals that earn page-one placement, and the gap between them is measurable and consistent across categories. A brand can hold the top organic position for a category keyword and still be absent from every AI recommendation on the same query. The reverse is also true.
The core difference is what each system is optimizing for. Search engines rank pages. AI engines name brands. The inputs that drive each outcome overlap partially but diverge significantly in the signals that carry the most weight.
The two characteristics shared by brands AI engines consistently recommend
Across every category we've measured, the brands that earn the highest AI recommendation rates share two structural properties. Neither is brand size. Neither is marketing spend.
The first is direct Q&A content: pages that answer the exact questions buyers ask AI engines, written in the same language and framing buyers use. The second is consistent off-site presence: editorial mentions, review platform profiles, and third-party citations that appear across the sources AI engines index most heavily. Brands that have both get named. Brands that have one or neither don't, regardless of traffic or domain authority.
Why fixing schema and fixing content move different query types
Schema improvements and content improvements both raise recommendation rates, but they affect distinct query types at distinct funnel stages. Schema fixes tend to move brand and comparison queries first: queries where the buyer already knows your name and is evaluating whether to shortlist you. Content fixes tend to move discovery queries: queries where the buyer is asking which brands to consider and yours needs to be surfaced unprompted.
Understanding which to prioritize first depends on where you're losing, not on a general best-practice checklist. A brand losing primarily on discovery queries should build buyer-question content before investing in schema. A brand already being discovered but losing at comparison should do the reverse.
How Indian-market queries expose gaps that English-only measurement misses
AI engines trained on predominantly English data show measurable, consistent patterns when answering queries in Hindi, Tamil, Kannada, and other Indian languages. Brands that track both surfaces see a different and often more actionable gap than brands that only track English queries.
The most common pattern: a brand that is reliably recommended in English-language queries for its category is absent or inconsistent in the equivalent Hindi or Tamil query, even when the brand is well-known in that market. The gap is not awareness. It is structured content and citation presence in the languages and sources AI engines weight for regional queries. This is a fixable signal gap, not a brand recognition problem.
What third-party citations actually signal to AI engines, and which sources carry weight
We've tracked citation impact across engines and categories over time. The source types that move recommendation rates are fewer than most teams expect, and several widely targeted platforms have minimal measurable effect on AI recommendations while consuming significant effort.
The sources that consistently carry weight share a common property: AI engines treat them as editorially credible rather than user-generated or paid. Review platforms where editorial curation is part of the product model tend to carry more weight than open-submission directories. Independent editorial roundups on established publications tend to outperform sponsored placements on the same publications. The pattern holds across categories and engines.
The difference between being mentioned and being recommended
A brand can have strong editorial presence, appear across dozens of sources, and generate consistent mention volume and still lose recommendation rate to a smaller competitor with a narrower but more structured footprint. Mentions and recommendations are related but not the same thing.
AI engines are not counting mentions. They are assessing whether a brand is the answer to a specific question from a specific type of buyer. A brand mentioned frequently in general coverage but rarely cited in direct response to buyer questions will have high mention volume and low recommendation rate. The brands that close the gap between mention and recommendation are the ones that structure their content and citations around the exact queries buyers are asking.
How recommendation rates decay when teams stop shipping fixes
Brands that improve their recommendation rate and then stop shipping fixes see it decline in a predictable pattern. The decay is not immediate. In most categories, a brand that reaches a materially improved recommendation rate will hold that position for four to eight weeks before the decline becomes measurable. After that, the rate drops toward baseline as competitors keep shipping and AI engines re-weight their sources.
The implication is that recommendation rate is not a position you win once. It is a rate you maintain through consistent work. The minimum threshold to hold a position varies by category competitiveness, but in every category we've measured, stopping the work entirely leads to measurable decay within one to two re-crawl cycles.
See where AI recommends your brand - and where it doesn't
Enter your website and get a full recommendation rate report across ChatGPT, Perplexity, Google AI, and Claude. Takes 2 minutes.
Get your free report →No credit card required