Cited or Invisible? A Multilingual, Multi-Platform Audit of Brand Citation in Generative Search

Abstract

Generative search engines increasingly answer commercial questions directly, citing a handful of sources rather than returning a ranked list. Which businesses those systems cite is becoming consequential for market access, yet the question is mostly discussed through vendor marketing rather than measurement. We report two audits of brand citation across four generative answer surfaces (Google AI Overviews, Gemini with search grounding, Perplexity, and ChatGPT with web search) in six European languages. Prompts are authored natively in each language rather than translated, because translation measures the translator and not the market, and each is issued three times per platform. Study 1 covers five commercial verticals with a set of replica furniture retailers as focal brands (4,320 observations); Study 2 covers authenticity-seeking queries with eight design rights holders as focal brands (1,080 observations). Neither run had a failed call. Four findings. First, the platform a question is asked on dominates the language it is asked in, and the effect replicates across two independent brand sets: citation rates run 16.2% to 91.2% across platforms in Study 1 and 6.3% to 63.3% in Study 2, with the same platform ordering in both, while the spread across six languages is 41.0-61.1% and 31.1-46.1% respectively with intervals that overlap almost throughout. Second, we show by resampling how readily prompt selection manufactures a language gap: an eight-prompt single-pass study reports a difference of at least 20 points 16.4% of the time when the corpus-wide difference is 8.3 points, and 12.5% of the time on a platform where the true difference is exactly zero. Our own pilot produced a 38-point gap by this mechanism. Third, repeated identical queries disagree with themselves at rates differing by more than an order of magnitude across platforms, and that ordering also replicates. Fourth, on queries explicitly seeking the authentic article, Google AI Overviews cite the actual rights holder in 6.3% of answers and produce an overview at all for 30% of queries; a replica retailer is the fourth most-cited domain, ahead of four of the eight rights holders. The measurement harness, prompt corpora and observation logs are released. Total data-collection cost for the reported studies was $57.65. Four domains are pseudonymised as R1-R4: they are replica retailers operated by the author, redacted for legal reasons, and this is disclosed in the paper because R1 is the fourth most-cited domain in Study 2. No reported quantity depends on the identity of those domains. Companion methods paper: doi:10.5281/zenodo.22480733.