Ranked is not cited: what 240 brands show about AI answers

The average brand shows up in about a third of the AI answers where it should.

Sourceable tracked 240 brands across 1,800 category prompts on ChatGPT, Gemini, Claude and Perplexity, from March to May 2026. The average brand appeared in 34% of relevant answers. By engine, ChatGPT mentioned brands 41% of the time, Gemini 36%, Perplexity 33% and Claude 29%.

The engines also disagree with each other. A brand named first on Perplexity appeared in only 42% of equivalent ChatGPT answers.

71% of mentions came in answers that sent no traffic. 9% of mentions carried outdated or wrong information. Brands that were cited often held their position 2.3 times more steadily over time.

Two cautions. Sourceable sells a tracking tool, so it has a stake in the result. And the study shows correlation, not proven drivers.

What I take from it is that ranked and cited are different problems. At Iterate we built AVA (AI Visibility Architecture) around the second one: structured data, entity optimisation and llms.txt, so brands get cited inside the answer, not just ranked.