AI Search

AI search supports conversational questions and follow-ups. A person may describe a situation ("my AC died and it's 105 out, who's good near me?") instead of typing keyword fragments. Complete, current, and specific business information gives the system evidence it can repeat.

AI answers often present fewer options than a traditional results page. An answer may name two to five businesses, while a search page lets the user review more listings. A business excluded from the answer may receive no visibility during that interaction.

Across 540 answers in the Routeless study, assistants named about 47 different businesses per market, but roughly 60% appeared only once in the sample. Repeated measurements are needed to distinguish frequently recommended names from one-time mentions.

Different products also require separate measurement. ChatGPT, Perplexity, and Google's AI features use different retrieval systems, interfaces, and source displays. A business visible in one may be absent from another. Keep the prompt wording, location, date, search setting, and number of runs attached to each result instead of combining unlike samples into one rank.

Archive the raw answer and its citations because a later run may use different wording or sources.

Measure AI search by asking representative buyer questions, recording the answers, and repeating the process. Traditional rank and impression reports do not fully describe these answers, and repeated samples reduce the effect of non-determinism.

Do people use ChatGPT to find local businesses? reviews the adoption evidence and its limits.