Who Does AI Recommend When Your Customers Ask?

We ask Perplexity (sonar-pro) and OpenAI web search (the Responses API benchmark) four real buying questions about your category and city, for example who to hire, who is most trusted, and who is best for a specific need. Each question is asked twice: the way a buyer would phrase it, and again with an instruction to rank recommendations in order. Every question runs twice on each platform, so a single lucky or unlucky answer cannot decide the result. That is 32 measured answers plus 2 brand-recognition answers, all shown verbatim with their sources. Google AI Overviews and Gemini are not part of the free check; the full assessment covers them.

Yes, it is completely free. There is no credit card required, no obligation, and no hard sell. Each check costs us real API calls, so it is limited to one run per business per day. The check tells you who wins the AI recommendation on four questions today. It deliberately does not tell you why, because answering that responsibly is the full assessment: 30 or more questions across four platforms with confidence intervals and a scored diagnosis of your site and off-site authority.

The Observed Recommendation Placement Score uses mean reciprocal rank at five, a standard ranking measure. When AI is asked for an ordered list, being placed first scores 1.0, second 0.5, third 0.33, fifth 0.2, and absent or beyond fifth scores 0. The score is the average across ordered answers that parsed, times 100. Bands are LEADING (60 and up), COMPETITIVE (30 and up), PRESENT (12 and up), MARGINAL and ABSENT. They are internal descriptive bands, not industry percentiles. Brand recognition (can AI describe you when asked directly) is reported as a badge and never counted in the score.

A failed call or an answer with no parseable ordered list is excluded from every number and disclosed on the results page. It is never counted as a miss. If too few ordered answers survive, the score is withheld for that run rather than published on thin data, and the mention results still stand.

Every answer is logged with the exact question, the platform, the model, the timestamp and the sources it cited, and shown verbatim on your results page. You can open any question in Perplexity or ChatGPT with one click and read the answer yourself. AI answers vary between runs and phrasings, which is exactly why each question is repeated and why the results report how stable the outcome was across repeats.

ChatGPT personalizes responses based on your account history, memory, and prior conversations. When you are logged in, you may see results that reflect what ChatGPT knows about you, not what a stranger sees. The check queries Perplexity and OpenAI web search through the API in clean sessions, the unbiased view a brand-new prospect gets. OpenAI's API web search is related to but not identical to the consumer ChatGPT product. To verify a finding, sign out of ChatGPT or open an incognito window and re-run the question.

We send your results to your email and may follow up once to see if you have questions. That is it. We do not sell your information, we do not add you to a marketing blast list, and we do not share your data with anyone. If you want to hear from us, you will need to reach out.

Start with the two lists on your results page: who AI named instead of you, and which sources it built the answers from. Those sources are where the recommendation is decided, and a business that is absent or thin on them rarely gets named no matter how good its website is. The full assessment turns that into a prioritized plan. Book a free 15-minute walkthrough and we will read your results with you.

To keep it free, we limit checks to one run per business per day. If you want a deeper measurement, book a free strategy call and we will scope the full assessment.