Skip to content

Guide: What is an AI search audit?

What an audit covers, how it is built, and what it is for.

In two paragraphs.

An AI search audit reads one market the way a buyer moves through it. It finds the questions customers ask before they buy, asks them of ChatGPT, Gemini, Perplexity and Google the way customers do, and records who is named, how each business is described, and which sources the answers cite. Some call it an AI visibility audit. The result is a picture of where you win, where a competitor wins, and why.

It ends in a ranked list of what to change: on your website, in your content and product information, and in the sources that describe you. An audit is a reading of one market on one date, so it is built to be repeated, and the second reading is what proves the change.

Six readings of one market.

  • Demand mapping: the questions customers ask and the decisions underneath them, grouped and ranked by how much each matters commercially.
  • AI search and recommendations: how Google, ChatGPT, Gemini and Perplexity answer those decisions, which brands they name, and which sources they cite.
  • Competitor intelligence: who owns which decision, how each competitor is described, and where they are ahead of you.
  • Buyer decision research: how customers compare and choose, in their own words, from search, communities, reviews and your own records.
  • Information consistency: whether your products, prices and claims read the same on your site, on your listings and in the answers AI gives about you.
  • Agentic commerce: how AI assistants and shopping agents find, compare and present products, and whether yours can be found and chosen.

The questions come from customers. The answers are asked for several times.

The questions are written from what customers say, in volume. For our perfume study in the United Arab Emirates we collected 8,102 shopper posts, comments and search suggestions and wrote 100 questions from the patterns in them, 50 in English and 50 in Gulf Arabic. For the running-shoe study in Lebanon, 3,666 buyer comments, posts and searches became 105 specific buying situations.

Each question is then asked of each assistant several times, in a fresh conversation each time, with web search on and the market as the location, because answers vary from one run to the next. The perfume study asked each question five times on each of three assistants, 1,500 answers. The running-shoe study asked 38 decisions three ways each on three systems, 342 answers. A brand's share is then a share of questions, with the number of questions beside it.

What the two published studies read, and how much
StudyEvidence collectedQuestionsAI answers recorded
Perfume, United Arab Emirates, August 20268,102 shopper items1001,500
Running shoes, Lebanon, September 20263,666 buyer items38 decisions, asked three ways342

Audit, strategy, recommendations.

  • The audit: where you win and where you lose across the decisions that matter, with the figures and their denominators.
  • The strategy: which decisions to go after first, weighed by commercial value and by how realistically each can be won.
  • The recommendations: a list of what to change, in order, across SEO, content, sources, product information and the technical reasons an assistant cannot read you.

A tracking tool shows where you appear, how often, and against whom. The audit shows which customer decisions matter, who is ahead on each one, and which changes to make first. Tracking can be part of the work; the audit is what says what to do.

Asked often, answered here.

  • How long does an AI search audit take?

    One market or category, from the evidence to the ranked list, takes a few weeks: the collection and the question writing come first, then the runs, then the reading. The question map on this site is the first minute of it.

  • What is the difference between an AI search audit and an SEO audit?

    An SEO audit reads your pages and their rankings. An AI search audit reads the answers: what ChatGPT, Gemini, Perplexity and Google say when your customers ask, who they name, and which sources they cite. It includes the SEO reasons, because a page that cannot be found cannot be read.

  • Does the audit cover Google's AI Overviews?

    Yes. The same decisions are read in Google search, where AI Overviews answer a share of queries before any result is shown, and in the three assistants.

  • What does an audit cost?

    It depends on the market and on how many decisions it holds. Senpai.ai scopes it on a call, after the question map has shown what the market looks like.

What are your customers asking AI?

They ask ChatGPT what to buy and who to trust before they ever reach you. Type your website and see the questions they ask, and who AI names instead of you.

No website? Start with your business name. Free. One minute. No email to start.

People ask AI things like

  • Where can I try on running shoes in Dubai today?
  • Which running shoes stay cool in very hot weather?
  • How do I know my running shoe size?

Find what's worth changing next.

Bring the market, the product or the question. We'll start there.

Or start with the question map: see what your customers ask AI.