What the scan checks — and what it cannot prove.

1. Start with your business

You enter a public website, one city, and one service. Before provider requests run, the scanner reads the public home page to identify the business; if it cannot read a name, it stops before spending on a scan.

2. Collect repeated local answers

The scan records 27 AI-search requests: nine Google AI Mode requests through SerpApi, nine Perplexity Sonar requests, and nine OpenAI Responses API requests with web search. It also makes one Google local-business-listing request to establish the only candidate set from which a competitor may be named.

Google requests are fresh SerpApi requests, meaning SerpApi refetched rather than served its cache. That does not establish that Google generated each answer independently. The OpenAI method is an API measurement, not what a person sees in consumer ChatGPT.

3. Keep evidence separate from inference

The report distinguishes business mentions, cited source pages, local-listing candidates, and unavailable requests. A method that could not be completed is shown as unavailable; it is not treated as evidence that a business was left out.

4. Point to one useful next action

When the collected evidence supports it, the report identifies a first documentation or crawlability gap to investigate. It does not prescribe a rebuild because of a website platform, and it does not guarantee an AI mention or customer outcome.

Report availability and privacy

Reports use an unguessable link and are retained for seven days. They are excluded from search indexing, but a link is not the same as an account login; do not share a report publicly unless you are comfortable sharing its contents. Email is optional, unlocks nothing, is retained for 30 days when consented to, and does not send a report automatically.