Free, no sign-up to see the rate
Most tools ask once. We ask each buying question five times, on ChatGPT, Perplexity, Claude or Gemini, and show how often you are named, where you sit in the list, who gets named instead and which pages the answers cite.
Up to 25 answers per assistant per run, about 30 to 60 seconds. Three runs a day per visitor. Email unlocks every answer and a link you can share.
Example of the method. Not a client result.
Ask an assistant the same buying question twice and you get two different lists. The brands move, the order moves, the cited pages move. A tool that asks once and prints a score is reporting a coin flip to two decimal places.
Five runs per question give you a rate and a range. The report also tells you how many runs it took for your rate to stop moving, so you know whether to trust it or run it again with more questions.
Per assistant and per question, with the range around each rate and your average position in the list when you do appear.
Every other brand the answers recommended, ranked by how often, so you know who you are being compared with in the assistant's head.
The domains behind the recommendations: review sites, comparison posts, your competitors' own pages. That list is your off-page to-do list.
Then fix the on-site part. The site audit checks whether AI crawlers can read you at all, and the product page check tests whether an assistant can read your price, stock and shipping. Both are free.
Run the free check first. On its report page you can buy a monthly or a weekly plan: the same questions re-run on a schedule with 13 runs each across ChatGPT, Perplexity, Claude and Gemini, every answer unlocked, and an email that says what moved and whether the move is bigger than the noise. Pay once for the period, no card on file.
Because assistants do not give the same answer twice. The brands named, their order and the pages cited change from run to run. One run can show you at 100% or at 0% by luck. Five runs per question give a rate with a range around it, and the report tells you how many runs it took for your rate to stop moving.
ChatGPT through the OpenAI API with web search on, Perplexity through its Sonar API, Claude through the Anthropic API with web search, and Gemini through the Gemini API with Google Search grounding. These are the same models behind the consumer apps, asked with a plain buying question and no system prompt that favours anyone. If an assistant answers without web search, the report says so, because an answer from memory is not what a shopper sees today. Results can differ from what a signed-in user with history sees.
We look for your brand name (spacing and punctuation ignored) and your domain in every answer. Position is where you appear in the list of brands the answer names. Other brands are pulled from bold names and list items, which is how assistants write recommendations.
Yes. Three runs a day per visitor, up to five questions and five runs per question per assistant. The mention rates, your average position and the top three brands named instead are free. Your work email unlocks every answer, the full competitor table and the cited pages, and gives you a link to share.
If the assistants cite review sites, directories or comparison posts, get your brand onto those pages. If they cite your competitors' own sites, check that yours is readable: run the free site audit and the product page check here, which test crawler access, structured data and whether prices and stock are in the HTML. The agentic SEO audit from ROAS Expert does both for the whole catalogue.
The ROAS Expert agentic SEO audit runs your real buyer questions at higher volume across all three assistants, reads the cited pages, and ends with a plan for the on-site fixes and the off-page mentions that move the rate. $1,500, two to three weeks.
.svg){kind=small}