RankedByAi.dev

how to check if ChatGPT recommends your product (step by step)

A practical way to see whether ChatGPT, Claude, Perplexity, Gemini and Grok name your product when buyers ask for a recommendation, and what to do with what you find.

· 3 min read · by the RankedByAi.dev team

more buyers now skip the search results page and ask an AI assistant instead: "what's the best invoicing tool for freelancers?", "alternatives to [the tool I use]", "which CRM is simplest for a two-person team?". the assistant answers with a short list of names. if yours isn't on it, that buyer never hears of you, and nothing in your analytics tells you it happened.

so the first question is simple: when your buyers ask, does the AI name you? here's how to find out properly.

1. write down the questions your buyers actually ask

don't start with your brand name. start with the situation a buyer is in before they know you exist. three kinds of question cover most of it:

  • category questions: "best [category] for [audience]", like "best uptime monitor for small SaaS teams"
  • alternative questions: "alternatives to [a well-known competitor]"
  • problem questions: "how do I [the job your product does] without [the usual pain]?"

aim for five to ten questions, written the way a person types them, not the way your homepage describes you.

2. ask more than one AI

ChatGPT, Claude, Perplexity, Gemini and Grok don't read the same pages or rank things the same way. a product can be named every time on one and never on another. ask each question on each assistant you care about, with web search on where there is a switch for it: that is how most buying questions get answered today.

3. ask more than once

answers change between runs. the same question can return three names one minute and five the next. ask each question at least twice, and treat "named once out of two" differently from "named both times".

4. record three things for every answer

  • were you named? yes or no.
  • where? first, third, or "also worth a look" at the end.
  • who was named instead? these are the competitors the AI trusts for your category, which may not be the ones you think.

if the assistant shows its sources, write those down too. they tell you which pages it read to build the answer, and those pages are where the work is.

5. ask about yourself by name

finally, ask "what is [your product]?". this is a different test: not whether you're recommended, but whether the AI knows what you are. a wrong or vague answer here usually means your own site doesn't say it plainly enough, or the pages that mention you describe you differently.

what the results usually mean

  • named on some AIs, not others: the AIs that miss you read different sources. look at the sources the others cited.
  • never named, competitors always named: you're missing from the lists, reviews and comparison pages the AIs lean on. see how AI search picks its sources.
  • named, but described wrong: fix how your own site and listings describe you first. see when AI describes your product wrong.
  • crawlers can't read your site: check your robots.txt with the free AI crawler checker.

doing it without a spreadsheet

doing this by hand is slow: ten questions, five assistants, twice each is a hundred answers to read and tally. that's the job RankedByAi.dev does for you. the free scan asks your main search on four AIs and shows whether you're named, who is named instead and the first thing to fix. the full report runs twenty searches on all five AIs, twice each, and turns the answers into a score and a fix list.

whichever way you do it, do it before you change anything, so you have a baseline to compare against.

keep reading