How to check if ChatGPT recommends your business

Typing one question into ChatGPT and looking for your name tells you very little. Answers change from week to week, differ between engines, and depend on exactly how the question is worded. A single result can make you feel safe or worried for no good reason.

It matters because buyers have changed where they start. G2's April 2026 survey of software buyers found 51% now begin research in an AI chatbot more often than in Google. The method below gives you a baseline you can trust and compare over time.

Write 10 to 15 questions in your customers' words, not yours. Good places to find them are sales-call notes, support tickets, and the Queries report in Google Search Console.

Mix four kinds of question:

• Category: "best invoicing software for small agencies"
• Comparison: "alternatives to [your biggest competitor]"
• Problem: "how do I stop clients paying invoices late?"
• Local or niche: "accountant for startups in Manchester"

Leave your own brand name out. You want to know whether the assistant suggests you unprompted, which is what happens when a buyer hasn't heard of you yet.

Run every question in ChatGPT, Gemini, Claude and Perplexity. Turn web search on wherever the assistant gives you the choice, because that's the answer most buyers see.

Keep the test fair:

• Use a fresh or temporary chat, so your past conversations don't steer the answer.
• Paste the exact same wording every time. Small changes produce different answers.
• Ask once and don't follow up with "what about [your brand]?". That's not what a buyer does.
• Note the date. You'll compare against it later.

Set up a spreadsheet with one row per question per engine, and these columns:

• Date and engine
• Named? Yes or no
• Position: first recommendation, somewhere in a list, or only in the sources
• Competitors named: every other business the answer mentions
• Sources cited: the links the assistant shows

Your visibility score for each engine is simple: the share of answers that name you. If you're named in 3 of 15 answers on Perplexity, your score there is 20%.

Run the same questions on the same day each week for at least four weeks. One week is a snapshot; four weeks shows a trend, and you'll start to see which engines favour which competitors.

The sources column is where the useful work is. It shows which review sites, comparison articles, directories and forum threads each assistant trusts for your category. If competitors appear in a "best tools" article or a Reddit thread and you don't, that's your to-do list:

• Get listed on the review and directory sites that keep appearing.
• Contact the authors of comparison articles that leave you out.
• Answer real questions in the forum threads that get cited, as yourself.
• Make your own site state plainly what you do, who it's for and what it costs. Assistants quote pages that answer directly.

• Including your brand name in the question
• Testing only ChatGPT, when buyers use several assistants
• Judging on one run instead of a four-week trend
• Testing in a long-running chat that already knows about you
• Rewording questions between weeks, so you can't compare results

If you'd rather not run this by hand every week, Mentioned asks your questions across all four engines each week and emails you the score, the competitors named and the sources cited. You can also read our plain-English definitions of AI visibility and generative engine optimisation.