flareOps
Guide

How to check if ChatGPT recommends your store

Last checked: · Updated · By flareOps (how we write)

In short: Write 20–30 questions your real shoppers ask, without your store name. Ask each one in ChatGPT and Perplexity several times, in a clean session, and count how often your store is named or linked. Track that rate weekly, not the position: AI answers change almost every run. Add a few store-name questions to check that AI describes you correctly.

"Does ChatGPT recommend my store?" sounds like a yes-or-no question. It is not. The same question gives a different answer almost every time. This guide shows a free way to get a number you can trust, for any store on WooCommerce, Wix or a custom site. Facts were checked on 4 Nov 2026.

Why one test tells you nothing

SparkToro and Gumshoe ran 2,961 prompts in November and December 2025, with each test prompt repeated 60 to 100 times per platform. They found that ChatGPT and Google's AI returned the same list of brands less than 1% of the time, and the same list in the same order less than 0.1% of the time.

But one thing was steadier: how often a brand showed up across many runs. Strong brands appeared in a large share of answers even when the wording changed.

So:

  • do not measure your "position" in an AI answer;
  • measure your mention rate: the share of answers that name or link your store, over many runs.

(The study was part-sponsored by an AI tracking vendor and is not peer-reviewed. The authors said so.)

Step 1: Write the questions

Write 20–30 questions the way your shoppers would type them. Do not put your store name in them.

Mix three kinds:

KindExample (a store selling ceramic planters)
Product need"Best ceramic planter with a drainage hole for a small apartment"
Comparison"Glazed or unglazed ceramic planters for succulents, which is better?"
Problem"My plant pot has no drainage hole, what should I buy instead?"

Add the country if you only ship to one ("…in the UK"). Use your shoppers' language. If you sell in Hebrew, ask in Hebrew.

Then write 2–4 store-name questions in a separate list, such as "What is [store name] and is it reliable?". These do not measure recommendations. They check whether AI describes you correctly (price, location, what you sell).

Step 2: Ask in a clean session

Personal context changes answers. OpenAI says ChatGPT considers Memory and custom instructions when it shops. So:

  • use a logged-out or temporary chat, or turn Memory off;
  • do not ask follow-up questions in the same chat;
  • ask the same day each week.

Ask each question at least twice in ChatGPT and twice in Perplexity. That is 80–120 answers a week for a 20–30 question set. A spreadsheet is enough.

Step 3: Decide what counts, before you look

Write the rule down first, so you are not tempted to count near-misses:

  • Counts: your store name, your web address, or a link to your site.
  • Does not count: a similar product name from another shop, or your name only because the question contained it.

Record for each answer: question, engine, date, found (yes/no), how (name, address, link), and which other shops or sites were named. The "who else" column is gold. It shows who AI trusts for your products and which websites it reads.

Step 4: Get a starting number before you change anything

Run the full set once or twice before you fix anything. This is your baseline. Most small stores start at or near 0%. That is normal and not a failure.

Step 5: Compare with your baseline, honestly

Ask the same questions every week and compare the rate with your baseline. Be honest about what that shows: if your rate goes up, you cannot be sure whether your work did it or whether the AI changed that week.

  • Look for a rise that holds for several weeks, not one good week.
  • Note the dates when you made changes, so you can see what came before and after.
  • Write down known AI updates (a new model, a new shopping feature) next to your numbers.

If you want stronger evidence, you can leave 2–3 products unchanged for a while and compare. Most small stores would rather fix everything, and that is a fair choice.

Step 6: Add the free first-party data

Your weekly test is a sample. Real data adds context:

  • Bing Webmaster Tools → AI Performance. In public preview since February 2026, it shows how often your pages are cited in Copilot and Bing AI answers, which pages, and sample "grounding queries". No clicks. Verify your site and import from Search Console to set it up.
  • Search Console. Since June 2026 it has a generative AI report (still rolling out) that shows impressions in AI Overviews and AI Mode, but not clicks or queries. Clicks stay inside the normal Web report.
  • Analytics. Links from ChatGPT search usually carry utm_source=chatgpt.com. Perplexity shows as perplexity.ai.
  • Server or plugin logs for OAI-SearchBot, ChatGPT-User, PerplexityBot. A ChatGPT-User visit means a real user's chat opened your page.

Step 7: Read the trend monthly

Look at the four-week average, per engine. Expect slow change. Ahrefs found that Perplexity links a brand it mentions up to about half the time, against about 10% for Google AI Overviews, so links may show up on Perplexity first.

How flareOps automates this

flareOps runs exactly this method every week, so you do not need the spreadsheet:

  • Claude writes about 30 shopper questions in your store's language and market, plus 2 store-name questions;
  • ChatGPT and Perplexity answer each one twice a week through their official APIs;
  • a baseline run starts before any change;
  • a store counts as found only by name, web address or link, never by a similar product name, and not when the question already held the name;
  • every product is improved, so the report is a before and after: the rate per week against your baseline, the full answers and who AI named instead.

It does not run Gemini: its terms forbid analysing search-grounded answers. The store check is free; the weekly AI check is part of Pro at $29 per month per store.

Questions people ask

Why does ChatGPT give a different answer every time?

AI answers are generated fresh each time and can search the web differently on each run. A SparkToro study found ChatGPT and Google's AI returned the same brand list less than 1% of the time for the same prompt. So measure a rate across many runs.

Should I include my store name in the question?

Not in the main set. A question that contains your name will usually echo it back, which proves nothing. Keep a few separate store-name questions only to check that AI describes your store correctly.

How many times should I ask each question?

At least two runs per question per week, across 20 or more questions. More runs give a steadier number. Judge trends over four weeks or more, not one week.

Does a similar product name count as a mention?

We say no. Many shops sell similar items. Count it only when the answer names your store, shows your web address or links to your site.

Can I see when Copilot cites my site?

Yes. Bing Webmaster Tools has an AI Performance report, in public preview since February 2026, that shows how often your pages are cited in Copilot and Bing AI answers. It does not show clicks.

Why not just use a tool's AI visibility score?

A single score hides how it was counted. Ask any tool which questions it asked, how many times, what counts as a mention, and what it compares with (a baseline taken before any change). If it cannot say, the score may be luck.

Is it against the rules to automate these checks?

Using the official APIs within each provider's terms is fine. Scraping the consumer apps may break their terms. Some providers, such as Google's Gemini, do not allow analysing their search-grounded answers at all.

Sources

  1. Search Engine Journal: SparkToro study, AI recommendations change with nearly every query (Jan 2026)
  2. Search Influence: Inside Bing's AI Performance report
  3. Ahrefs: AI citations vs mentions study
  4. Ahrefs: self-promotional content experiment (Jul 2026)
  5. Google Search Central: AI features and your website (Search Console reporting)

Sources checked on 4 November 2026.