How to check what ChatGPT and Perplexity say about your store
To check what ChatGPT and Perplexity say about your store, write down the questions a buyer would ask, put each one to both assistants in a setup that is as neutral as you can make it, and save what comes back. ChatGPT and Perplexity both show where an answer came from, so the check is less about guessing and more about reading. The sections below cover why your result may differ from someone else's, how to read the sources, and how to turn one look into a baseline you can repeat.
Start with a plain list. Write the questions a shopper might type, such as the kind of product you sell, a problem it solves, or a comparison between types of product. Keep the wording fixed. If you rephrase a question each time, you are testing your phrasing, not the assistants.
Run each question in both assistants. Give each answer its own row: question, assistant, date, market, store mentioned and page linked. That makes it possible to compare what ChatGPT and Perplexity say about your store even when one answer mentions the brand and the other links directly to a product. For more on that distinction, see how to measure product visibility in AI.
Why two people get different answers
Two people can get different answers to the same question because both assistants can use personal context. If memory is enabled, ChatGPT may use relevant saved memories when rewriting a search query. ChatGPT may also use an approximate location based on the user's IP address to provide relevant local results. Perplexity may draw on a user's memories and search history depending on the question.
That is why your own account is a poor test of what a stranger sees. If you have talked about your own store with these assistants before, that history can shape the reply.
To reduce it, use the neutral modes each product offers. In ChatGPT, a temporary chat is set to Personalized or Unpersonalized before the first message, and neither option creates or updates memories. Pick Unpersonalized for your check. In Perplexity, memories and search history are always off in incognito mode, so use incognito. Location can still apply, so note the place you tested from alongside each answer.
This does not make results identical for everyone. It makes your result fairer and easier to compare with your own later runs.
Read the citations and check your crawler access
The sources attached to an answer tell you where the assistant looked, so read them before the prose. ChatGPT may search the web automatically when a question would benefit from current information. Its web answers may include citations, and Sources, when available, shows cited sources and other relevant links. Perplexity says each answer includes numbered citations linking to the original sources.
Open the cited page. An assistant might name your brand but send the buyer to another retailer. Record the destination beside the mention, as in this walkthrough of which retailer gets the link. If the link is yours but the answer quotes the wrong price or size, compare the visible offer with the structured data for that variant. Our guide to the product fields AI can read shows where to look.
Then check your access settings. OpenAI says sites opted out of OAI-SearchBot are not shown in ChatGPT search answers, and that results can take about a day to adjust after a robots.txt update. Perplexity says PerplexityBot surfaces and links websites in Perplexity search results, and that robots.txt changes may take up to a day to apply. Open your robots.txt and confirm neither OAI-SearchBot nor PerplexityBot is blocked. Allowing a crawler is not a promise that your store will be cited or shown. It only removes one reason it could be left out.
Build a baseline you can repeat
A baseline is a saved record of the same questions, asked the same way, so a later run can be compared with it. You do not need special tools to make one.
Here is a worked example of the structure, using the sample run Arbling publishes. This is Arbling's sample data, not a real store's result. Arbling's page shows a sample follow-up run that asks the same 12 buyer questions in ChatGPT, Gemini and Perplexity on 14 November. The pieces are what matter: a fixed list of buyer questions, named assistants, a stated date, and a repeat of the same list later.
To track what ChatGPT and Perplexity say about your store next month, keep these details with this month’s answers:
- A fixed list of buyer questions, written once and never reworded.
- ChatGPT in an Unpersonalized temporary chat and Perplexity in incognito mode.
- The date and the place you tested from, saved next to each answer.
- The cited sources for every answer, saved as links.
- The same list asked again on a later date.
Compare the two runs question by question. Do not expect the answers to match, and do not read a change as proof of its cause. The record shows what the assistants displayed on each date, nothing more. Keep the list steady and the comparison stays fair.
Where Arbling fits. Arbling collects AI answers for a declared set of buyer questions, assistants, markets and dates, turns a catalog into buyer questions and records which products, variants and sellers assistants show for each, and fixes the product facts behind each gap in the catalog, with the merchant previewing each change and publishing it on confirmation.
Start with one buyer question you hear regularly. Save both answers and their links today. The next time someone says your store disappeared from an answer, you’ll have a dated example to compare.
All sources were checked on one date, 21 November 2026.