How to check if ChatGPT cites your store: a 6-step GEO audit you can run weekly

Most store owners ask the same question in the same way: "Is my shop actually showing up in ChatGPT?" It's the right question, but the usual way of answering it — typing your brand name into a chatbot and hoping you appear — tells you almost nothing. What matters is whether ChatGPT and Perplexity cite *your product pages* when a buyer asks a buying question you don't control.

This guide gives you a repeatable weekly check. It takes about fifteen minutes per product, and it produces something far more useful than a yes/no: a list of the specific fields and pages that are missing from your answer surface.

## Step 1: Build a question list, not a brand list

Chatbots are asked questions, not brand names. Write down five to ten questions a real buyer would type before they know your store exists:

• "best [product category] for [use case] under $[budget]"
• "does [product] work with [compatibility requirement]"
• "[product] vs [alternative] — which ships faster to [country]"
• "where can I buy [product] with a [return window] return policy"
• "is [product] [specification] or [other specification]"

Cover the full funnel: discovery, comparison, and last-mile trust questions. The last category — shipping, returns, warranty — is where most independent stores are invisible, because their pages describe products but never state the service facts assistants need to compare.

## Step 2: Run the questions in a fresh session

Open a new chat with no prior context. Ask each question exactly as written. Then read the answer for three things:

1. Are you cited at all? If yes, capture the exact URL the assistant links to — often it's a category page or a blog post, not the product page you wanted. 2. Are you mentioned without a link? A mention with no citation is a partial win. It means your text was readable but not specific enough to be worth linking. 3. Who is cited instead? Marketplaces, review sites, and forums are the usual winners. Note which competitor claims the assistant is repeating.

Log all three columns in a spreadsheet. Do this in a private/incognito window where possible; personalization and prior shopping history can change results between sessions.

## Step 3: Ask the assistant to explain its choice

This is the step almost everyone skips, and it's the most informative. Follow up with:

> "Why did you choose those sources, and what information would you need to consider [your domain]?"

Assistants are usually willing to describe the signals they prioritize: explicit specifications, stated prices, clear shipping and return terms, structured data, and unambiguous product naming. Their answer is a direct hint about which fields on your pages are doing the work and which are silent.

## Step 4: Inspect what the crawler actually sees

A beautiful product page can still be empty to an assistant. Check the rendered text:

• View the page with JavaScript disabled, or use a text-only renderer. If your specifications, price, and policies only appear after client-side hydration, some crawlers never see them.
• Confirm structured data is present and valid. Product, Offer, and FAQ markup give assistants a machine-readable version of facts you already display.
• Verify that each important fact exists as visible text, not only inside an image or a PDF. Citations originate from text, not from your hero banner.

If you want a faster read on this, [Visora's free audit](/audit) scans a product URL and shows which fields are parseable, which are missing, and how the page reads to an assistant.

## Step 5: Fix one gap, then re-test the same question

Resist the urge to rewrite everything. Pick the single highest-value gap from your log — usually missing return terms, an absent price range, or specifications trapped in an image — and fix only that. Wait a few days for re-crawling, then ask the identical question again in a fresh session.

This one-change/one-retest loop is what turns GEO from guesswork into measurement. It also tells you quickly whether a gap was a *content* problem (the fact was never on the page) or a *readability* problem (the fact was there but unparseable).

## Step 6: Repeat on a schedule

AI answers shift as assistants re-index, as competitors publish, and as your catalog changes. A citation you had in July can disappear in September. A monthly pass over your question list is enough for a small catalog; weekly is reasonable if you're actively optimizing.

The pattern to watch is not "am I cited today" but "which questions did I lose, and what changed on the page since I last won them."

## Frequently asked questions

### How do I know if ChatGPT cited my website?

Ask a buying question in a fresh session and look for your domain in the response or its source list. If your brand appears but the link points elsewhere, you were mentioned but not cited — a different problem with a different fix.

### Does appearing in ChatGPT cost money?

Organic citations are not purchased. You earn them by making product facts explicit, current, and machine-readable. Ads in AI surfaces are a separate, paid channel and should not be confused with citation visibility.

### How long does it take to see changes after fixing a page?

It depends on re-crawl timing, but plan on days to weeks rather than hours. Change one thing, wait, then re-ask the same question so you can attribute the difference honestly.

### What's the single highest-impact fix for most stores?

Stating service facts in plain text — price, shipping time, return window, warranty — because these are the comparisons assistants make most often, and they're the facts most product pages omit.

Run the check, log the gaps, fix one thing at a time. That's the whole discipline — and [our FAQ](/faq) covers the common questions that come up along the way.