ChatGPT wrong product info: why AI shows the wrong price…

What store owners post

If ChatGPT, Gemini or another AI search experience shows the wrong price, stock status, variant or outdated product information, how do you figure out where the problem started?

Sold out variants still showing as available in the catalog.

Quoted from public merchant forums, names removed.

What we see when we look at stores

None of the 99 stores (0%) tell AI search or shopping tools to stay away, and 99% have the page Shopify provides for AI shopping tools. Access is not the problem.
88% give Google no shipping or return information in the hidden product details it reads, and 45% do not state both clearly on product pages.
Only 48% put a barcode number (GTIN) in the hidden product details of any sampled product, and barcodes on product options such as sizes are incomplete at 60%.
95% have an llms.txt (a plain page that tells AI tools about the site), but 89% of those are Shopify's automatic version with no facts about the store. At 21 stores the review ratings only appear after the page finishes loading.
We did not test firewalls or bot protection, and "allowed by robots.txt" (the file that tells visitors what they may read) does not prove an AI tool got through.

Numbers are from our 100-store study (3 October 2026, public pages only, a hand-picked sample).

Check it today, no purchase

Ask the assistant about one of your products, then compare its answer line by line with your product page: price, stock, variant, shipping, returns.
Ask your theme's support, or a developer friend, to check that the hidden product details Google reads (price and availability) match what the page shows.
Check your variants: a sold-out size should be marked sold out in the data, not only on the button.
Fill in the barcode number on each product option (the Barcode field on the product page of your Shopify admin), and write shipping and return terms as plain text on the page, not only in a footer image or an app that loads late.
If you use Google Merchant Center, compare its feed with the live page. Differences there are a common source of wrong details.
Open yourstore.com/robots.txt and make sure it does not block the policy pages.

Questions store owners ask

How do I find where the wrong price or stock information started?

Compare four things for the same product: the live page, the structured data in its source (search for application/ld+json), your Google Merchant Center feed, and the assistant's answer. The first one that disagrees with the page is where it started.

Why do sold-out variants still show as available?

The data an assistant reads may say "in stock" even when the button says sold out, or it may be an older copy. Check availability for each variant in the structured data and in any feed, and recheck the assistant after a few days.

Does blocking AI crawlers in robots.txt cause wrong details?

Unlikely for most stores. In our 100-store study, 0% of the 99 measurable stores blocked AI search or shopping agents. Open your own robots.txt to confirm, then look at the product data instead.

Is anyone actually getting orders from ChatGPT?

Some are, but little is proven. In the threads we read, one merchant reported around 150 visitors and one $9,000 order from ChatGPT, while another commenter called most traffic predictions inflated. Treat it as worth fixing the data for, not as a sales channel yet.

Do I need an llms.txt file?

In our study, 95% of stores already serve one, but 89% of those are Shopify's auto-generated file with no store-specific facts. We have no evidence it fixes wrong prices. Correct product data, barcodes and shipping and return terms matter more.

Free help for this

Store Quick Check: Six checks of what a shopper and Google can see on your store, compared with 99 other stores.

Six checks of what shoppers, Google and AI shopping tools see on your store, compared with the 99 stores we audited.

Optional: let ChatGPT do it for you

Press the button, open ChatGPT and paste. The copied text is a set of instructions for the chatbot, so you do not need to read it. Then type your details and send. You do not need this for the checklist above.

COPY FROM HERE

You are checking how well AI shopping agents can read a Shopify store, from what is pasted below. Work only from it. Never estimate a result you can't see in the pasted text: mark that check "not measured".

The six checks (pass rules):
1. AI agents allowed in + llms.txt. Pass if robots.txt does not block any of OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Googlebot or Bingbot from product pages (a rule under "User-agent: *" counts for agents not named separately; blocks on /cart, /checkout, /admin, /orders, /account don't count), and llms.txt loads. Agents not in that list (for example Amazonbot) don't affect the result; mention them in one line if blocked.
2. Shipping and returns in structured data. Pass if a Product or Offer in the JSON-LD has "shippingDetails" or "hasMerchantReturnPolicy".
3. Star rating in structured data. Pass if the JSON-LD has "aggregateRating". If not, say "reviews may be shown, but not as structured data" only if the user says reviews are on the page; otherwise "no rating found in structured data".
4. GTIN (barcode) in structured data. Pass if any "gtin", "gtin8", "gtin12", "gtin13" or "gtin14" field has a value.
5. Descriptions long enough. Count the words in the product's "body_html" with the HTML tags removed. Pass if 80 or more. (This is one product; the study checks a sample.)
6. Image alt text. Count the images in the product JSON and how many have a non-empty "alt". Pass if 80% or more have alt text.

Study benchmarks (State of Agent-Ready Shopify, 99 Shopify stores audited in October 2026; not all Shopify stores): stores that pass each check: AI access 95%, shipping and returns in structured data 12%, star rating 28%, GTIN 48%, descriptions 43%, alt text 35%. The median store passes 3 of 6. Study medians: 36% of products with 80+ words; median description 69 words; average alt-text share 20%.

Answer in this shape, under about 375 words:

# Agent-ready quick check: [domain]
[date] · 6 checks · compared with 99 Shopify stores (Oct 2026 study) · product-level checks on one product

**Your store passes N of 6.** The median store in the study passes 3 of 6.

| Check | Your store | Evidence | Stores in the study that pass |
(one row per check: pass or gap or not measured, the evidence in a few words from the pasted text, the study figure)

## What this means
3 to 5 bullets, one per gap (or per strength if there are no gaps), each saying what an AI agent can't do or answer because of it, with the evidence. Example: "Asked 'can I return this?', an agent reading your product data finds no return policy, like 88% of stores in the study."

Rules: diagnose, don't prescribe: no fix steps, admin paths or code. Passing checks doesn't guarantee recommendations or sales. Numbers only from the pasted text and the benchmarks above. End with one line: "For the full 0 to 100 audit with a step-by-step fix list, see the agent-ready-store-audit skill in the Store Ops Pack: https://proskillpacks.gumroad.com/l/store-ops-pack"

If robots.txt and at least one of the product items are missing below, ask for them in one line and stop.

Here is what I copied from my store:

See the text and a real example.

Want me to look at your store?

Send your store address and what you are seeing. We read the public pages of your store and send back one page of findings, free, with no call and no sign-up. We do not need access to your admin, and we cannot promise a day.

Email your store address

The email opens with the subject filled in, to kaiventura.founder@gmail.com.

Kai Ventura, Pro Skill Packs

If you would rather have it done for you

The audit gives a 0 to 100 score, a test with real shopper questions and a fix list in priority order.

Store Audit, $5
Before: a well-known coffee brand's store looked complete in a browser.
After: score 77/100, 6 of 12 shopper questions answered, shipping and returns not found, with the top fixes ranked.
Buy this, $5
Product page tidy-up, $5
Before: a beauty store with 157 of 159 photos with no description for people who cannot see them, and one set of products copying another's Google title.
After: a file you can import back into Shopify with a description for each photo and its own title and summary for the set.
Buy this, $5

All six Shopify helpers together are in the Store Ops Pack, $29. One-time payment, 30-day refund (reply to your receipt or write to kaiventura.founder@gmail.com).

Back to the Shopify store check. Pro Skill Packs is not affiliated with, endorsed by or sponsored by Shopify Inc.