ChatGPT is showing the wrong price for your product

You ask ChatGPT about one of your own products and there it is: the photo is right, the name is close, and the price is three months old. Or worse — it says "out of stock" when you have stock, or it sends the shopper to a reseller charging 20 % more. Your first instinct is to go find where you edit that. You cannot: that data never came from your site. What follows are the four different causes that hide behind the same symptom, each with the specific fix that works for it, and the one check that tells you which of the four you are dealing with.

Worth saying up front to save you an afternoon: this is not the same problem as ChatGPT getting facts about your business wrong. Those are entity facts — who you are, where you are, what you do — and you fix them by fixing the sources the model reads about your brand. This is catalog data — price, stock, variant, listing — and it travels through completely different plumbing.

Where that price actually comes from

OpenAI's documentation is explicit, and worth reading literally because it dismantles the most expensive assumption in the room: product listings "may include prices, which we receive from third-party providers". Not from your site. From third-party data providers that aggregate merchant catalogs.

And there is a second detail that explains most of the panic. OpenAI notes that prices in the initial response "typically reflect the price from the first merchant listed, which may not be the lowest available price". When several merchants sell the same product, ChatGPT builds a merchant list; the number the user sees first belongs to whoever landed on top. If that is a reseller carrying your product at a different price, that is the price your prospective customer just read as if it were yours.

There are exactly two documented routes for your data to be the source rather than a middleman. If your store runs on Shopify, you are already in: OpenAI says product data "is already integrated into ChatGPT through Shopify Catalog" and that "no additional work is required from individual merchants". Otherwise the route is the direct product feed, which exists precisely, in the documentation's own words, to "ensure ChatGPT always reflects the most up-to-date information about its products".

Cause 1: the lag (and it is official)

Before you touch anything, rule out the most common and least actionable cause. OpenAI admits it in writing: "when merchants update their pricing or shipping terms, there may be some delay before it is reflected in the information you see", adding that it is actively working on faster methods to update it.

No published window. The operational reading is simple: if your product changed price this week, stale data in ChatGPT is probably not a misconfiguration but the lag. And it carries a concrete business consequence: short promotions — a weekend, a 48-hour flash sale — almost certainly will not be reflected while they run. Planning campaigns on the opposite assumption is planning a disappointment.

Cause 2: the wrong source wins

This is the one that costs the most money and gets mistaken for a bug most often. Your product shows up, but with another merchant's price and link. That is not a model error: it is the result of a ranking. OpenAI documents that the merchant list "is generated by ChatGPT based on merchant and product metadata from third-party providers or directly from the merchants themselves", and that merchants are ranked on factors like availability, price, quality, and whether they are the maker or primary seller of that item.

That last factor is your strongest card and almost nobody plays it. Being the maker or primary seller of your own product is something no reseller can claim and you always can. We treat it separately, because it deserves its own conversation: why marketplaces show up ahead of your store in AI answers.

For the problem in front of you right now, what matters is this: when a reseller takes the top spot, it is almost always because their data is better than yours — clearer stock, more consistent pricing, a fuller listing — not because they are bigger. Bad data loses to good data even when the brand is yours.

Cause 3: the title and labels are not yours

This is where teams burn time trying to correct something that is not correctable. OpenAI documents that ChatGPT "may generate simplified product titles and descriptions based on information it receives from third-party providers", and explains why: merchants use varying titles for the same product, so the model normalizes them for readability.

Same with the labels over the images — "Budget-friendly", "Most popular". They are model-generated, and OpenAI states they are "not guarantees or verified statements, and may not reflect all available market data". Its own example is refreshingly honest: a "Budget-friendly" label might mean reviewers frequently mention good value, not that it is the lowest price available. And review summaries are built from reviews on public websites and, verbatim, are "not verified by OpenAI".

Practical translation: there is no field that forces your brand copy through. What you do control is the raw material. A clean, consistent title with no promotional noise across all your sources gives the model less to normalize and less to invent.

Cause 4: your own data arrives incomplete

If you send a direct feed, the problem may be yours. The product feed spec defines 79 fields, 19 of them required for a non-Ads feed. The ones that matter here:

• price — required, number plus currency (79.99 USD).
• availability — required, with values like in_stock. This is the field behind the phantom "discontinued".
• sale_price with sale_price_start_date and sale_price_end_date — optional, but the correct way to declare a dated promotion instead of overwriting the base price.
• is_eligible_search — required. Set to false, your product simply cannot surface in ChatGPT results.
• expiration_date — optional, removes the product after that date.

A feed that validates but ships a stale availability is worse than no feed at all: it hands the model an authoritative source that happens to be wrong.

The button that does exist

There is an official correction path and it is not widely known. OpenAI documents that you can report "any products that have incorrect information or violate our policies by using the menu in the product", alongside the thumbs-up/thumbs-down icons and a feedback form.

It is not an admin panel and it promises no turnaround, so treat it for what it is: a signal, not a switch. But it is the only route OpenAI publishes for flagging incorrect product data, and it costs two minutes. Do it, and fix the source in parallel — because if the source stays wrong, the data comes back.

Where to start tomorrow

Take your five highest-margin products, not your five best sellers. Ask ChatGPT about each one the way a customer would — "where can I buy X?" — and note three things: what price it shows, which merchant it links to first, and whether the stock status is right. That fifteen-minute grid tells you which of the four causes you are in and, more importantly, whether your problem is data or ranking.

If the price is right but the link goes elsewhere, this is not your article: the one next door is. If the price is wrong everywhere, start with your sources. And if you run a direct feed and it still breaks, check availability before anything else — it is the field that quietly breaks a listing most often.

Frequently asked questions