What ChatGPT shopping research actually does
When a buyer types 'best waterproof work boots for site work under £120', ChatGPT treats that as a research task, not a lookup. It breaks the request into web searches, reads the results, and assembles a comparison from what it finds. You'll often see it return two or three named options with a line on price, fit, and why each one made the cut.
The key point for a merchant: ChatGPT is reading real pages to build that answer. Retailer product pages, category pages, review sites, and buyer's guides. If your page is one of the pages it reads and trusts, your product can be named. If it can't find or make sense of your page, it names someone else's.
It runs searches, then reads trusted pages
Behind the scenes, ChatGPT rewrites the buyer's request into its own search queries, runs them, and works from the results. The words it searches for are usually not the exact words the buyer typed, and they're usually not stock codes. They read like how a person describes a product: material, size, use case, price band.
It also leans on sources it treats as reliable. A product page that clearly states what the item is, well-known review platforms, and established retailers all carry more weight than a thin page that hides the detail behind an image or a stock reference. So two things decide whether you're in the answer: can ChatGPT match its search to your page, and does it trust what it finds there.
Clean product data is the price of entry
Most stores lose here before anything else. If the buyer-facing detail lives in an image, a spec PDF, or a stock description like 'BT WP S3 43 BLK', a machine reading the page can't extract it. It can't tell that's a size 43 black waterproof S3 safety boot, so it can't put you in a comparison about waterproof safety boots.
Write the important facts as plain text on the page: what the product is, the material, the size or dimensions, the key use, the price, and whether it's in stock. On Shopify that means real body copy and filled-in product fields, not just a gallery. On ERP-driven merchant catalogues it means overriding the visible title and lead copy so they read as words, not warehouse codes, with the stock reference tucked in the body where shoppers don't see it.
Match the language of the searches it runs
ChatGPT shopping research rewards pages that mirror the query. If it's running 'waterproof safety boots under £120 uk', the store most likely to be named is the one whose page title, heading, and first line say waterproof safety boots and the price, not the one titled by SKU.
You can see the real searches ChatGPT runs, then close the gap. Ask ChatGPT a buying question in your category, let it search, and pull the queries out of your browser's developer tools. Then line up the words a machine weighs most, the page title, the H1, and the opening sentence, against that query. That isn't keyword-stuffing. It's closing the distance between what ChatGPT searched for and what your page plainly says it is.
Review signals and structured data tip the pick
When several products fit, ChatGPT tends to reach for the ones with visible proof. Ratings and reviews, on your own product page and on third-party platforms, give it something concrete to cite: 'rated 4.6 from 200-plus reviews'. A product with no review signal anywhere is easy to leave off a shortlist. Collect reviews and show them as readable text, not just a star image.
Structured data helps the machine read your page cleanly. Product schema that states the name, price, availability, and aggregate rating in a format built for machines makes it easier for ChatGPT, and Google's AI results, to lift the right details. It doesn't guarantee a mention, but it removes ambiguity, and ambiguity is what gets you skipped.
What to do this week
Pick three products you'd want named in a buying question. For each, ask ChatGPT the question a real buyer would ask to end up buying it, and note which stores it named and why. That tells you the standard you're up against.
Then fix your own page to that standard: plain-text specs and price, a title and first line that speak the buyer's language, at least a few genuine reviews on show, and Product schema in place. Re-test in a few weeks by asking the same question. The aim is simple. When ChatGPT does the buyer's shopping research, your product is one it can find, read, trust, and name.