How to Rank in ChatGPT Shopping Is Really About Product Data, Not Persuasion

The mistake most store owners make: trying to persuade the model

If you want to show up in ChatGPT Shopping, stop writing copy that sounds persuasive and start giving the system facts it can read. A human shopper may be nudged by a clever line, a brand story, or a polished promise. A shopping system reads the data instead.

That distinction is the whole game. It wants clean product data it can parse, match, and trust. If the record is messy, vague, or incomplete, the item becomes harder to classify and easier to skip. Machines are consistent that way, which is more than can be said for most product catalogues.

This is why title polish, storytelling, and keyword stuffing matter less than structured facts like product type, variant name, material, size, colour, compatibility, and availability. Google has said structured data helps search engines understand page content more accurately, and Schema.org exists for the same reason. Machines need explicit product facts.

A page can sound beautiful and still be hard for a shopping system to use. A product record can sound plain and still be easy to surface, because the facts are there and lined up clearly.

Think about the difference between a page written for people and a record written for systems. The page can explain benefits, answer objections, and sell the product. The record has to state what the item is. Both matter, but product data determines whether the item is eligible to surface well.

If one listing says lightweight travel bottle, 500 ml, stainless steel, blue, and another says premium everyday essential with a lifestyle paragraph, the first one is easier to classify. The system can tell what it is, who it fits, and whether it matches a query. The second one is a mood board with a price tag.

You are not trying to write better sales copy for a model. You are making the product data machine-readable and consistent everywhere it appears. When the title, attributes, and availability line up across the page, feed, and markup, the system gets a clean signal. When they do not, it gets noise, and noise loses.

What ChatGPT Shopping is actually reading

Shopping systems read product titles, descriptions, attributes, category signals, prices, stock status, images, and merchant-provided structured data. That is the input set. They do not read your brand story the way a person does, pausing for emotional beats or appreciating a clever turn of phrase. They match product facts to a shopper query and decide whether the item fits.

When a shopper asks what size to order, they want a direct answer. Shopping queries work the same way: the system wants direct product facts rather than a paragraph that reads like a marketing slogan.

This is where many stores create their own confusion. The product page says one thing, the feed says another, and the structured markup says a third. A page may call an item a black running shoe, while the feed says trainer, and the markup leaves out colour entirely. That inconsistency makes the item harder to trust.

Systems do not guess well when the same product has three identities. They prefer records that agree on the basics, because agreement signals confidence. In catalogue terms, consistency is not a nice-to-have; it is the difference between being understood and being passed over.

Schema.org Product markup shows how much shopping visibility depends on explicit data. It includes fields like name, description, brand, SKU, offers, availability, aggregateRating, and image. That list tells you exactly what matters: the system needs to know what the product is, who makes it, how it is identified, whether it is in stock, and what it looks like.

A vague title like premium everyday essential fails because it hides the product type. An exact title like stainless steel water bottle, 500 ml, blue gives the system something it can match to a query without guessing.

People search in direct language all the time. They ask whether a jacket runs small or how much a product weighs, because they want the fastest path to the answer. Shopping systems are looking for the same thing: exact product facts that line up with the query. If your title, attributes, and markup are clear, the system can do its job. If they are muddy, the item gets passed over. The machine is not being difficult; it is being literal.

Fix the product title first, because it carries the most weight

Start with the title and use this formula: product type first, then the most important differentiator, then the variant details that matter for matching. That means something like running shorts, men’s, 7 inch, black, or ceramic mug, 12 oz, white, rather than a brand-heavy slogan or a sentence that buries the item.

The first words matter most, because they tell the system which bucket the product belongs in. If the bucket is wrong or vague, everything that follows has to work harder.

Titles should be specific without sounding stuffed, and that is the line to walk. A weak title says premium everyday essential, best-in-class solution, or something equally empty. Another bad pattern is internal jargon that means nothing outside the company. A third is hiding the product type at the end, as in ComfortFit Pro Series, black, size M, men’s t-shirt. By the time the system reaches t-shirt, the useful signal has already been diluted.

Put the product type up front. State the thing first, then the detail that separates one item from the next.

Variant-heavy catalogues need extra care. If you sell size, colour, pack count, material, or compatibility variants, each title has to stay distinct and easy to match. Blue cotton crew socks, 3 pack is a different record from black cotton crew socks, 6 pack. An iPhone case compatible with iPhone 15, clear is not the same as one compatible with iPhone 14, clear. If the title does not carry the difference, the system has to infer it from weaker fields, and that is where mistakes start.

Lean teams make the same errors over and over. They copy supplier titles unchanged, they use one naming style on collection pages and another in the product feed, and they change titles for SEO without checking feed consistency, which creates split signals.

Google Shopping guidance has long treated product titles as one of the most important signals for matching items to queries, and title quality affects relevance directly. If you want the item to match, make the title say exactly what the item is, every time, in every place it appears. Consistency is what wins here.

Fill in the attributes shopping systems need to classify the item

If the earlier sections were about cleaning up the title and the product family, this is where the catalogue starts doing real work. Shopping systems classify products from attributes, and the ones that matter most are the attributes a shopper would use to separate one item from another: brand, product type, material, size, colour, gender or age group where relevant, compatibility, condition, and variant identifiers.

Google Merchant Centre documentation keeps returning to required and recommended attributes for a reason: missing or inconsistent data reduces item quality and makes a product harder to match. If the system cannot tell whether a shirt is men’s or women’s, or whether a charger fits a specific device, it does not guess well. It usually loses confidence and moves on.

The mistake most stores make is treating attributes like a dumping ground for every detail in the warehouse. The job is to add the fields shoppers actually use to choose. Capacity matters for drinkware, wattage for electronics, fit for apparel, and dimensions for furniture, while for beauty and home goods the deciding factor can be ingredient details, finish, scent, or material.

A candle without scent notes, a lamp without dimensions, or a jacket without fit information leaves both the shopper and the system guessing, and guessing lowers the chance of a strong match. The details are not decoration. They are the decision points.

Standardise the values across the catalogue. One page saying navy and another saying dark blue creates avoidable inconsistency, and inconsistency is where product data starts to drift. The same rule applies to size labels, material names, and compatibility terms. If one product says cotton blend and another says cotton/poly, the catalogue looks messy even when the items are similar.

Keep the vocabulary tight and use it consistently throughout. That helps systems group products correctly and helps shoppers compare without second-guessing whether they are looking at the same item under a different label.

The point is not to add every possible field. The point is to add the fields that actually define the purchase. A shopper comparing a bottle, a chair, or a pair of shoes needs the facts that separate one option from another. If you leave those facts out, the system fills the gap with assumptions, and assumptions hurt visibility and conversion. Clear attributes make the item easier to classify, easier to match, and easier to trust.

Write descriptions for clarity, not persuasion

Product descriptions should explain what the product is, who it is for, what it includes, and what makes it different, in plain language. Long blocks of brand copy, lifestyle language, and vague benefit claims do very little when the factual details are missing. A shopping system cannot use a paragraph about “elevating everyday rituals” if it still does not know the size, material, or compatibility.

The same goes for a shopper comparing items. Nielsen Norman Group research has long shown that people scan for concrete details and specifications when a purchase carries risk or fit concerns, which covers most ecommerce categories. People want the facts before they want the atmosphere.

Use a simple structure: one sentence on the product, one on the main use case, one on key specs, one on what is included, and one on care or compatibility. That gives the system and the shopper the facts in a predictable order.

For example, a description should say what the item is, who should buy it, what size or material matters, what comes in the box, and how it should be used or maintained. That format works far better than a polished paragraph that sounds nice and says almost nothing.

Keep the description consistent with the title and attributes. If the title says stainless steel, the description should not casually refer to aluminium. If the size field says 12 oz, the description should not mention 16 oz. Contradictions create trust problems for both systems and shoppers, and sloppy listings lose visibility. Persuasion still matters for human readers, but it belongs after the facts are clear. The facts are what get the product surfaced; once it is in front of the right shopper, the copy can do its job.

Think of the description as the answer sheet rather than the sales pitch. A shopper who wants to know whether a jacket runs small wants that answer rather than a brand story. Put the plain facts first, then let the shopper decide. Clear descriptions help the system classify the item and help the shopper compare it against the next option in the list. That is what good product copy does: it removes friction.

Images, variants, and availability can make or break eligibility

A strong title and a clean description cannot rescue a product if the images are weak, the variants are messy, or the stock status is wrong. These fields carry a lot of weight, because they tell the system whether the item is real, current, and easy to buy. Google has documented that product-data quality problems, including mismatched price, missing availability, and poor image quality, can affect item approval and performance in shopping surfaces.

That is not a small issue. If the system does not trust the offer, it stops showing it.

Start with image quality, using clear product-on-white images where that format makes sense, especially for items where the shape, colour, or finish matters. Keep aspect ratios consistent across the catalogue so the grid stays tidy.

Show the actual product, not only lifestyle scenes, because lifestyle images hide the details shoppers need. A sofa in a styled room is useful, but the shopper still needs to see the arm shape, the fabric texture, and the exact colour. A product image should answer the practical question someone asks before they buy: what exactly am I looking at?

Variant hygiene matters just as much. Each variant should map cleanly to a real option, and the shopper should not have to guess which colour or size they are buying. If a listing has six colours, six sizes, and a few style differences, every option needs to be separated and labelled cleanly. Bad variant data creates ghost options, duplicate options, and mismatched thumbnails, which confuses the system and annoys shoppers. The same problem shows up with bundles, multipacks, subscriptions, preorder items, and products with multiple compatible parts. Those edge cases are where data errors pile up, because the catalogue rules stop being obvious and the cleanup gets deferred.

Availability and price consistency are the final gatekeepers. If the page says in stock but the feed says out of stock, or the price changes in one place and not another, the offer looks unreliable. Stale pricing and mismatched stock status suppress visibility, because they create a bad shopping signal. Many stores lose easy wins here: the product may be good and the copy may be clean, but the offer data is messy, and that is enough to keep it from surfacing when a shopper is ready to buy.

How to audit your catalogue without wasting a week

Start with the products that matter most, ahead of the ones that are easiest to edit.

1. Audit your top-selling items first, then your highest-margin items, then the products with the most impressions and the weakest click-through rate. That order catches the catalogue problems that cost real money.

Search Console and merchant reporting often show pages earning impressions without clicks when titles and snippets miss shopper intent. That pattern usually points to weak product data rather than a traffic problem. If a product is getting seen and ignored, the data is doing a bad job of answering the shopper’s question.

Once you have that list, look for patterns rather than one-off mistakes: repeated vague titles like “Classic Shirt” across multiple products; missing attributes such as material, size, compatibility, or fit; variant names that change from “Blue” to “Navy” to “Midnight” for the same shade. Duplicate products split reviews and attention, and descriptions that say the same thing in different words usually mean nobody owns the catalogue.

The product page, the feed, and any structured data should all state the same facts in the same way. If the page says cotton, the feed says cotton, and the structured data says polyester, the catalogue is broken.

Do not start by fixing everything. Fix the items that are both important and messy first. A low-value item with a perfect title is a vanity project; a top seller with a vague title, missing attributes, and a mismatched feed is a revenue leak.

Build a simple checklist that a small team can repeat every month or quarter: title, primary attributes, variant naming, duplicate check, description match, feed match, structured-data match. That keeps product-data quality from depending on memory, which is how catalogues drift in the first place.

What to do when your catalogue is messy, large, or inherited from an agency

A messy catalogue is normal. Many stores inherit supplier imports, old SEO edits, inconsistent naming, and agency changes made without a catalogue strategy. The first move is batching rather than chasing perfection. Clean by collection, by product type, or by revenue tier.

Pick one lane and finish it before moving on. If you try to fix the whole store at once, you will spend a week polishing low-value items while the products that actually sell stay messy. That is how teams end up with a catalogue that looks busy and still fails to answer shopper questions. Activity is not the same as progress.

Use clear decision points for the messy cases.

• Duplicate products should be merged when they compete for the same search intent and confuse the shopper.

• Discontinued items should be retired or redirected rather than left to collect dead traffic.

• Variant splits belong together when the shopper is choosing size, colour, or pack count, but separate products make sense when the difference changes the use case.

If a shopper would need to ask a question to understand the product, the data is incomplete. That rule is simple, and it works. It catches the same kind of friction Baymard Institute keeps finding: unclear product information and weak product-page structure make selection harder, especially when products are similar.

Treat product data as operations rather than copywriting. Copywriting can make a product sound better; operations make the catalogue usable. That difference matters when you inherit a store full of old edits and supplier junk.

A team that owns the data can clean the catalogue in batches, keep naming consistent, and stop new mistakes from entering the feed. A team that treats it like ad copy will keep rewriting descriptions while the same bad structure stays in place. If you want a catalogue that can rank in shopping systems, fix the structure first, then the words. The words matter, but they stand on the structure underneath them.

Frequently asked questions

Written by Richard Newton, Co-founder & CMO, Sprite AI.