A shopper's question is a list of requirements, and an assistant can only name you if your page contains facts that match them. Most product pages are written to persuade someone who has already arrived, which is a different job, and it is why pages that read beautifully lose to pages that read like a spec sheet.

This is the third gate in getting your shop recommended by AI assistants, and the sibling of the piece on structured data. That one is about labelling your facts so a machine knows which is the price. This one is about whether the facts are there at all. You can pass that gate and fail this one, and plenty of shops do.

A question is a list of requirements

Take a question someone would really type:

"best waterproof hiking boots for wide feet under €150"

Pull it apart and it is four separate requirements:

The words	What has to be on your page
hiking boots	the category, stated plainly
waterproof	the rating or membrane, not just the adjective
for wide feet	the fitting, in the words a buyer would use
under €150	the price, current and machine-readable

An assistant answering this needs to satisfy all four before it commits. Miss one and it cannot be confident, and the safe move is to name a shop where it can. You do not get partial credit for three out of four, because the shopper did not ask for three out of four.

Now read your own best-selling product's page against the questions your buyers actually ask. Most owners find the category and the price, and then start guessing.

The details that keep coming up

These are the ones that appear in real shopping questions across most categories. Not all apply to what you sell. The ones that do are the ones worth an hour.

Fit and size. Not just which sizes exist, but how it runs. Wide, narrow, true to size, sizes up. This is the single most common thing a page omits and a shopper asks.

Measurements and weight. Actual numbers with units. "Lightweight" is not a number and cannot be matched against "under 500g".

Materials and construction. What it is made of, and what the important part is made of. The upper, the sole, the lining, the frame.

The performance figure that defines the category. Every category has one or two. A waterproof rating. A capacity in litres. A battery life in hours. A thread count. If you sell in a category, you know which one it is, and the page should say it in numbers.

Compatibility. What it fits, works with, or needs. This is most of the question in anything technical.

What is in the box. Whether the stand, the cable, the case is included is a question people ask constantly and pages answer rarely.

Who it is for. The use, the level, the conditions. A beginner's version and a professional's version of the same object are different answers to different questions.

Care and durability. How it is cleaned, whether it can be repaired, what the warranty is.

Price, current, and stock status. Both need to be right and both need to be readable. The labelling side is here.

What a vague description costs you

Here is a description that is not badly written:

Our Classic Trail boot combines timeless design with modern comfort. Built to last and crafted with care, it is the perfect companion for your next adventure. Available in a range of sizes.

Every sentence is fine. Together they contain no fact. There is nothing here for a question to match against: no material, no waterproof rating, no fitting, no weight, no use it is suited to and none it is not.

The same product, written to be answerable:

A waterproof leather hiking boot for day walks and light multi-day routes. Full-grain leather upper with a Gore-Tex lining, rated for sustained rain. Vibram outsole. 480g per boot in a UK 8. Available in a wide fitting as well as standard, and buyers with wide feet report it fits true to size. Resoleable.

The second is shorter than most product pages manage and answers nine of the things a shopper might ask. The point is not to write more. It is to write fewer sentences that contain more facts.

The thing almost nobody does: say who it is not for

A page that says "not warm enough for winter hiking, and too stiff for casual wear" has not talked itself out of sales. It has made itself safe to recommend for everything else. An assistant that can see the edges of what a product suits can commit to it inside those edges with much more confidence than one guessing from marketing copy.

It has a second effect worth having: the people it turns away were going to return it. You are trading a recommendation you would have lost anyway for confidence in the ones you want.

If your page needs JavaScript to show a price, they see nothing

This is the failure that catches good shops with good product information, because nothing about it is visible to a person.

Your page is built in two steps. The server sends some HTML, and then code runs in the visitor's browser and fills in the rest. If the price, the stock status or the description arrive in that second step, then anything that reads the first step and stops sees an empty shell.

Google is not the problem here: Googlebot runs that code. The AI crawlers largely do not. Vercel, which runs the infrastructure a great many sites sit on, monitored crawler traffic across its network over several months and found that none of the major AI crawlers render JavaScript, naming OAI-SearchBot, ChatGPT-User, GPTBot and ClaudeBot. They fetch script files and do not execute them.

This is measurement by an infrastructure company rather than a statement from OpenAI or Anthropic, so treat it as strong evidence rather than a vendor guarantee. It also matches what Google recommends for its own shopping results, which is to put Product structured data in the initial HTML.

How to check yours in one minute. Open a product page, press Ctrl+U on Windows or Cmd+Option+U on a Mac to view the source, then search that raw text for your price. If the number is there, you are fine. If it is not, your product page is invisible to exactly the crawlers this whole subject is about, and it is a question for whoever maintains your shop rather than something to fix in the page editor.

Reviews, returns and delivery are evidence

An assistant is being asked to vouch for a shop it has no relationship with. Anything on the page that is not your own marketing carries weight out of proportion to its length.

Reviews are the obvious one, and the specific ones do the work: a review saying the boots stayed dry over three days answers a question the description also should. Returns and delivery terms matter for a different reason, which is that Google lists shippingDetails and hasMerchantReturnPolicy among its recommended properties for merchant listings, so they are facts the machinery already has a place to put.

Fifteen minutes, one page

Pick the product that matters most to your business rather than the first one in your catalogue.

Write down the three questions a real buyer would ask about it. Use their words, not your category names.
Read the page and mark which requirements in those questions it actually answers. Be strict. An adjective is not an answer.
View the page source and search for the price, to rule out the JavaScript problem above.
Add the missing facts as facts. Numbers, units, materials, fittings.
Add one sentence saying what the product is not suited to.

Then leave it alone for a few weeks. The crawler has to come back before anything can change, and changing it again in the meantime only makes it harder to tell what worked.

Questions people ask

Does length matter?

Not in itself, and chasing a word count is the wrong instinct. A dense two hundred words beats an evocative eight hundred. What matters is the number of answerable facts, not the number of words around them.

Will adding all this hurt how the page reads?

It can, if it goes in as a wall of prose. Most of it belongs in a specification table or a short FAQ block under the description, which is easier for a person to skim as well. The persuasion and the specification can both exist. They just should not be the same paragraph.

Should I use AI to write these descriptions?

That is your call and plenty of shops do. The thing to watch is that a model writing from your existing thin description will produce more words containing the same absence of facts, and often invent the ones it lacks. The facts have to come from you or from your supplier either way.

What about the manufacturer's copy?

If you sell the same product as fifty other shops using the same supplied description, you have nothing to be chosen on. This is where a small shop can genuinely beat a large one, because you can add what you know from actually selling the thing and they will not bother.

My category is not technical. Does this still apply?

Yes, with different facts. For clothing it is fit, fabric, care and measurements. For food it is ingredients, allergens, origin and quantity. Every category has the handful of things buyers always ask, and you already know yours from your customer emails.

Where we come into it

One product, read honestly against three real questions, is an afternoon you can spend today and it will tell you something. The pattern only shows up across the catalogue: which gaps repeat, which of them appear in the products that actually earn you money, and which ones a competitor filled in where you did not.

Shop on Radar does that reading for you, product by product, against the questions your buyers actually ask, and then puts those questions to ChatGPT to see whose page answered them. You get the list of facts yours is missing. We do not write your descriptions, and we do not touch your shop.

We are not open yet. Join the waitlist and we will tell you when we are.

The product page details AI needs and yours probably lack