In September, NIQ’s monthly shopper tracker crossed a line: 51% of US consumers said they had used at least one AI shopping tool in the past month. It is the first time the figure has passed half, after two quarters of steady growth. The most common use was product recommendations (20%), followed by AI personal shopping assistants (16%).
The more useful part for a small shop is how people use these tools. According to NIQ, shoppers are not handing the purchase over to a bot. They use ChatGPT, Gemini, Google AI Mode, Perplexity or Claude to compare options, judge value and narrow a long list down to two or three. NIQ calls it compressing the decision.
That changes what your product page is for. It is no longer only read by a person who already clicked through. It is also read by an assistant that is trying to answer “which of these is better for me?” If your page does not say the thing being compared, the assistant either guesses, or picks the rival whose page does.
What NIQ is saying, and how solid it is
The tracker is part of NIQ’s Quick Question research and surveys around 500 US consumers each month. That is a modest sample, so treat 51% as a direction, not a precise number; the trend over several months is more telling than one reading. NIQ’s own conclusion is that product content, AI discoverability and structured product data now matter as much as shelf placement and search ranking. NIQ sells measurement tools for exactly this (a new one with Similarweb is due in Q4 2026), so it has a reason to say so. Even discounted for that, the logic holds for a small online shop: an assistant can only compare what it can read.
The ten-line product-page checklist
Go through one of your best-selling product pages with this list. Each line is a fact an assistant is likely to need when it compares your product with another one. If the fact is only in a photo, a video or a review, count it as missing: write it in the page text.
- What it is, in one plain sentence. “Hand-thrown stoneware mug, 350 ml”, not just a brand name or a mood.
- Size and weight with units, including the packed size if it matters for shipping.
- Materials, and for anything worn, eaten or put on skin, the full composition or ingredient list.
- What is in the box. Cables, refills, spare parts, gift wrap: say it, and say what is not included.
- Who it is for and who it is not for. “Fits wrists 15–19 cm”, “not dishwasher safe”, “not suitable for children under 3”. Honest limits make the comparison fair and cut returns.
- Price, and what the price includes. Taxes, duties for international orders, any subscription.
- Shipping: where you ship, typical delivery time and cost, and the cutoff for same-day dispatch if you have one.
- Returns: how many days, who pays return postage, any exceptions. Put a short version on the product page, with a link to the full policy.
- Care and lifespan: how to clean it, how long it lasts, whether you sell spares or repairs.
- The one difference that matters versus the obvious alternative, stated as a fact you can back up: “made in small batches in Poland”, “uses standard AA batteries”, “refillable”. Not a claim like “best quality”.
Most shop platforms also let you fill in product attributes (size, colour, material, GTIN/barcode) as structured fields, separate from the description. Fill them in. They feed product feeds such as Google Merchant Center, and they are the “structured product data” NIQ is talking about.
A 20-minute test: ask the assistants to compare you
You do not need tools to see how an assistant reads your page. Pick one of your products and one direct rival that a customer might weigh it against. Then, in two or three assistants your customers are likely to use, ask:
- “Compare [your product] from [your shop] with [rival product]. Which is better for [a real use case]?”
- “What is the return policy for [your product] from [your shop]?”
- “How long does delivery take from [your shop] to [a city you ship to]?”
Write down three things for each answer: whether the facts about you are right, which pages it cites, and where it says it does not know. Each “I couldn’t find that” or wrong fact points at a line on the checklist above. Fix the page, wait a couple of weeks for the pages to be re-read, and ask again.
Two honest limits. Assistants do not all read your site live, and some lean on marketplaces, reviews or product feeds instead, so a perfect page will not fix every answer. And you cannot control what an assistant concludes. What you can control is whether the facts it needs are there and unambiguous.
Where to start this week
Do not rewrite the whole catalogue. Take your top five products by revenue, run each through the ten lines, and run the comparison test on two of them. That is an afternoon’s work, and it covers the pages most likely to come up when a shopper asks an assistant to help them choose.
