Beauty stores: can an AI agent read ingredients, size and reviews, and tell your product apart?
Shoppers ask agents for products by skin type, ingredient, volume and rating. Agents need the exact product (brand, line, volume), what is in it, and reviews they can read.
Requests a shopper might give an agent
- “A fragrance-free moisturiser for sensitive skin, 50 ml, rated 4.5 or higher.”
- “Vitamin C serum under ₺800 without alcohol in the ingredients.”
- “The 100 ml version of this shampoo, delivered tomorrow.”
Examples written by us to show the kind of constraints shoppers set. They are not measured question frequencies.
The questions an agent needs your product page to answer
The free test reads one of your product pages and checks whether it answers the questions expected for that type of product: first in the structured data, then in the visible page text (a rule-based dictionary in Turkish and English, no AI). The result appears as the “Product question coverage” line, which doesn’t affect the score.
| Question | Why it matters here | Expected for |
|---|---|---|
| size / dimensions | Volume or weight (ml, g) separates a travel size from the full size; the agent quotes it in comparisons. | beauty and personal care |
| material / ingredients | For cosmetics this is the ingredient list; the test looks for “ingredients:” in the data or text. | beauty and personal care |
| returns | Many stores restrict returns of opened products; the agent needs that rule in words, not only in the sales contract. | beauty and personal care |
| delivery time | “Delivered tomorrow” requests are only answerable with a stated delivery time. | beauty and personal care |
| stock / variants | Shades and sizes are variants; the agent must know which one is in stock. | beauty and personal care |
Where the test looks for each answer
- size / dimensions: additionalProperty named size/dimension, or size, width, height, depth; in text: a size chart, “dimensions:”, measurements like 40 × 60 cm or 50 ml.
- material / ingredients: material in the product data or additionalProperty named material/fabric/ingredients; in text: “material:”, “ingredients:”, fibre percentages.
- returns: hasMerchantReturnPolicy in the product or offer data; in text: “14-day free returns”, return period.
- delivery time: shippingDetails with handling and transit time; in text: estimated delivery, ships within N days, same-day dispatch.
- stock / variants: availability in the offer, or variants (hasVariant); in text: in stock, out of stock, only N left.
The checks that matter most in this category
All 27 checks run on every store and the weights are the same for everyone; these are the ones most often decisive in this category, and why. Each name opens the check’s own page.
| Weight | |
|---|---|
| Product structured data (schema.org) critical blocker Product structured data is how agents identify the exact product and its offer. | 7 |
| Product identity (GTIN/MPN, brand, SKU) GTIN, brand and SKU let an agent match the same product across stores and the brand’s own catalog. | 5 |
| Reviews and rating Agents often shortlist by rating threshold and review count; ratings must be in the structured data. | 3 |
| Price, currency, availability and URL Price, currency and availability for the exact size the shopper asked about. | 6 |
| Price consistency critical blocker The price in the data must match the visible price; a mismatch is a critical blocker. | 6 |
| Returns page and window Hygiene-related return exceptions should be stated plainly on the returns page. | 4 |
Run the free test on your store → How the free test works →
Simulator scenarios that fit
The agent simulator sends AI agents through your store in a real browser. They stop at checkout: they never order, sign in or type personal data.
Free simulator
- Conditional search: Budget plus a condition (volume, shade) in one request.
- Policy reading (return period, shipping cost): Reads the return period and shipping cost before deciding.
Paid project panel (one extra scenario each month, in rotation)
- Sorting (most reviewed, best rated): Sorts a category by most reviewed or best rated and picks a top result.
- Coupon field: Checks whether there is a coupon field on the way to checkout, without typing a code.
- Honesty when out of stock: A sold-out shade should be reported, not swapped for another.
Send the agents to your store → · Pricing →
What we don’t claim
- The test checks whether an ingredient list exists, not whether it is complete, correct or compliant with cosmetics regulations.
- Skin-type suitability and claims such as “dermatologically tested” are not verified.
- We don’t publish averages for this category: no figure without our own sample size and date.
- The question check is a dictionary signal on one product page, not a verdict: a phrase it finds may not fully answer the question, which is why the result shows the excerpt.
- Nobody can guarantee that an AI assistant recommends your products; we measure whether agents can read and use your store.
Questions
Do you check our ingredient lists against regulations?
No. We only check whether an ingredient list is present in a form agents can read. Regulatory compliance is outside what we measure.
Why does product identity matter so much here?
Beauty products come in many sizes and versions. A GTIN, brand and SKU let an agent be sure it is comparing the same product across stores.