The short answer
When a shopper asks an AI shopping assistant for a kettle that is quiet, holds enough for four mugs and fits under a low cupboard, the assistant can only answer from facts it can find. If your product page says "sleek, compact and whisper-quiet", there is nothing to check and nothing to quote. This guide shows how to write product descriptions that carry usable facts, without making them dull for people.
Why adjectives fail
Shopping assistants match a shopper's need to product evidence. Words like "compact", "powerful" or "long-lasting" are relative. Compact compared with what? Powerful enough for what? An assistant either ignores them or repeats them without being able to back them up. Neither helps the shopper, and neither helps you.
A fact has a value and, where it applies, a unit and a condition: "1.5 litre capacity", "24 cm tall with the lid closed", "up to 30 hours of playback with noise cancelling off". Facts can be compared, filtered and explained. That is what makes a product show up for the right question.
Five rules for usable descriptions
1. Facts over adjectives
Replace each adjective with the measurement behind it. If you cannot measure it, consider cutting it.
2. Units on every number
Write "2,400 W", "45 dB", "1.5 L", "UK size 9". A bare number is ambiguous, especially across markets.
3. Name the use cases
Say who and what the product is for: "suits a household of two to four", "designed for road running, not trail".
4. State compatibility explicitly
List the models, sizes or standards it works with, and the ones it does not.
5. Keep facts in text
Do not leave dimensions, size charts or spec tables only inside images.
Units and conditions
Units sound obvious, but they are where most copy goes wrong. "Holds 12" might mean place settings, cups or bottles. "Battery lasts all day" says nothing about the test conditions. Use the unit a shopper would use, keep it consistent across your range, and add the condition that changes the number. For a dishwasher that means place settings and the programme used for the energy rating. For headphones it means whether noise cancelling was on.
If you sell in more than one market, keep the market's convention. UK and EU shoe sizes differ, and so do plug types. An assistant serving a UK shopper should not have to guess.
Use cases and compatibility
Shoppers rarely ask for specifications directly. They ask for outcomes: a pram that folds small enough for a hatchback boot, trainers for flat feet, a replacement filter for a specific fridge. A short use-case line connects the product to those questions. Keep it honest: "suitable from birth to 22 kg" is useful; "perfect for every family" is not.
Compatibility deserves its own line or list. Name the exact models, part numbers or standards. If a charger works with USB-C Power Delivery devices up to 65 W, say so. If a vacuum bag fits only certain series, list them. Stating what it does not fit prevents returns and wrong recommendations.
What should you not bury in images?
Many product pages put the most useful information in a picture: a size chart, a dimension diagram, a table of what is in the box. People can read these. Text-based systems often cannot, or read them unreliably. Keep images for showing the product, and repeat the facts in text.
- Dimensions and weightHeight, width, depth and weight in text, with units and whether packaging is included.
- Size chartsPublish the chart as an HTML table, not only as an image.
- What is in the boxList every included item and anything sold separately.
- Spec tablesKeep the specification table as text, with the same attribute names across the range.
- Certifications and ratingsWrite out energy ratings, safety standards and age ranges rather than showing only a badge.
Before and after: an electric kettle
Here is a typical description, followed by a rewrite. The product is invented for illustration.
Before: Our stunning new kettle combines sleek design with powerful performance. Whisper-quiet and super fast, it's the perfect addition to any modern kitchen. Compact enough for any space!
After: 1.5 L cordless electric kettle, 3,000 W, UK plug. Boils one mug (250 ml) using the minimum fill line. 45 dB rated noise level. 24 cm tall with the lid closed, so it fits under most wall cupboards with 30 cm clearance. Removable limescale filter, 360° base, automatic shut-off and boil-dry protection. Suits a household of two to four. Matt black finish; also available in white (separate product code).
The rewrite is longer, but every sentence answers a question someone might ask. It can be matched to "quiet kettle under 25 cm", compared with another kettle on capacity and power, and explained back to the shopper with confidence.
| Shopper question | Before | After |
|---|---|---|
| Will it fit under my cupboard? | "Compact" | 24 cm tall, lid closed |
| Is it quiet? | "Whisper-quiet" | 45 dB rated |
| How much does it hold? | Not stated | 1.5 L, minimum 250 ml |
| Other colours? | Not stated | White, separate product code |
Descriptions are one layer
Good copy works best alongside structured data: stable identifiers such as a GTIN, consistent attribute names, and schema.org Product markup on the page. Our product data quality checklist covers the full catalogue audit, and structured retailer data for AI search explains what these systems read from public pages.
Simlir is a structured product-data API for AI shopping. It returns products as clean JSON with titles, identifiers, specifications, images and dated prices, so the clearer your facts, the more accurately a shopping assistant can represent them. For the catalogue side, see how retailers can make catalogues easier for shopping assistants to understand.
