Use case · AI content creation

AI product descriptions for whole catalogues.

AI can write distinct product descriptions for an entire catalogue, replacing the manufacturer text every other retailer uses. Quality depends on the input: descriptions generated from structured PIM attributes are accurate, while those generated from a product name are guesses. Each marketplace has its own rules, and claims such as 'eco-friendly' or health benefits need control, not creativity.

Replacing duplicate manufacturer text

Most retailers receive the same supplier copy as their competitors, so their product pages read identically. Google does not penalise duplicate content as such, but when many pages say the same thing it picks one version to show, and there is no reason it should be yours. Unique copy gives search engines and shoppers a reason to choose one page: answers to the questions buyers actually ask, comparisons within the range, use cases, care and fit advice. Rewriting the supplier's sentences with synonyms does none of that. Write from the attributes and the buyer's questions, not from the old description.

Attribute accuracy starts in the PIM

A description is only as accurate as its source. Generate from structured attributes (dimensions, materials, compatibility, capacity, care instructions, certifications) and forbid the writer, human or machine, from adding facts that are not in the data. A missing attribute should raise a flag, not a plausible guess. Some information is mandatory before purchase: the EU's textile regulation (Regulation (EU) No 1007/2011) requires fibre composition to be visible before an online sale, and Regulation (EU) No 1169/2011 requires food sold at a distance to show its mandatory information, apart from the best-before or use-by date, before purchase. Convert units and formats per market from the data, not in the prose.

Marketplace rules and claims control

Each channel sets its own limits. Amazon now caps titles at 75 characters for most product types, with a separate 125-character item highlights field, and bars promotional phrases and certain special characters from titles; Google Merchant Center allows titles up to 150 characters and descriptions up to 5,000, with no promotional text. Claims need a register: which environmental, health, safety and origin claims are approved, in which markets, in what wording. In the EU, Directive (EU) 2024/825 bans generic environmental claims such as 'eco-friendly' or 'green' without recognised excellent environmental performance, from 27 September 2026. Give the register to whatever writes the copy, and check every output against it.

Multilingual catalogues

Translating a catalogue line by line produces text that is grammatically right and commercially flat. Localise instead: sizes and units per market, local names for product types, the search terms shoppers actually use in that language, and legal wording that differs by country. Keep a terminology base so a feature has one name per language across the whole range. Synthetic White runs batches from a PIM or product feed and localises copy into 30 languages with terminology and approved claims enforced, while the brand hub applies tone of voice and banned words at generation rather than in review.

Updated 25 September 2026

Questions

AI product descriptions, answered.

Will AI product descriptions hurt our SEO?

Not if they are accurate, specific and useful. Google judges content by its helpfulness, not by how it was produced. The risk is mass-produced text that adds nothing: thousands of pages saying the same thing in slightly different words. Descriptions built from real attributes and real buyer questions avoid that.

Do we need to label AI-written product text?

In most markets there is no general legal duty to label AI-written product copy, but some channels ask for it. Google Merchant Center requires titles and descriptions created with generative AI to be submitted in its structured title and description attributes. Check each channel's current specification before a feed goes live, because these requirements are still changing.

How do we stop AI inventing product features?

Constrain the input and check the output. Generate only from the product's structured attributes, instruct that no facts outside them may appear, and compare each claim in the output with the data before publishing. Missing information should trigger a request to the data owner, not creative filling.

Get started

See your AI studio generating this week.

Thirty minutes. A live studio in your colours, your brand hub loaded, a real campaign brief.

See plans