Cosmetics product sheets: write faster without taking risks on claims

Implementation example. This scenario describes a request we receive in this form. It is not a project delivered to a named client, and no results are claimed.

The context

A cosmetics brand (face, body or hair care) sells on its own website, on marketplaces and through a few online retailers. It launches several products a year, with variations in size and gift sets. The marketing team has two to five people. A regulatory affairs manager, in-house or external, approves what can be said about each product.

The tools in place:

an online store (Shopify, PrestaShop, Magento or similar);
sometimes a PIM, more often a product spreadsheet;
the product information file (PIF) and the safety assessment reports, held by the regulatory affairs manager;
launch briefs as PDFs or slide decks.

The day-to-day problem

Every product sheet goes through several rounds of back-and-forth. Marketing writes copy that makes people want the product. The regulatory affairs manager removes “anti-aging,” “hypoallergenic” or “100% natural” for lack of evidence or correct wording, and the copy comes back. Lead times stretch out before every launch.

When the team tries a generic AI tool, it's worse: it writes promises the brand can't keep, and invents active ingredients or percentages. No one dares to publish without rereading everything line by line.

On marketplaces, the sheets are copied or shortened by hand, and wording removed from the website stays online elsewhere.

What the brand wants

“I want sheets that make people want to buy, written fast, but that only use what we're allowed to say. And I want the regulatory affairs manager to approve once, not five times.”

The solution, step by step

1. Build the library of approved claims. With the regulatory affairs manager, we list for each product what the brand can say, with the associated evidence (test, consumer study, ingredient data) and the exact approved wording. Rejected wording and the reason for rejection are recorded too. This library is the reference for everything that follows.

2. Gather the product data. Name, range, size, skin or hair type, directions for use, INCI list, precautions for use, price and visuals. Everything factual comes from the spreadsheet or the PIM, never from the AI.

3. Have AI write the sheets, within a closed framework. The AI writes the title, the hook, the description, the benefits and the usage routine in the brand's tone. It can only use the claims in the library, in their approved wording. If the brief asks for a benefit that isn't in the library, it doesn't write it and flags it.

4. Check automatically before review. A second pass checks each sheet: sensitive terms (“free from,” “natural,” “hypoallergenic,” “anti-aging,” medical references), figures not found in the library, consistency with the INCI list. Each issue raised is highlighted for the reviewer.

5. Have the regulatory affairs manager approve. She receives sheets whose claims already match the library, with only the points to decide highlighted. Her decisions feed the library for the next sheets.

6. Adapt for each channel. Once approved, the sheet is adapted to the formats of the website, marketplaces and retailers, without rewriting the claims.

7. Keep a record. Each published sheet is linked to the version of the library used, which is useful if an authority or a distributor raises a question.

The regulatory framework to follow

In the European Union, cosmetic claims fall under Regulation (EC) No 1223/2009 and the common criteria of Regulation (EU) No 655/2013: legal compliance, truthfulness, evidential support, honesty, fairness and informed decision-making. The AI does not judge compliance. It strictly applies what the regulatory affairs manager has approved, and the final decision remains human.

What the AI does, what the brand does

The AI writes, adapts and checks consistency with the library. The brand provides the data, decides on the claims, reviews and publishes. The regulatory affairs manager keeps control over everything promised to the consumer.

Data and access required

the product spreadsheet or read access to the PIM;
the approved claims and their substantiation, or a summary of them;
a few existing sheets considered successful, to set the tone;
the list of sales channels and their format constraints;
a regulatory reviewer available during setup.

Project timeline

Weeks 1 and 2: diagnostic. Analysis of a sample of sheets, the current back-and-forth and the channels. This is the format of our audit (8 to 10 days).
Weeks 3 and 4: library and tone. Building the claims library for a first range, tone guide, checking rules.
Weeks 5 and 6: production. Writing and checking the sheets for the first range, approval, adjustments.
Then: extension. The other ranges, then new launches as they come.

Deliverables

a reusable library of approved claims;
a tool for writing and checking sheets;
the approved sheets for the first range, for each channel;
a brand tone guide;
a documented approval procedure;
a short training session for the team.

Metrics to track

the time between a launch brief and the publication of the sheets;
the number of rounds between marketing and regulatory per sheet;
the number of issues raised by the automatic check, then corrected;
the share of sheets that are up to date and identical across all channels;
the store's product page metrics (add-to-cart rate, conversion).

These metrics are recorded at the start, then tracked at each launch. No results are promised in advance.

Limitations and prerequisites

Without approved and documented claims, the library must be built first: it is the longest part, and the most useful.
AI does not replace the regulatory assessment or the tests that substantiate a claim.
For borderline products between cosmetics and medicines, specialized expertise is still required.
Rules outside the European Union (United Kingdom, United States, Asia) require one library per region.

Indicative budget

The diagnostic corresponds to our audit, from €1,000 excl. VAT. The next phase is quoted on request, depending on the number of products, channels and languages.

Frequently asked questions

Can the AI tell whether a claim is compliant? No. It applies the approved claims and flags anything outside the framework. The decision belongs to the regulatory affairs manager.

What happens if marketing wants a new promise? It first goes through regulatory approval. Once approved, it enters the library and can be used everywhere.

Do we need a PIM? No. A well-maintained product spreadsheet is enough to get started.

Can the sheets be translated? Yes, keeping the approved wording for each language. That is the subject of another scenario on content localization.

What next?

If every product sheet goes back and forth several times between marketing and regulatory, the audit shows you in 8 to 10 days how to frame the writing around your approved claims.

Client challenge