AI Terms

1. What the AI-assisted features do

When you turn on AI assistance for product descriptions, the Service may (a) suggest values for product attributes (such as ingredients, weight, warranty or usage instructions) by reading the description text already present in your own supplier feed, and (b) rewrite the tone of the assembled description text. That is the entire scope: the AI-assisted features work on text only.

2. No image generation

The Service does not generate, alter or select product images with AI, and does not offer an AI image feature of any kind. Every image that reaches your store comes from the URL your supplier feed provides.

3. AI output is a suggestion, not a verified fact

AI-suggested attributes and rewritten text are assistive suggestions, generated by a third-party language model from the text you supplied. They are not independently verified, and we make no warranty of accuracy, completeness or fitness for a particular purpose regarding AI-suggested or AI-rewritten content. A language model can misread a source description, omit a detail, or phrase something in a way that changes its meaning.

4. You review before it reaches your store

AI-touched content is never pushed to your store automatically. It is staged and requires your explicit action — pressing "Apply" in Templates — before the next sync can publish it, exactly the same gate that applies to non-AI content. Pressing Apply is your confirmation that you reviewed what would be published. You remain solely responsible for what appears in your store, including content produced or modified with AI assistance, in the same way you are responsible for content you typed yourself (Terms §4, §7).

5. What we record as evidence of your review

For every applied product description, we record which specific fields were AI-assisted (as opposed to assembled deterministically from your feed), which version of these AI Terms was in effect, and who applied it and when. This is the record that a human reviewed and approved the content before it published — kept because "we assume someone checked" is not evidence.

6. Flagged content — surfaced, never silently handled

Three kinds of suspected problem are detected automatically and shown to you with a reason, rather than being fixed or ignored on your behalf:

• Suspected watermarked or third-party images — an image URL that matches a known stock-photo service or carries a watermark/preview marker is refused: it is not sent to your store, and you are told why.
• Missing manufacturer identification — a product with no brand name in its feed data is flagged, because EU GPSR requires you to be able to identify the manufacturer or responsible importer of what you sell.
• Text that reads as a regulated claim — description text matching a health, safety or certification claim pattern (for example a medical claim or a "certified" claim) is flagged for your own legal review before it publishes.

These are heuristic checks, not a legal or forensic determination, and they do not replace your own review duty under §4 and Terms §7. A flag clears itself once a later sync no longer finds the condition true for that product.

7. Provider

AI-assisted features are powered by a third-party AI provider, named — together with the country it processes in — in the sub-processor table in our Privacy Policy once the feature processes anything of yours on our account (as opposed to a provider key you supply yourself, in which case your own agreement with that provider governs).

8. Turning it off

AI assistance is opt-in per template and can be turned off at any time in Templates. Turning it off stops future AI suggestions; content already applied to your store is unaffected until you change it.

9. Contact

Monine AS, Norway — org. no. 926 029 215 — hello@ovanap.no.