
Platforms offer AI tools to generate or adapt ad creatives: backgrounds, text variations, images and videos. They save time, but call for brand and risk controls: human review, fidelity to the product, rights and transparency when synthetic content may be mistaken for real.
The essentials in a definition
Generated creatives on Meta and other platforms are ads created or adapted by AI from supplied assets. They require brand controls, fidelity to the product (UCPD), transparency (AI Act) and knowledge of the options (Meta’s help pages).
What the tools do
Meta’s creative tools and those of other platforms can generate variations of images, backgrounds, text and formats from assets you provide. Consult Meta’s help pages to know what is available, how to enable or disable each option and what controls you have. Features vary and evolve.
Responsibility for the published ad remains with the advertiser.
Brand controls
Set clear rules: tone, forbidden terms, visual elements that cannot be altered, and when a variation is acceptable. Disable options that alter the product’s appearance unfaithfully. Keep a set of approved assets as a base and review all variations before publishing, or in samples if volume is large.
Check default changes before activating campaigns.

Generated creatives with brand control
- Map active AI options
- Define brand and product rules
- Review variations before publishing
- Compare with base creatives
Legal and transparency risks
The Unfair Commercial Practices Directive prohibits misleading: the image must match the product sold and claims must be true. The AI Act provides transparency obligations for certain synthetic content, with gradual application, and platforms have their own labelling rules. Confirm the applicable requirements.
Avoid realistic AI-generated people in testimonials.

Whoever publishes the ad answers for it, whether or not AI generated it.
Measure and review
Compare generated creatives with base creatives under equivalent conditions and track performance, complaints and disapprovals. Keep a record of the tools and versions used. Review the policy every quarter.
If a tool generates results you do not approve, switch it off for that case.
Decision table
| Risk | Example | Control |
|---|---|---|
| Unfaithful image | Product altered or idealised | Disable product alterations |
| Improper text | Unsupported claim | Review and text rules |
| Realistic synthetic content | Generated person in testimonial | Avoid or label |
| Loss of identity | Style different from the brand | Assets and brand rules |
Common mistakes to avoid
- Accepting all suggestions. Changes can affect the brand.
- Not reviewing. Errors reach the public.
- Not recording. Compliance is hard to prove.
How to measure the result
Track performance, disapprovals and complaints of generated variations against base creatives.
Limits and exceptions
Tools and rules change fast, and this content is not legal advice. The quality of generated variations is uneven. Always keep human review and confirm current options.
Illustrative example
Illustrative example: a cosmetics brand enables automatic background variations, but disables those that alter the product’s look. It reviews a weekly sample, withdraws two variations that change the product colour and keeps a log. It compares performance with base creatives and approved variations.
In practice: checkpoints
Write a short guide on what the tools may and may not change in your creatives, and share it with agencies. Keep approved and rejected examples, to calibrate the team. Review samples of variations weekly at first and ask agencies to declare their use of AI. Check that images of products, people and places are faithful and that no unsupported claim has crept into the text. Keep a record of the options enabled in each account, because defaults can change, and compare the performance of generated variations with your own base creatives over several weeks before enabling more automation.
Frequently asked questions
Can I enable all automatic options?
You can, but with controls and review; disable those that alter the product.
Do I need to label generated content?
It depends on content type and applicable rules; when in doubt, be transparent.
How do I avoid misleading?
By ensuring fidelity to the product and true claims.
How often should I review?
Weekly at first and then monthly, in samples.
What about client accounts?
Inform and agree controls and responsibilities in writing.
Use AI creative tools while keeping brand rules, fidelity to the product, review, transparency and comparison with base creatives.
Sources and recommended reading
- Artificial Intelligence Act (AI Act)
- Unfair Commercial Practices Directive (2005/29/EC)
- Meta Business Help Centre
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.