EraseAILabelHome / AI info on Facebook ads
AI Info on Facebook Ads: What Advertisers Should Know
Meta labels some ads that were made or significantly edited with AI, and it says it will also detect ads made with third-party tools. Here's what Meta has said, what it hasn't, and how to prepare.
Published by Km Anjali · Last reviewed November 2026
Which ads get labeled
Meta's ads transparency post, first published in February 2025 and updated in June 2026, sets out the approach. The label depends on what the AI did and on which tool was used.
| Ad creative | What Meta says |
|---|---|
| Significantly edited with Meta's generative AI features | Labeled |
| Includes a photorealistic person from Meta's tools | Labeled, shown next to Sponsored |
| Meta's tools, no significant edit, no photorealistic person | No label |
| Made or edited with third-party AI tools | Meta says it will begin detecting these through industry-standard signals, with AI info shown in About this ad |
Meta adds that the experience "may vary in certain regions due to legal requirements," so if you advertise in several countries, check a live ad in each one.
Where the label appears
Meta describes "About this ad" as a unified place for transparency information, inside the three-dot menu on every ad. So most viewers will need to open the menu to see the label. The exception that matters most for creative teams is a photorealistic person made with Meta's tools, where the label sits next to "Sponsored" and is visible straight away.
What Meta hasn't said
Being clear about the gaps will save you from planning around guesses. In the pages we read, Meta doesn't say:
- How complete its detection of third-party AI tools is, or how far it has rolled out.
- Exactly which signals count as "industry-standard" for ads.
- Whether advertisers can appeal or remove a label that Meta applied.
- What effect, if any, a label has on delivery or cost.
Marketing blogs often state these points with confidence. Unless they link to Meta, treat them as opinion. Run your own comparison if the answer matters to your budget, and change only one thing at a time.
Political and social issue ads have stricter rules
Since 2024, Meta has required advertisers to disclose when a political or social issue ad contains a photorealistic image or video, or realistic-sounding audio, that was digitally created or altered in certain ways. If you run this type of ad, read the current policy in Ads Manager before launch, because the consequences of getting it wrong are more serious than a label.
A simple creative review
Labels tend to surprise teams because nobody tracked which tools touched an asset. A short routine fixes that:
- Keep a tool log. Note every AI feature used on each asset, including upscaling, background extension, and object removal.
- Ask your agency and freelancers. Make it a standard question on each delivery, and ask for the source files.
- Scan final exports. Run a sample through the AI metadata checker to see whether files carry Content Credentials or XMP data.
- Check live ads. Open the three-dot menu and About this ad on your running ads, since that's what your customers can see.
- Decide who owns the answer. If a customer asks about a label, someone should know how the image was made.
Should advertisers strip metadata from AI creative?
We don't recommend it as a way to avoid a label. Removing records from AI-made creative doesn't change how it was made, and it can undermine trust if a customer, a partner, or a regulator later asks. Advertising rules in your market may also require disclosure regardless of what any platform shows, so ask a qualified adviser if you're unsure.
Cleaning a file does have legitimate uses in ad production. Examples include removing GPS data from photos taken at a customer's home, clearing internal file names and editing history before files go to a vendor, and shrinking files by dropping unused metadata. None of these depends on hiding AI use.
Where AI fits well in ads
A label isn't a verdict on your ad. It's likely to matter most when people expect to see something real. Customer testimonials, before-and-after results, product demonstrations, and photos of your actual team all depend on being genuine, so use real photography there. AI is better suited to backgrounds, concept tests, and abstract visuals, where nobody assumes that they're looking at a record of something that happened.
An entrepreneur on the EraseAILabel team who writes about photo privacy and AI labels. How we test
Common questions
Do all AI-made Facebook ads get an AI info label?
No. Meta labels ads significantly edited with its own generative tools or showing a photorealistic person from them, and says it will detect some third-party AI ads through industry-standard signals.
Where do people see AI info on an ad?
Usually in About this ad, inside the three-dot menu. A photorealistic person made with Meta tools gets the label next to Sponsored.
Does a Meta AI info label hurt ad performance?
We could not find a Meta statement on this. If it matters to your budget, run your own controlled comparison.
Can I remove AI info from an ad?
Meta pages we read do not describe a way to remove or appeal a label it applied. Check Ads Manager for current options.
Check your final exports
See which records a file carries, without uploading it.