In this guide · 8 sections
- 01How Meta decides to label AI content
- 02Why an edited product photo can get an AI label
- 03When you are required to add an AI label
- 04How AI labels work on Facebook and Instagram ads
- 05An honest-disclosure rule for small business posts
- 06What to do if a post is labelled and you think it is wrong
- 07Do it yourself or get help?
- 08Frequently asked questions
Short answer
Meta labels content when it detects industry AI signals (such as C2PA) or when you self-disclose. On posts, you must add a label to photorealistic video or realistic audio made or altered with AI; images are exempt. Ads made or edited with AI tools get an AI info label under About this ad.
How Meta decides to label AI content
Meta first announced its approach in April 2024 and has changed the wording since. Its newsroom post on labelling AI-generated content explains the history: labels began as "Made with AI", became "AI info" on 1 July 2024 after complaints that lightly retouched photos were being labelled, and from September 2024 content that was only edited with AI had its label moved into the post menu.
Meta's current help page on how AI-generated content is identified and labelled lists three routes:
- Detected signals: files made with third-party tools may carry industry-standard signals. Content with signals showing it was created with AI is labelled AI content on the post; content modified with AI shows the detail in the post's menu instead.
- Self-disclosure: people can, and sometimes must, label their own AI content when posting.
- Meta's own AI tools: content created or edited with Meta AI and shared as a post, story or reel may be labelled automatically or carry a visible watermark.
Meta also says not all AI content contains the information needed to identify it, so labels are not a reliable AI detector.
Why an edited product photo can get an AI label
Most surprise labels come from editing tools, not from fully generated images. Meta's help page gives the example directly: if a photo editing tool uses AI to change the size or colour of an image, the file may carry a signal that it was edited with AI, and Meta's systems read that signal.
| What you did | Likely result on an organic post |
|---|---|
| Shot a photo on a phone, posted as is | No AI label |
| Used an AI tool to remove a background or object | Modified-with-AI detail in the post menu, if the tool wrote the signal |
| Generated a whole image with an AI image tool | AI content label on the post, if the tool wrote the signal |
| Generated with Meta AI and posted | May be labelled automatically or watermarked |
| Removed the metadata before posting | May avoid a label, but does not change whether disclosure is required for video or audio |
When you are required to add an AI label
Meta's rule for organic posts is narrower than many people think. On Facebook, Instagram and Threads, Meta may require an AI label when content has photorealistic video or realistic-sounding audio that was digitally created, modified or altered, including with AI. Meta says the requirement does not apply to images, though images can still be labelled if detected. There may be penalties for content shared without a label when one is required.
Label required
- Realistic AI video of people in a market
- An audio clip of two people talking, AI-made
- A song with AI-generated vocals
- A reel with a realistic AI voiceover
Best for: Meta's own examples of content that needs a label.
Label not required
- Cartoon-style AI video of a landscape
- An AI image of a person riding a bull
- Ordinary photo edits such as cropping or colour
Best for: Meta's examples where no label is required (images may still be auto-labelled).
To disclose, use the AI label option when you create the post, reel or story. Accounts that mostly post AI content can also add an "AI creator" label to their profile.
How AI labels work on Facebook and Instagram ads
Ads follow a different system. Meta's business help page About AI info on ads says it automatically labels ad content created or edited with some of Meta's generative AI tools (Background Generation, Image Generation, Add Animation) or with third-party tools such as Photoshop or DALL-E, which it detects using standards like C2PA.
- The AI info label usually sits in the three-dot menu under About this ad, sometimes next to the Sponsored or Ad label.
- Ads Manager shows a message before publishing when an AI info label will appear.
- Minor changes such as resizing or colour correction do not get a label, per Meta's ads labelling help.
- Boosted posts follow the organic labelling rules, and an organic label may stay on the boosted ad.
- Self-disclosure options exist for ads delivered to regions with specific AI rules: currently the European Region, California, New York, India and Taiwan. Singapore is not on that list as of October 2026.
An honest-disclosure rule for small business posts
Meta's rules set the minimum. Customer trust sets a higher bar, especially for food, beauty, property and renovation businesses where the photo is the promise.
AI disclosure rule for your page
For more on producing AI social content that still looks like your brand, see our guide to AI social media content for small businesses. Commercial rights to AI images are a separate question, covered in AI image generators for commercial use in Singapore.
What to do if a post is labelled and you think it is wrong
Open the label to see why
Tap the label or the post menu. Meta says it shows whether the label came from detected signals or self-disclosure.
Check the original file
Run it through the Meta AI Detection tool or check the editing app's export settings for content credentials.
Re-export without AI-edit steps if they were not needed
If a simple crop triggered it through an AI-powered tool, redo the edit with a non-AI tool and repost if it matters.
Accept labels on ads
Meta says advertisers cannot request removal of automatically detected labels on political ads; treat ad labels as permanent once applied.
Do it yourself or get help?
Most small businesses only need the disclosure rule above and a quick check of which editing tools write AI metadata. That is an afternoon of testing.
If you want ad creatives produced with real product photography and honest AI use where it helps, our Extra Ad Creative Pack (S$290, paid upfront) delivers ready-to-run images and copy. For a full Facebook and Instagram ads setup with verification and tracking, see Meta Ads Setup (S$590). IT Sorted is a Singapore-based team of developers with over 10 years of experience.
Frequently asked questions
Usually because the file carries an industry-standard signal, such as C2PA metadata, that an AI tool created or edited it. Meta's help centre says even AI features that resize or recolour an image can add that signal. Fully generated images get a label on the post; AI-edited ones show the detail in the post menu.
Meta's rule requires a label for photorealistic video and realistic-sounding audio created or altered with AI, and says the requirement does not apply to images. Images may still be labelled automatically if Meta detects AI signals. Labelling realistic AI images anyway is a good trust practice for businesses.
If you use Meta's generative features such as Background Generation, Image Generation or Add Animation to create or significantly edit an ad, Meta says it automatically applies an AI info label, usually under About this ad in the three-dot menu. Ads Manager shows a message before publishing when this will happen. Minor edits like resizing are not labelled.
If you added it yourself, you can edit the post. If Meta applied it because it detected AI signals in the file, the practical fix is to re-create the image without AI-edit steps and post again. Meta says advertisers cannot request removal of automatically detected labels on political and social issue ads.
Meta's ads help page lists the regions where it offers specific AI self-disclosure for ads: the European Region, California, New York, India and Taiwan, as of October 2026. Singapore is not listed. General advertising rules still apply, so AI images must not mislead customers about the product. This is not legal advice.