Why Instagram shows an AI Info label
Meta shows AI Info on Instagram and Facebook from C2PA or IPTC signals, or from your upload disclosure. It is not a detector. A real photo can be labelled.
The AI Info label on Facebook and Instagram is usually explained as if it were a detector: Meta looks at your image, decides whether it is AI, and labels it. That model predicts the wrong things. It cannot explain why a photograph you shot yourself gets labelled, and it cannot explain why a fully generated video sails through without one.
What Meta has actually published is narrower. It began applying "AI Info" labels when it detected industry-standard AI image indicators, or when people disclosed they were uploading AI-generated content. Its current help page still describes those two families: signals in the file, and self-disclosure. The file signals are at least two different metadata systems, put there by tools you did not necessarily choose for that purpose.
The signals Meta has named
| Signal | Where it lives | Who puts it there | What removes it |
|---|---|---|---|
| C2PA manifest | Cryptographically bound file metadata | The generating or editing tool, if it signs | Any re-encode, screenshot or non-aware edit |
| IPTC DigitalSourceType | Standard image metadata field | Whatever tool wrote the file | A metadata strip, a re-encode, a screenshot |
| Self-disclosure | Meta's own records for the post | You, at posting time | You, by not using the toggle |
In its February 2024 announcement, Meta named C2PA and IPTC as the industry-standard indicators it would read, and added a disclosure tool for AI-generated video and audio that did not yet carry those indicators at scale. Content made with Meta's own AI tools is labelled separately, and sometimes carries a visible watermark as well.
C2PA manifests are signed Content Credentials, hard-bound to the exact bytes of the asset. When a manifest survives to upload, it is the highest-confidence signal available: the file declaring itself rather than a platform guessing.
IPTC DigitalSourceType is an ordinary, unsigned metadata field indicating how the media was produced, including values for media produced by a trained algorithm. Any tool in your chain can set it, and plenty do automatically, without surfacing that they have. It has no notion of degree. There is no "12% generative fill" value.
Self-disclosure is the posting-flow toggle. It is the only signal that is entirely yours. For photorealistic video and realistic-sounding audio, Meta's help page says the label may be required even when the file carries no industry-standard signal. That requirement does not apply to images, which may still be labelled if a signal is detected.
Two things get treated as extra signals and should not be. Meta said in 2024 that it was developing pixel-level classifiers to catch AI content that lacked invisible markers; its current organic-labelling help does not list a classifier as a separate trigger. Do not plan as if a stripped file will still be labelled because "the detector will catch it." And Meta now distinguishes created from modified: signals that the content was created with AI put an AI Info label on the post; signals that it was only modified with AI put the note under Options instead. A camera-original photograph that went through a generative edit can land in either bucket, depending on what the editor wrote into the file.
The two you actually control
You control the metadata and the toggle. Those cover both failure modes people actually care about.
The surprise label is a metadata problem. If any tool in your chain writes a DigitalSourceType value indicating algorithmically produced media, that assertion is now in the file. A camera-original photograph that passed through one editing operation arrives carrying a flat claim about its origin, and the on-post label can fire if the tool wrote the field as "created." The check before you post is not to look at the image. It is to look at the file's metadata.
The missing label is the same problem inverted. A fully generated video re-encoded on its way to upload arrives with no C2PA manifest and no DigitalSourceType, because neither survives a transcode. What remains is the toggle. If you do not flip it, a wholly synthetic asset can publish with nothing on it. Meta's help page concedes that not all AI content contains the information needed to identify it. That gap is not a permission: for photorealistic video and realistic audio, Meta may still require the disclosure. The full walk of where metadata dies in a pipeline covers the mechanics. Assume the metadata is gone unless you have looked.
Avoiding a surprise, in five minutes
- Inspect the export, not the project. Open the file you are about to upload and look at its metadata. That is the diagnostic.
- Find which tool writes the field. If DigitalSourceType is present and you did not expect it, export the same asset from each stage. The first stage where it appears is your writer, almost always an editor with a generative feature.
- Decide, rather than discover. For anything with a real AI contribution, letting the file declare itself is doing you a favour. For a trivial generative touch, that is a conversation to have on purpose rather than to lose to a metadata field.
- Use the toggle regardless of what the file carries. The toggle and the metadata are independent, and a mismatch between them is what looks bad.
- Keep the copy consistent across destinations. The same video should not read as candid on one platform and disclosed on another. The cross-platform labelling checklist is the per-launch version, and profile-level versus post-level disclosure decides which one you need where.
Ads are a separate rulebook
Everything above is about the organic label. For ads about social issues, elections or politics, Meta requires advertisers to disclose when the ad uses AI to create or edit photorealistic images, video or realistic audio in specified cases. That is a duty you owe, not a label inferred from the file. From 1 June 2026 Meta has also said it uses automated detection on this category and applies an AI Info label in "About this Ad" when it finds AI-generated or edited media. Do not let a clean organic-post inspection train the reflex that silence is the default on ads. Meta ads creative workflow is the production side of that split.
New York added a narrower layer: since 9 June 2026, an advertisement reaching New York consumers that features a fully synthetic human performer needs a conspicuous on-ad disclosure. A platform label does not discharge it. New York's synthetic performer ad disclosure is the scoped version.
What the label does and does not mean
Read the AI Info label as "a file signal or a disclosure fired," not as "this content is AI." It can fire on a real photograph with an eager metadata field. It can fail to fire on a fully generated video whose metadata was stripped. It is a compromise between provenance data Meta can trust and detection it cannot fully rely on, not an authoritative verdict. What AI detection and labels actually establish is the broader argument; what an AI content label is is the one-page definition.
Your disclosure obligation is defined by law and by your own disclosure policy. The platform label is a signal that may or may not coincide with it. Build the policy first. Let the label be a consequence.
FAQ
Can I remove the metadata to avoid being labelled?
Technically yes, and it is a bad trade. Stripping metadata produces a file indistinguishable from one that never carried provenance, while leaving any pixel-level watermark a provider applied fully intact. You have removed the cooperative signal platforms actually read and kept the robust one. Machine-readable marking under Article 50 in the EU and California's provider-side marking rules both applied from 2 August 2026; those duties sit primarily on the tools that generate the file, not on you as a reason to strip what they attached. For photorealistic video and realistic audio, Meta may still require the toggle after the metadata is gone.
Why did my photo get labelled when I barely used AI?
Because the signal that fired was almost certainly a metadata field asserting algorithmic origin, and that field has no notion of degree. One generative operation anywhere in your editing chain can set it. Inspect the file, find which stage writes it, and check whether the tool marked the file as created with AI or only modified — Meta treats those differently.
Does a C2PA manifest guarantee a label?
No, and the reason is that manifests frequently do not survive to Meta at all. A manifest is hard-bound to the exact bytes of the file, so any re-encode between your export and the upload invalidates or removes it. A manifest that arrives intact is a strong signal. A manifest that existed when you exported is not evidence of anything about what arrived.
Should I flip the toggle on stylised, obviously synthetic work?
Yes, on consistency grounds rather than deception grounds. Nobody is fooled by an obviously rendered piece, but an inconsistent disclosure policy across your own feed makes the disclosed posts look like the exceptions. The threshold for a visible on-screen disclosure differs by content type. The threshold for using the platform's own toggle does not need to.