You took the photo yourself. You removed one stray cable with a brush stroke, posted it, and Instagram put “AI info” on it. To your followers, that reads like the picture is fake. It is not. The Instagram AI info label is a note about how the file was made, and in most cases it comes from a few bytes of metadata your editing app wrote into the file. This guide explains what the label means, what trips it, and how to check a file before you post it. Checked 15 November 2026: label rules change often and quietly, so the dates and the steps below are the current ones.
Short answer: The Instagram AI info label is a note Instagram adds to a post when it detects industry standard AI signals in the file, such as C2PA Content Credentials or IPTC metadata, or when the person posting discloses AI use. It says AI was involved in making or editing the image. It does not say the whole photo is fake.
What does the Instagram AI info label mean?
Definition: The Instagram AI info label is a note Instagram puts on a post, reel or story to say that the media in it was created or changed with artificial intelligence. It is a piece of information about the file, not a verdict on the person who posted it.
Tap the label and Instagram tells you more about why it is there. In the current app, the note says whether the label came from industry signals found in the file or because the person posting said AI was used. That distinction matters, because the two routes mean different things. A signal in the file can be triggered by a single small edit. A self disclosure means the poster chose to say so.
The name itself has changed twice, which is why older guides disagree with each other. Meta first called it “Made with AI”. Photographers complained that lightly retouched photos, images that had only been edited and not generated, were being tagged as AI made. On 1 July 2024 Meta renamed it “AI info”, a softer name meant to read as information rather than an accusation. The wording still varies by surface: a reel we checked on the Instagram website on 10 November 2026 showed “AI content”, a small grey button directly under the username in the post header, instead of “AI info”. Do not be surprised if you see both wordings in different places. They are the same label doing the same job.
The label is also not the only AI note on Instagram. An account built around an AI generated person can carry a separate “AI-generated profile” label, which sits on the profile and next to that account’s content. That label is about the account, not about one post. Our guide to AI labels on social media covers the same label on YouTube, TikTok and Facebook, and how each platform words it. This article stays on Instagram only.
Answer capsule: the label says AI was involved in making or editing the media. It does not say the photo is fake, and it does not say the poster tried to deceive anyone.
Where does the Instagram AI info label come from?
Instagram adds the label in two ways, and Meta has described both in its own announcements. The first is detection. Meta reads the “AI generated” information in two industry standards, C2PA Content Credentials and IPTC metadata, which image generators and many editors write into files automatically. The second is self disclosure. If the person posting switches on the AI label when they share, the post is labelled whatever the file says.
Meta also says that images made with its own Meta AI feature carry invisible watermarks, so some images are recognised from their pixels and not only from metadata. Its February 2024 announcement said it was building tools to read the C2PA and IPTC signals in images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock, and that it would require people to disclose photorealistic video or realistic sounding audio that was digitally created or altered. That announcement is the starting point for everything that followed, and you can read it in full on Meta’s newsroom post on labelling AI generated images.
The rules have changed several times since, and the changes explain most of the confusion about this label. The table below is the timeline as Meta published it.
| Date | What changed |
|---|---|
| 6 February 2024 | Meta announces it will label images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock using the “AI generated” information in the C2PA and IPTC standards. |
| 5 April 2024 | The “Made with AI” label is announced, based on industry shared signals or the uploader’s own disclosure. |
| 1 July 2024 | “Made with AI” is renamed “AI info”, after photographers complained about labels on lightly retouched photos. |
| 12 September 2024 | For content Meta detects was only edited with AI, the label moves into the post’s menu. Content detected as generated by AI keeps the label on the post. |
| 31 August 2026 | Instagram renames the account level “AI creator” label to “AI-generated profile”. This is a separate label for whole accounts, not for single posts. |
The September 2024 change is the one most people miss. Before it, an edited photo and a fully generated image got the same visible label. After it, they do not. Content Meta detects was generated by AI keeps the visible “AI info” label on the post, and the label says whether industry signals or self disclosure caused it. Content it detects was only modified or edited with AI gets the note tucked into the post’s menu instead, under Options, where almost nobody looks. Meta set this out in its approach to labelling AI generated content, which carries both the July rename and the September update.
Answer capsule: the label comes from file signals or from the poster’s own disclosure, and since September 2024 an edited photo and a generated image are labelled in different places.
What trips the Instagram AI info label?
The short version is that the label is tripped by what the file declares, not by what the photo looks like. Instagram does not look at your image and decide whether it is AI. It reads the signals in the file. Two kinds of metadata can carry an AI declaration, and they behave differently.
Content Credentials, known as C2PA, is a signed record of how a file was made: which app or camera created it, what was done to it, and with which tools. Adobe apps, image generators and some cameras write it. In a JPEG it sits in APP11 segments, in a PNG in a caBX chunk. Inside the record, each action can carry a digital source type. An action like c2pa.created with the source type trainedAlgorithmicMedia says, in so many words, that the image was created by generative AI.
The same vocabulary also exists as a plain metadata field, written in the XMP block of the file. The IPTC digital source type field is the one that matters most for still photos, because many editors write it even when they do not write a full C2PA record. Only a few values say generative AI. The matrix below is the Ranking Asset for this article: every trigger, what Instagram does with it, and what you can do about it.
| What happened to the file | Signal in the file | What Instagram shows |
|---|---|---|
| Image fully generated by an AI tool, such as Firefly, Midjourney or Meta AI | trainedAlgorithmicMedia, or a C2PA record with c2pa.created | “AI info” label on the post |
| Real photo edited with a generative tool, such as Generative Fill, Generative Expand or a generative remove | compositeWithTrainedAlgorithmicMedia | “AI info” note in the post menu, not on the post |
| Real photo edited on a phone with Magic Editor, Magic Eraser or Zoom Enhance | IPTC field saying “Edited with Google AI” | “AI info” note in the post menu |
| Image made with Meta AI, re-uploaded from a download | Invisible watermark in the pixels, plus IPTC metadata | “AI info” label on the post, even though the file was re-saved |
| Poster switches on Add AI label when sharing | Self disclosure, whatever the file says | “AI info” label on the post |
| Account built around an AI generated person | “AI-generated profile” label on the account | Profile label, shown next to content without its own label |
| Real photo, no generative step, no C2PA record | No AI declaration anywhere in the file | No label |
| File with a C2PA record but no AI declaration, such as a camera photo with colour adjustments recorded | C2PA present, digital source type absent or not AI | No label |
Most complaints come from photographers whose pictures are real but were finished with one generative tool. Generative Fill and Generative Expand are the usual suspects. Adobe says Photoshop automatically attaches Content Credentials to images made with these features when they are exported as PNG or JPG, or saved with Save As. Generative remove tools count too: removing a person or a wire with a tool that fills the gap with generated pixels counts as a generative edit, even if the result looks untouched. Phone gallery editors do the same. Since October 2024, photos edited with Magic Editor, Magic Eraser or Zoom Enhance carry IPTC metadata saying “Edited with Google AI”, which Google Photos shows in an “AI info” section of the photo’s details. Our guide to Google Photos and how a photo was made walks through that panel and what each entry means.
Two traps catch people out. The first is an edit you undid. Photoshop users report that a generative step, once used and then undone, can still be declared on export. The record keeps the action even when the pixels are back to how they were. The second is the re-save. A screenshot of your own photo, or a copy run through a messaging app, usually loses the metadata, and with it the label. That is not a reliable fix, and it is not an honest one when the image really was generated. It also strips the proof that helps you: Content Credentials are also how a real camera photo can show where it came from.
In all of these cases, the declaration is technically accurate. Some pixels were generated. What the label does not tell viewers is how much. A whole fake scene and a removed cable produce the same field, and that is why the label feels unfair to photographers. It is a note about the file, not a measure of how much AI was used.
Answer capsule: the label is tripped by AI signals written into the file, and a single generative brush stroke is enough to write one.
We checked four real files: what each one declares
Claims about this label are easy to make and hard to test, so we tested four real files on 15 November 2026 and read the bytes ourselves. Each file was inspected for the two things Instagram reads: a C2PA Content Credentials record and an IPTC digital source type. Nothing below is a guess. The file sizes and the findings are exactly what the inspection returned.
File one was a real camera photo: the Sydney Opera House in December 2008, a 2,544 by 1,406 JPEG of 1,271,018 bytes from Wikimedia Commons, with EXIF data and a Photoshop CS3 XMP block from the day it was taken. It has no JUMBF segments, no C2PA record, no C2PA actions and no digital source type anywhere in the file. There is nothing for Instagram to read, so no label is expected. A photo this old, finished in a pre AI editor, is the cleanest possible case.
File two was a real AI generated image: a 1,024 by 1,024 PNG of 1,025,247 bytes from the Adobe Firefly gallery, iridescent holographic headphones by Liz Hayes. It carries one caBX chunk of 84,798 bytes, a c2pa.created action, and the digital source type trainedAlgorithmicMedia. The claim generator is named in the file as Adobe Firefly, signed by Adobe Product Services. This file declares AI generation in plain terms, so it should get the “AI info” label on the post. Our guide to telling if an image was made with Adobe Firefly reads this same file in full, including the signed manifest.
File three was a C2PA test image, CA.jpg, 1,024 by 683 pixels and 166,864 bytes, made by the C2PA reference tools. It carries a JUMBF segment of 64,010 bytes, with the actions c2pa.opened and c2pa.color_adjustments, and the claim generator make_test_images with c2pa-rs version 0.33.1. It has no digital source type at all. This is the case most guides get wrong: a C2PA record is present, but it declares no AI. A file can carry Content Credentials and still get no label, because the record only says what was done, and what was done here was opening the file and adjusting the colour.
File four was another C2PA test image, C.jpg, 2,048 by 1,365 pixels and 132,518 bytes, from the same test set. It carries a JUMBF segment of 45,882 bytes and a c2pa.created action with a digital source type of algorithmicMedia. That value means pure algorithmic media, and it is not a generative AI declaration. Even a digital source type value does not always mean AI. Only trainedAlgorithmicMedia, compositeWithTrainedAlgorithmicMedia and compositeSynthetic say generative AI was involved.
| File | What it is | C2PA record | Digital source type | Label expected |
|---|---|---|---|---|
| Sydney Opera House photo, Dec 2008 | Real camera photo, 1,271,018 bytes | None | None | No label |
| Firefly headphones, 1,024 by 1,024 PNG | AI generated image, 1,025,247 bytes | caBX chunk, 84,798 bytes, c2pa.created | trainedAlgorithmicMedia | “AI info” on the post |
| CA.jpg, C2PA test image | Test file, 166,864 bytes | JUMBF, 64,010 bytes, opened plus colour adjustments | None | No label |
| C.jpg, C2PA test image | Test file, 132,518 bytes | JUMBF, 45,882 bytes, c2pa.created | algorithmicMedia, not a generative AI value | No label |
Three lessons fall out of that table. First, the label follows the declaration, not the picture. A real photo with no declaration gets nothing, and a generated image with a clear declaration gets the label. Second, Content Credentials on their own do not cause a label. They are a record, and the record has to say AI was used before anything happens. Third, a screenshot of any of these files would carry none of this. Screenshots have no C2PA record and no IPTC block, which is why a screenshot of a labelled post does not itself get labelled when it is reposted.
Answer capsule: we read four real files byte by byte, and only the one that declares trainedAlgorithmicMedia should get the label.
How do you check your file before you post?
Check the exact file you are going to upload, not the original and not a screenshot of it. The file on your phone after export is the file Instagram reads, and an export can add or drop metadata that the working file did not have. Run this check before you post, while you can still fix the file.
- Open the C2PA verify tool and drop the exact file in. The official tool at verify.contentauthenticity.org reads the Content Credentials in a file and shows the claim, the actions and the tools that touched it. It is free and it runs in the browser. If it shows a c2pa.created action with a generative source type, Instagram is likely to label the post.
- Read the verdict for the action, not just the headline. A file can have a valid C2PA record and still be fine to post. What matters is whether any action says AI created or edited the image. An “opened” or “colour adjustments” action with no AI declaration is not a problem. A “created” action with trainedAlgorithmicMedia is.
- Check the IPTC field too, because it travels separately. Some editors write the IPTC digital source type without writing a C2PA record, and some write both. A metadata viewer that shows EXIF, XMP and IPTC will show the field as DigitalSourceType. The values that matter are trainedAlgorithmicMedia for a generated image and compositeWithTrainedAlgorithmicMedia for a generative edit. Our guide to checking Content Credentials on a photo walks through the verify tool step by step.
- If the file declares AI and the photo is real, fix the file at the source. Re-export from the original without the generative step, and use a non generative healing or clone tool for small fixes. Then check the new export again. Do not just strip the metadata and post the same file, because the label is only one of the signals. Meta says its own AI images also carry invisible watermarks, and Google adds SynthID to images from its models. Those live in the pixels, and no metadata edit removes them.
- Check your editor’s Content Credentials settings before the next export. Some apps attach credentials automatically whenever a generative feature was used, and some let you turn the attachment off. Know which one yours is, so the next export does not surprise you. If the image really was made with AI, the label is correct, and keeping it is the honest choice.
Keep in mind what a metadata check cannot see. It cannot see invisible watermarks in the pixels, and it cannot see your own disclosure when you post. A clean file check means the file carries no AI declaration. It does not mean Instagram will never label the post, because a disclosure toggle or a watermark can still trigger one.
Answer capsule: check the exact export file with the official verify tool, read the actions rather than the headline, and re-export without the generative step if the declaration is wrong.
How do you add the label yourself?
Instagram’s Help Center sets the rule plainly. Meta requires you to label content you share that has photorealistic video or realistic sounding audio that was digitally generated or altered, including with AI. The label goes on before you share, and Meta says there may be penalties if you do not label content as required.
Images are treated differently, and this is the point most creators get wrong. Meta does not require you to label images that have been created or altered with AI. It may still label them itself when its systems detect AI signals in the file. So the toggle is optional for a still image, and the automatic label can appear whether you used the toggle or not.
For a post or a reel, create the post as usual and, on the screen where you share it, turn on the toggle next to Add AI label. Share the post, and viewers will see an AI content label on it. For a story, share the photo or video to your story, tap Options, then Label AI, then turn on the toggle next to Add AI label. The story path has one extra step, so check it separately rather than assuming it matches the post path.
The profile label is a third control, and it is not the same toggle. If your account regularly posts content featuring an AI generated person instead of a real human, go to your profile, tap Edit profile, and turn on the toggle next to AI-generated profile. Instagram says labelled accounts see no change in reach, while accounts it detects as AI generated and unlabelled can become ineligible for recommendations. The label then appears on the profile and alongside content that does not already have its own AI label.
Two practical notes. The labelling feature is not available on computers, so use the mobile app: Instagram’s Help Center says so directly. And an AI label is not an ad disclosure. A sponsored post still needs its own paid partnership label, and in the US a clear disclosure under the FTC rules. The AI label does not cover that.
Answer capsule: turn on Add AI label before you share a post or reel, use Options then Label AI for a story, and use the separate profile toggle for an AI generated person.
What does the label not mean?
The label causes three common mistakes, and all three come from reading more into it than it says.
First, a label does not mean the photo is fake. A real photo that went through Generative Fill to remove a cable is still a real photo of a real place, and it can still carry the label. The label says a generative tool touched the file. It says nothing about whether the scene in front of the lens was real.
Second, no label does not mean a photo is real. Most photos online carry no AI signals at all, because platforms strip metadata on upload, screenshots never have it, and a file can be re-saved to drop the record. An image made entirely by AI can reach Instagram with no label if every signal was stripped before upload. The absence of a label proves nothing either way, and that is the single most important thing to remember when you are judging someone else’s post. Our guide to telling if an image is AI generated covers the checks that work when the label is missing, including reverse search and the visual tells that survive a re-save.
Third, the label does not say how much AI was used. A whole fake scene and a removed cable produce the same field in the file, and the label looks the same in both cases. Tapping the label sometimes says whether the cause was industry signals or self disclosure, and whether the content was generated or only edited, but it never says how many pixels changed.
There is one more confusion worth clearing up. Meta also adds a “Captured with a camera” label when secure metadata confirms a camera took the photo or video. That label is the opposite signal: it says the file came from a camera. A missing camera label does not mean AI was used, any more than a missing AI label means the photo is real.
Answer capsule: the label does not mean fake, no label does not mean real, and the label never says how much AI was used.
You saw the label on someone else’s post: what now?
Start by tapping the label. It may tell you whether the cause was industry signals or the poster’s own disclosure, and whether the content was generated or only edited. That is more useful than the label on its own, because a disclosure and a detection mean different things. A poster who disclosed is being open. A poster who was detected may not have known the file carried a signal.
Then look at the account, not just the image. An account built around an AI generated person should carry the profile label as well. An account that posts a mix of real and generated work, with no pattern and no disclosure, deserves more care than one that labels everything. Check the profile for the same person or product appearing in different places, and check whether the account has a history or was created days ago.
Reverse search the image before you decide anything. A generated image often appears nowhere else, while a stolen real photo often appears on other profiles or stock sites. A zero on a reverse search never clears a photo, because a new image may simply not be indexed yet, but a match to a stock site or another account settles the question fast. Save the post first if you think you may need to report it, because a poster can delete it.
If the post is deceptive, report it on Instagram rather than arguing in the comments. Use the report option on the post itself and choose the reason that fits, such as false information or scam. If the image is of you and was made without your consent, use the platform report and the removal routes in our reporting guides, and keep the original files safe. Do not share the image on, even to warn people, because that spreads the very thing you are trying to stop.
Answer capsule: tap the label, check the account, reverse search the image, and report rather than argue.
Can you remove an AI info label?
There is no remove button. Instagram’s Help Center, as read by several independent checkers in October and November 2026, does not describe a way to remove a label it applied automatically. If the label is wrong, the fix is the file, not the post. Delete the post, fix the export so it no longer declares AI, and post the corrected file.
Some sites sell “AI label removers” that simply delete the metadata fields. There are good reasons not to treat that as a fix. It does not remove invisible watermarks, and Meta has said its own AI images carry them. It does not change the rules, and Meta requires a disclosure for realistic video and audio that was digitally created or altered, and says it may apply penalties when people do not disclose. And it removes the proof that helps you, because Content Credentials are also how a real camera photo can show where it came from.
The honest route depends on what the image really is. If the image was made with AI, the label is correct, and keeping it is the honest choice. Your followers may trust you more for it, not less. If your photo is real and the label came from one small generative edit, the better fix is a file that never declared AI in the first place: re-export from the original, use a non generative tool for the fix, and check the new file before you post it.
Label rules change often and quietly. Meta renamed this label once, moved edited content into a menu once, and renamed the profile label in August 2026. What is true in November 2026 may not be true in six months. Check the current Help Center page before a campaign, and re-check a file before every important post, not just the first one.
Answer capsule: there is no remove button, so fix the file at the source and repost, and keep the label when the image really was made with AI.
FAQs
Why did Instagram put “AI info” on my real photo?
Because a tool you used wrote an AI declaration into the file, even if the photo itself is real. Generative Fill, a generative remove tool, or a phone AI edit writes a field that says AI edited the image. Instagram reads that field when you upload. Re-export from the original without the generative step and check the new file first.
Does an AI info label mean a photo is fake?
No. The label says AI was involved in making or editing the file, not that the whole photo is fake. A real photo that went through Generative Fill to remove a cable can still carry it. It never says how much AI was used, so judge the post on more than the label alone.
Can I remove the AI info label from my post?
There is no remove button, and Instagram’s Help Center does not describe a way to take off a label it applied automatically. Delete the post, fix the file so it no longer declares AI, and post the corrected version. If the image really was made with AI, the label is correct, and keeping it is the honest choice.
Why is the label on the post for some photos and in the menu for others?
Since September 2024, Meta has split the two cases. Content it detects was fully generated by AI keeps the visible label on the post. Content it detects was only modified or edited with AI gets the note tucked into the post’s menu instead, under Options. The split is why a lightly retouched photo can seem unlabelled at first glance.
Do I have to label AI images on Instagram?
No, not for still images. Meta requires the label for photorealistic video or realistic sounding audio that was digitally created or altered, and says there may be penalties if you skip it. Images are not required, but Instagram may still label them itself when it detects AI signals in the file. Using the toggle anyway is the safer choice.
How do I check if my photo will get the label before I post?
Check the exact file you plan to upload, not the original. Drop it into the official C2PA verify tool and read the actions it lists. A c2pa.created action with trainedAlgorithmicMedia means a label is likely, while opened or colour adjustments with no AI declaration mean the file is clean. Re-export without the generative step if the declaration is wrong.







