How to remove the AI label from an image on Instagram, Facebook and Threads

Removing a file’s metadata removes the label that metadata triggers. Meta says its systems read industry-standard signals in the file to decide whether to add “AI info”. NoWatermark removes the C2PA, XMP and IPTC signals from your copy of an image and then re-scans it, so you can see the file no longer carries them. That is the one source of an AI label a file cleaner can change. The others are invisible watermarks, visual classifiers and self-disclosure, and none of them lives in file metadata. Cleaning also does nothing for a post that is already labelled: it changes the next file you upload.

Our prediction: high confidence. Meta says the label comes from signals in the file or from your own disclosure. We expect an image cleaned with NoWatermark using Everything removable, with the re-scan showing Content Credentials, XMP and IPTC as removed, not to receive the AI info label those signals trigger on Instagram, Facebook or Threads. One caveat: images made with ChatGPT or Gemini also carry a SynthID watermark, which no file cleaner can remove, and Meta has not said whether it reads it. This is a prediction from Meta’s documentation, not yet our own test result. We are running that test and will publish the result on this page.

Try it: check and clean your image before you post. It runs in your browser, and your file is never uploaded to us.

How to remove the AI label from an image before posting on Instagram

1. Open Remove AI Label Metadata and drop in the exact file you plan to post. It runs in your browser, and the file is never uploaded.
2. Read the scan. The rows Meta’s systems can read are C2PA / Content Credentials, XMP, IPTC and AI generator metadata. Any marked Detected is a signal Meta may use to add AI info.
3. Choose Everything removable and click Clean metadata.
4. Check the report. Each of those rows should read Removed, which means a second, independent scan of the cleaned copy could not find it. SynthID always reads Unable to verify: no file cleaner can check or remove it.
5. Download the cleaned image and post that file. The same file works for Facebook and Threads. A post that is already labelled is not changed; the cleaned file goes up as a new post. For an unedited camera original, skip cleaning so it can still qualify for the camera capture label.

What Meta calls it

Meta’s Help Centre and July 2024 newsroom update say “AI info”. Instagram’s own Help Center page for the same subject now says “AI content label” throughout. This page uses “AI info” and means both. Meta runs one labelling system across Instagram, Facebook and Threads, and its Help Centre page speaks of a “post, story, reel or thread”. Where a fact below comes from an Instagram page only, it says so.

Two ways the label gets on: automatic or self-disclosed

Meta said it adds AI info “when we detect industry standard AI image indicators or when people disclose that they’re uploading AI-generated content.”

• Self-disclosure (Instagram). Instagram’s help says to create a post and toggle “Add AI label”. Stories and reels have their own steps. The page is for the mobile apps and says the feature is not available on computers. It describes the toggle at posting time only.
• Automatic. Meta’s Help Centre says signals in the file are “read by Meta’s systems to determine if a label is applied”. If you choose not to disclose, Instagram’s help says an AI content label may still be applied if Meta detects your content was AI-generated or AI-edited.

The four sources, and which one cleaning changes

1. Metadata read at upload. In February 2024 Meta said it was building tools to read “AI generated” information in the C2PA and IPTC standards, so it could label images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock as those companies implemented their plans. Its current help pages say only “industry-standard signal” and do not name a standard. This is the source cleaning changes. See what are Content Credentials and IPTC DigitalSourceType.
2. Invisible watermarks. Meta says it puts invisible watermarks and metadata on images made with its own Meta AI feature. Whether it reads anyone else’s watermark, such as SynthID, is not something Meta has said in the pages we found. Cleaning cannot change a watermark, and SynthID is always Unable to verify.
3. Classifiers. Meta said in February 2024 that it was “working hard to develop classifiers” for AI content that lacks markers. That was a statement of intent, and its current pages do not confirm classifiers are used for labels. We cannot measure or influence them.
4. Self-disclosure. Your choice at posting time, described above.

Meta itself notes that “not all AI content contains the information needed to identify it,” and that its labelling methodology “is still evolving”.

Created versus modified: where the label appears

Meta describes two placements. Content with signals indicating it was created with AI is labelled AI info. Content with signals indicating it was modified with AI “will not have the AI info label directly on the post, story, reel or thread”, and you can see how it was modified by clicking Options. The change dates from September 2024, when Meta said it would move the label to the post’s menu for content it detects was only modified or edited by AI, and TechCrunch reported it the same day. Instagram’s help adds that information about AI used to create or modify content can be found in the Options menu regardless of the type of signal.

Meta has not said which IPTC or C2PA value decides between “created” and “modified”. IPTC defines trainedAlgorithmicMedia as “Created using Generative AI” and compositeWithTrainedAlgorithmicMedia as “Edited using Generative AI”, but no source connects Meta’s two placements to those values. Our scanner shows the XMP DigitalSourceType as “Created by a generative AI model” or “Composite including AI-generated elements”, but it does not read the value inside a C2PA manifest, so it cannot tell you which placement you would get.

If a real photo of yours carries the label, start with why is my photo labelled AI. Meta said in July 2024 that minor modifications using AI, such as retouching tools, had been labelled “Made with AI”.

What removing metadata changes, verified

Use the Remove AI Label Metadata tool. It runs on your device, removes what you choose, and scans the cleaned file again. A signal is marked removed only when that second scan cannot find it. Image data is copied byte for byte, not re-encoded. It finds C2PA (with manifest size and claim generator), XMP, IPTC and AI generator metadata.

• Everything removable: EXIF, IPTC, XMP, Content Credentials and embedded text. One EXIF rotation field is kept by default.
• Privacy-safe: location and device data. Provenance is kept.
• Provenance-light: Content Credentials, XMP and embedded text. EXIF and IPTC are kept.

Check the report for anything listed as remaining. The preset matters for one reason in particular. Instagram has a positive “Captured with a camera” label, based on secure metadata that certain cameras attach at capture, and says it will not add the label if that information is missing. Stripping Content Credentials from an unedited camera original can lose it. So can any other cleaning. Under the C2PA specification the credential is bound to the file by a hash that covers everything except the manifest itself and padding, so even Privacy-safe, which keeps the manifest and removes EXIF and IPTC, leaves a credential that no longer matches the file, and Instagram says it will not add the label when that information is “unverified”. To keep the camera label, upload the untouched original. We have not tested this on Instagram. Instagram also says a photo created or edited with AI will not get the camera label and may get an AI label instead.

What cleaning cannot change

• ChatGPT images now carry two signals. OpenAI’s help says supported images from ChatGPT, Codex and its API include C2PA metadata and SynthID watermarks. It also says that if metadata is removed, an embedded watermark may still provide a signal. Google’s May 2026 post says OpenAI is bringing SynthID to its content. Cleaning changes the C2PA half of that, confirmed by re-scan. It leaves the watermark, and we cannot tell you whether Meta reads it. See can you remove SynthID.
• Watermarks, classifiers, self-disclosure and live posts, as above.

Meta’s disclosure rules

Instagram’s help says: “Meta does not require you to label images that have been created or altered with AI. Images will still receive a label if Meta’s systems detect they were AI-generated.” Meta’s Help Centre says images may still be labelled if the systems detect they were AI-generated or modified using AI.

For video and audio, the two pages differ. Instagram’s help says “Meta requires you to label content you share that has photorealistic video or realistic-sounding audio that has been digitally generated or altered, including with AI.” The Meta Help Centre says Meta “may require an AI label” for that content and adds that “this requirement does not apply to images.” Both pages note there may be penalties for content shared without a required label; none is named. We do not process video, so this page covers images only.

If the post is already labelled

• Cleaning the file does not change a live post. A cleaned file uploaded again is a new upload, and we have not tested how Instagram treats one.
• Editing your own label on an existing post is not documented in any official page we could open. Instagram’s help describes the toggle when creating a post, a story or a reel. Walkthroughs elsewhere are not sources we accept, so we give no steps.
• We did not find an appeal or review process for a label Meta applied on Meta’s pages.
• Adobe’s FAQ says that for content showing an unexpected label you should visit the platform’s own support resources.
• Meta AI downloads stay labelled. Instagram’s help says content created with Meta AI tools, downloaded and then uploaded also receives an AI content label.

Other platforms have their own rules: see X, LinkedIn and Instagram and EXIF for what Instagram does with ordinary metadata.

Try it on your own image

Drop your image into Remove AI Label Metadata. It lists every signal the file carries, removes the ones you choose, and scans the cleaned copy again so you can see what is gone and what remains, including anything it can only mark Unable to verify. It all runs in your browser: your file never leaves your device. For an unedited camera original, skip cleaning and upload the original, so it can still qualify for the camera capture label.