Check before you post
AI label checker: will Instagram label this photo?
The checker reads a file's C2PA, IPTC, XMP and EXIF data without changing it, then rates the label risk on Instagram, TikTok and Facebook.
Drop files here or paste an image. Photo on your phone? Open scrubphoto.com there.
- No upload. Works offline once the page loads.
- Read only. Your file is not changed.
- Free. No sign-up, no limits.
No photo handy?
Pick a photo or video and the checker shows what Instagram, TikTok and Facebook can read inside it before you post. It lists the AI markers, camera data and location flags, gives a verdict per platform, and leaves the file untouched. The Scrub this file button hands it to the home tool if something needs to go.
Will Instagram label this photo? The three verdicts
Each platform tag in the result shows one of three verdicts:
- Will likely be labeled: the file carries C2PA Content Credentials, an IPTC digital source type saying AI created or edited it (or that it is a composite with AI parts), or an XMP generator string from a known AI tool.
- Unlikely: the only hint is an AI editor named in the EXIF Software field, with no C2PA. Meta names C2PA and IPTC as its signals, so a software name alone is weak.
- No metadata trigger found: nothing in the file matches what the platforms say they read.
Tap a verdict for its reason and source. On Instagram a likely verdict points to the AI info label (also worded as AI content, and called Made with AI when it launched in 2024). Meta said in September 2024 that photos only edited with AI show the label in the post menu, not on the post. Platform rules: Instagram, Facebook, Threads, TikTok.
Why "no metadata trigger found" never means no label
The checker reads metadata, and metadata is only one way a platform decides. Meta AI images carry an invisible watermark as well as IPTC data, and Google DeepMind's SynthID sits in the pixels, so neither appears in any metadata readout. Platforms can also use their own classifiers, and anyone can add the label by disclosing AI use.
Video works differently on Meta. Meta said it cannot yet read these signals in most third-party video and audio, so it relies on disclosure, which it requires for photorealistic video and realistic audio that was digitally created or altered. That is why Meta video verdicts are marked as disclosure-based. A clean readout means the file carries no known trigger, nothing more. The limits page covers the rest.
What each line of the readout means
C2PA and who signed it
Content Credentials are a signed record of where a file came from and how it was edited. The checker shows whether a C2PA manifest is present and names the app or device that wrote it, for example Adobe Photoshop, OpenAI or Samsung. It does not validate the signature. The C2PA remover page explains who writes them and why.
IPTC digital source type values
This field states how the media was made, using a fixed list of terms the IPTC publishes. The checker translates these values:
- trainedAlgorithmicMedia: created by AI. The IPTC names it "Created using Generative AI".
- compositeWithTrainedAlgorithmicMedia: edited with AI, for example inpainting or outpainting on a real photo.
- compositeSynthetic: a mix of several elements where at least one is generative AI.
- algorithmicMedia: made purely by an algorithm with no training data, like a pattern drawn from a formula.
- digitalCapture: recorded from real life with a digital camera or recorder.
The first three count as AI triggers. The last two do not.
XMP creator tool, EXIF camera and software
XMP CreatorTool names the app that saved the file. EXIF adds the camera, software and capture date. Together they trace a surprise label to its app; see Lightroom, Photoshop and Canva.
GPS, PNG text, trailing data and video boxes
Location shows as "GPS: present", never as coordinates. PNG text keys can reveal prompts saved by Stable Diffusion WebUI or ComfyUI. Trailing data is anything stored after the image, like a motion-photo clip. For video, the checker lists which metadata boxes exist.
Questions about checking photos for an AI label
Will Instagram label my photo as AI?
If the Instagram tag says "will likely be labeled", the file carries C2PA or IPTC data Meta says it reads. If it says "no metadata trigger found", the file holds no such marker, though disclosure, watermarks and classifiers can still lead to a label.
Why does my Instagram draft say AI info?
Meta labels a post when the file carries AI indicators or when you disclose AI use. Check the photo here and make sure you did not turn on an AI disclosure yourself. If the file is the cause, scrub it and start a fresh draft. The AI info guide goes further.
How do I check if a photo has C2PA?
Choose the photo above. The first line of the readout says whether a C2PA manifest is present and names the app that signed it.
Is this an AI image detector?
No. It reads metadata and cannot judge the pixels. A stripped AI image shows no trigger, and a real photo exported with Content Credentials can show one.
What does trainedAlgorithmicMedia mean?
It is the IPTC digital source type for media created with generative AI. Meta lists IPTC metadata among the indicators it reads, so this value points to a likely label on Instagram, Facebook and Threads.
Can I check a video before posting it to TikTok?
Yes. Choose an MP4, MOV or M4V file. TikTok said in May 2024 that it labels images and videos carrying Content Credentials, so a C2PA box there means a likely TikTok label. The video guide covers Reels too.
What happens when I tap Scrub this file?
The file moves to the home tool, already inspected. Tap the Scrub button there, then save the clean copy.
Sources
Official pages this page was checked against in October 2026: