Open Instagram or TikTok today and you will see small grey labels — "AI info", "AI-generated" — sitting on images that look, to the eye, like ordinary photographs. The platforms are not guessing from the pixels alone: in most cases they are reading hidden data written into the file by the tool that created it, and deciding the label before your post ever reaches a feed. That is why checking a file before you post it matters. Our free AI Label Checker reads exactly that hidden data — C2PA Content Credentials, IPTC and XMP fields, EXIF, PNG text chunks and video metadata boxes — and tells you, platform by platform, what Instagram, TikTok, YouTube and Pinterest are likely to do with the file. This guide explains how to check if an image is AI-generated with it, what the markers actually mean, the three layers of detection the platforms use, and what the "AI label remover" tools in the search results will not tell you.
Step one: drop in the file. Any photo or short video you are about to post — JPG, PNG, WebP, MP4, MOV or WebM. The file is opened by your browser only: there is no upload, no account and no history, and the scan takes about a second. Step two: read the findings. The checker parses the same metadata containers the platforms read. A C2PA Content Credentials manifest is a signed provenance record — the strongest signal there is. The IPTC DigitalSourceType field is the plain-language one: the value trainedAlgorithmicMedia literally means "created with AI". The XMP CreatorTool field names the generating tool, and EXIF, PNG text chunks and video metadata boxes can carry the same story in older or simpler files. Step three: read the per-platform verdict. You see exactly which markers are present and what Meta, TikTok, YouTube and Pinterest are therefore likely to do with the post — before it goes live, while you can still change your plan.
The markers are not a secret code — they are open standards the platforms themselves announced they would read. Meta announced in February 2024 that it reads C2PA and IPTC indicators and applies its "AI info" label; TikTok began reading Content Credentials in May 2024 and labels matching uploads "AI-generated"; YouTube auto-labels C2PA-carrying uploads, and labels arising from C2PA content cannot be adjusted in YouTube Studio; Pinterest reads the IPTC field and labels matching posts "AI modified". So a file carrying a C2PA manifest and an IPTC trainedAlgorithmicMedia value is not a borderline case — it is a file every major platform is set up to label. The worked example on the tool page makes this concrete: a creator generates an image with Google Gemini and checks it before posting. The checker finds three things — a C2PA manifest, the IPTC DigitalSourceType reading trainedAlgorithmicMedia, and an XMP CreatorTool naming the generating tool. The verdicts follow directly: Meta very likely labelled, TikTok very likely labelled "AI-generated", YouTube label locked in, Pinterest very likely labelled "AI modified". One check, four answers, no surprises after posting.
Platforms decide on a label in up to three ways, and this checker reads the first one. The first layer is file metadata — the structured data above, written into the file by the tool that created it. The second layer is invisible watermarking: signals such as Google's SynthID are woven into the pixels or audio itself and survive cropping, filters and recompression. They are not stored in metadata, so no metadata reader — including this one — can see them; even Google routes detection through its own detector. The third layer is the platform's own AI classifiers, which judge the content directly; Pinterest states openly that its classifiers detect generative AI "even if the content doesn't have obvious markers". A complete picture needs all three layers, and this tool gives you the first one, honestly labelled as such. That honesty is the point: "No markers found" means no metadata trigger — it is not a guarantee of no label. A clean metadata result on a file whose pixels carry a SynthID watermark, or whose content a classifier recognises as generated, can still be labelled. Check the metadata layer, then use your judgement about the other two.
Run the same check on a straight iPhone photo and the story flips. The checker finds camera make and model, a date — and GPS coordinates. No AI markers at all, so no metadata-driven label is likely on any platform. But the GPS finding matters for a different reason: posting that file would publish the photographer's home location to anyone who downloads it. Two files, two completely different risks — both visible in seconds, before anything is posted. That GPS case is exactly what the Photo Privacy Cleaner exists to remove: it strips location and device metadata from photos before you share them. The two tools are a pair — one tells you what a platform will say about your file, the other removes the details you never meant to share. And if the question is about words rather than pictures, our AI Tell Highlighter does the equivalent job for text: it flags the 70+ patterns that make writing read as AI-generated, so you can fix a draft before it goes out.
Search results for this topic are full of tools promising to strip markers so posts pass unlabelled. Three facts from the platforms themselves are worth knowing before trusting any of them. First, TikTok states that an applied label "cannot be edited or removed after posting", and its Community Guidelines prohibit "automation tools, scripts, or other tricks designed to bypass our systems", with removal or bans as consequences. Second, YouTube documents that labels based on C2PA metadata are among the cases that "can not be adjusted" — deleting and re-uploading the same pixels changes nothing once its classifiers have seen the content. Third, metadata is only one of three detection layers: a cleaned file can still be labelled by SynthID-style watermark detection or by a platform classifier judging the content itself. Stripping metadata does not make a generated image real, and it can cost you the account you were trying to protect. This tool is built on the opposite principle: it never removes markers or helps hide AI content. It exists for transparency — so you know what a platform will say about your file before you post it, and can label it yourself, choose a different image, or post it openly as AI-made.
Treating "no markers found" as a guarantee. It means no metadata trigger — SynthID watermarks and platform classifiers operate on a different layer and can still label the file. Checking after posting instead of before. Once a label is applied, TikTok says it cannot be edited or removed, and YouTube locks C2PA-based labels — the check is only useful while you can still change your plan. Stripping metadata to dodge a label. It does not work reliably (two detection layers remain), it removes useful provenance from your own work, and the platforms treat bypass tricks as a policy violation. Ignoring the GPS in a "clean" photo. A file with no AI markers can still carry your home location in its EXIF — check the privacy findings, not just the AI ones. Assuming one platform's rules apply everywhere. Meta, TikTok, YouTube and Pinterest read different combinations of markers and label differently; the per-platform verdict exists because one answer does not fit all four.
The honest limits are three. First, the checker reads file metadata only — it cannot detect invisible pixel watermarks such as Google SynthID, and it cannot predict a platform's own AI classifiers. Second, the verdicts are likelihoods drawn from what the platforms have publicly documented, not promises: platforms change their detection without notice. Third, the tool changes nothing about the file — it reports, it never removes markers, and it cannot change what a platform does once you post. Within those limits, it does one job completely: in about a second, on your own device, it shows you the hidden data inside a photo or video and what the four biggest platforms are likely to do with it. Pair it with the privacy side of the same question. The companion guides are our AI-written text guide, for the same "check before you publish" discipline applied to words, our PDF metadata guide, for the hidden fields inside the files you share every day, and our data broker opt-out guide, for removing your details from the databases that sell them.
The same check-before-you-publish discipline, applied to words instead of pictures.
The hidden fields inside the files you share every day — see them, fix them, strip them.
Remove your details from the databases that sell them — direct opt-out links for 15 brokers.