TikTok's updated guidelines turn AI labeling from a courtesy into a distribution rule: realistic AI or heavily edited photos without a label risk losing For You reach. The practical line is simple. If the edit changes what happened in the scene, label it. If it only changes how the photo looks, you don't need to.
- ✓From September 24, 2026, TikTok requires labels on AI-generated or substantially edited content showing realistic scenes or people
- ✓Unlabeled content can be removed, auto-labeled, or made ineligible for the For You feed, which is where most reach comes from
- ✓Routine edits such as cropping, exposure, color correction and compression are not 'substantial'; adding, removing or swapping people and objects is
- ✓TikTok reads C2PA Content Credentials to auto-label AI images; stripping metadata removes them, so label manually if you do
TikTok's Community Guidelines update, effective September 24, 2026, requires creators to label AI-generated or substantially edited content that depicts realistic scenes or people. Unlabeled content can be removed, labeled automatically, or made ineligible for the For You feed. For most creators, losing For You eligibility is the real penalty. The post stays on your profile but stops being recommended to people who don't already follow you.
TikTok has asked for AI labels since 2023. What's changed is how clearly the rules connect missing labels to lost distribution, and that "substantially edited" content now sits alongside fully AI-generated content. That second part matters most for photo creators, because it raises the question every Photo Mode carousel now has to answer: is my edit substantial?
The Line: Did the Edit Change What Happened?
TikTok's wording focuses on content that depicts realistic scenes or people. The most useful test is whether a viewer would come away believing something that didn't happen. If your edit changes what the photo shows, label it. If it only changes how the photo looks, you don't need to.
No label needed:
- Cropping, rotating, straightening
- Exposure, contrast, white balance, color correction
- Filters that change mood without adding or removing anything
- Resizing and compression
- Removing a small blemish or a dust spot
Label it:
- Adding or removing people or objects with generative fill or AI erase
- Face swaps, AI face or body reshaping
- Replacing the sky, background or setting with AI-generated scenery
- AI "enhancement" that invents detail, such as upscaling a blurry face into a sharp one
- Anything fully generated that looks like a real photo
Gray zone, label to be safe: heavy AI retouching of skin and features, AI-extended backgrounds that change a photo's aspect ratio, and "restored" old photos where AI filled in missing parts. If you have to think about it, the label costs you nothing. Getting it wrong could cost you For You distribution.
Why Metadata Matters More Than It Used To
TikTok reads C2PA Content Credentials, the metadata standard that tools including Adobe's apps, OpenAI's image generators and a growing number of cameras attach to images, and uses them to label AI content automatically. That's convenient: generate an image with a C2PA-supporting tool, upload it, and the label can appear without you doing anything.
But Content Credentials live in the file's metadata, and many everyday steps remove it: screenshots, some editing apps, messaging apps that recompress images, and privacy tools that strip metadata. That leaves two situations to handle:
- You strip metadata for privacy, for example to remove GPS location. That's reasonable. Just turn on TikTok's AI label yourself for any AI content, because the automatic signal has gone.
- Someone strips metadata to avoid a label. The updated guidelines specifically cover attempts to get around detection, and repeated attempts can lead to restricted posting or a ban. TikTok's detection doesn't rely only on metadata.
Treat metadata removal as a privacy tool, not a way to hide AI. Remove metadata when a photo carries location or device details you don't want public, and label honestly either way.
A Pre-Upload Checklist for Photo Mode
- Go through the carousel frame by frame. One AI-edited slide in a ten-photo carousel makes the whole post AI content for labeling purposes.
- Turn on the AI-generated content label in the post settings if any frame needs it.
- Turn on commercial disclosure if it's sponsored. The update also requires creators to switch on TikTok's commercial content setting before publishing brand promotions. Photo carousels for products count.
- Size the frames properly. Photo Mode displays best at 1080x1920 (9:16) or 1080x1350 (4:5). Resize every frame to the same ratio so the carousel doesn't jump between shapes, and compress each to a few hundred kilobytes so TikTok's own recompression has less to damage.
Four Common Cases, Decided
A wedding photographer posting a carousel of edited portraits. Exposure, color grading and skin smoothing that removes a spot: no label. Using AI to remove a guest from the background of the first kiss: label, because the scene has changed.
A travel creator posting a sunset from Santorini. A graded real sunset: no label. A real photo with the grey sky swapped for an AI sunset: label, because viewers will believe they're seeing what the sky looked like.
A product seller posting a lifestyle shot. A real product photographed on a real table, cropped and color-corrected: no label. The same product placed into a generated kitchen: label, and switch on commercial disclosure, since it's selling something.
A family account restoring a 1970s photo. Scanning, straightening, fixing the color and removing dust: no label. Letting an AI "restore" a damaged face by inventing features that weren't there: label, because the face in the result isn't fully the person in the original.
The pattern holds: the more the edit decides what's in the picture, rather than how it looks, the clearer the need for a label.
What This Means for Photographers
For people posting real photos, the update is mostly good news. Distribution now rewards honesty about AI, and an unedited or lightly edited photo carries no labeling burden at all. With feeds increasingly full of AI imagery, "no label" is starting to mean something.
The risk falls on the middle ground: travel creators who swap in dramatic AI skies, real estate accounts that stage rooms with generated furniture, and product sellers whose photos have AI-extended backgrounds. Those edits are common, often harmless, and now clearly in labeling territory. Label them and keep your reach, or leave them unlabeled and hope the detector misses them. Only the first is a strategy.
Related reading:
- TikTok's Invisible Watermark for AI Content: how TikTok marks AI images it generates
- TikTok Photo Mode 2026 Features and Image Quality: specs for carousel posts
- EU AI Act Article 50: Compression and C2PA: why compression can strip Content Credentials
Frequently asked questions
Do I need to label edited photos on TikTok?
Only if the edit is substantial and the result looks realistic, for example adding or removing people, swapping faces, or generating a background. Standard edits such as cropping, brightness, color correction, filters that don't alter content, and compression don't need an AI label.
What happens if I don't label AI content on TikTok?
Under the September 24, 2026 guidelines, unlabeled realistic AI or substantially edited content can be removed, restricted, automatically labeled by TikTok, or made ineligible for the For You feed. Repeated attempts to evade detection can lead to account bans.
Does removing metadata stop TikTok from detecting AI images?
It removes C2PA Content Credentials, which TikTok uses to auto-label AI content, but TikTok also uses other detection. Deliberately stripping credentials to avoid a label is the kind of evasion the new rules penalize. If you remove metadata for privacy, add the AI label yourself.
Continue reading
Try Optimage — it's free
Compress, convert, and optimize images in seconds. No sign-up, no limits.
Start Optimizing Free