AI Labels by Platform: What Triggers Them, Who Reads What, and What Changed
Updated October 30, 2026 (checked monthly)
Every major platform now labels AI content, but each one decides differently. Meta and LinkedIn read signals stored inside the file. TikTok, Pinterest and YouTube also run their own detectors on the content itself. This page keeps the current rule for each platform, quoted from the platform's own help pages and newsrooms, with a dated changelog at the bottom. Where a platform has published nothing, it says so instead of guessing.
Quick answer
The short version:
- 1Meta (Instagram, Facebook, Threads): "AI info" label from C2PA/IPTC signals or your own disclosure. No published pixel detector.
- 2TikTok: reads C2PA and runs its own detection models. Labels "AI-generated".
- 3YouTube: you must disclose realistic AI content; since May 2026 it also labels automatically from C2PA and its own detection.
- 4Pinterest: "AI modified" from IPTC metadata and classifiers, so it fires with no metadata at all.
- 5LinkedIn: a C2PA icon, only when a signed credential is in the file.
The table
| Platform | Label | What triggers it | Labels with no metadata? | Disclosure rule / penalty |
|---|---|---|---|---|
| Instagram / Facebook / Threads | "AI info" (was "Made with AI" until Jul 1, 2024); "Imagined with AI" on Meta AI images | C2PA and IPTC signals, invisible watermarks, or the poster's disclosure. AI-edited content: label in the post menu only | Not stated | May be required for photorealistic video and realistic audio, not images; "there may be penalties". Reach limits for unlabeled AI-person profiles (Aug 31, 2026) |
| TikTok | "AI-generated" | Content Credentials (C2PA) read since May 2024; TikTok's own detection models; creator toggle | Yes ("our own detection models") | Realistic AI content must be labeled; TikTok says harmfully misleading AI content is removed whether labeled or not (May 2024 post) |
| YouTube | AI label below the player (moved May 2026); overlay on Shorts | Creator disclosure; auto-label from YouTube AI tools, C2PA metadata, or internal detection (May 2026) | Yes ("internal systems detect") | Removal and Partner Program suspension for undisclosed realistic content |
| "AI modified" (Apr 30, 2025), bottom left of the Pin | Metadata checked against the IPTC standard; classifiers; owner indication | Yes (classifiers) | None stated; appeal form available | |
| C2PA icon on signed image and video | A C2PA credential in the file. Nothing else | No | None stated | |
| X | Not verifiable: help pages could not be reached | Not verifiable | Unknown | Unknown |
| Snapchat | Ghost-and-sparkle watermark on exported Snap AI images; context cards in app | Snap's own AI tools | No | Not verifiable |
| No post-level AI label (bot/app account labels announced Mar 2026) | n/a | No | Subreddit rules | |
| Discord | None | n/a | No | AI media covered only within sexual-content rules |
How to read the table
Meta: Instagram, Facebook and Threads
Meta announced in February 2024 that it would label images made with AI across all three apps when it detects "industry standard AI image indicators", naming the "AI generated" information in C2PA and the IPTC technical standards, as well as invisible watermarks and metadata embedded in image files (Meta newsroom, Feb 6, 2024). The label was called "Made with AI" until July 1, 2024, when Meta renamed it "AI info" (Meta newsroom, Apr 5, 2024, updated through Oct 2025). Since September 12, 2024, content that was edited with AI rather than generated by it shows the label only in the post menu, not on the post itself.
Meta's help page says signals in the file "are read by Meta's systems to determine if a label is applied" (Meta help). Two things Meta has not published: whether it strips or keeps metadata after upload, and whether it runs any detector on the pixels. Both claims circulate widely online; neither has a source.
Disclosure: Meta says "there may be penalties for content shared without a label when it is required". Meta's help page says that requirement covers photorealistic video and realistic-sounding audio and does not apply to images, though images may still show a label if its systems detect AI. Profiles: on August 31, 2026 Instagram renamed its "AI creator" label "AI-generated profile" and said profiles featuring an AI-generated person without it may have their reach limited and may lose recommendation eligibility; creators who only use AI tools do not need it (Instagram, Aug 31, 2026). Scale: Meta reported over 330 million labeled items on Instagram in one month of 2024 (Meta transparency report).
TikTok
TikTok has required creators to label realistic AI content since 2023 and, in May 2024, became the first video platform to read C2PA Content Credentials on uploads and auto-label them. It also attaches Content Credentials to content made with its own AI tools, which "will remain on content when downloaded" (TikTok newsroom, May 9, 2024). In November 2025 TikTok said it labels using "the labeling tools we offer creators as well as our own detection models", added a Manage-topics control to see more or less AI content, and began testing invisible watermarks on content made with tools like AI Editor Pro (TikTok newsroom, Nov 19, 2025). TikTok says it had labeled over 1.3 billion videos by then.
YouTube
YouTube requires creators to disclose realistic content that was meaningfully altered or synthetically generated; the label appears in the expanded description or, for sensitive topics, on the player (YouTube Help). Since May 27, 2026, YouTube may add the label itself for content made with YouTube's AI tools, content carrying C2PA metadata, and "content that our internal systems detect is AI generated or altered"; the label moved directly below the player, with an overlay on Shorts (YouTube blog, May 27, 2026). YouTube says the label does not change recommendations or monetization, but failing to disclose can mean removal or suspension from the Partner Program. A separate "captured with a camera" label appears when a file carries an unbroken C2PA 2.1 chain from a supporting camera (YouTube Help).
Pinterest introduced an "AI modified" label on April 30, 2025, shown at the bottom left of a Pin close-up. It is applied when image metadata matches the IPTC standard's generative-AI values, and also by "classifiers that automatically detect Gen AI content, even if the content doesn't have obvious markers" (Pinterest newsroom, Pinterest help). Pinterest does not mention C2PA. Owners can appeal a wrong label through a support form.
LinkedIn shows a C2PA icon on image and video content "that is cryptographically signed using C2PA Content Credentials"; tapping it shows the app or device, the issuer, the date and any AI-use statement (LinkedIn Help). There is no self-disclosure label and no published detector: no credential, no icon.
X, Snapchat, Reddit, Discord
X: its help pages could not be reached while checking this page, so the current label wording and any 2026 rule changes are unverified here. Snapchat: images made with Snap's AI tools carry a small ghost-and-sparkle watermark when saved or exported, and in-app context cards (Snap, Apr 16, 2024); Snap publishes nothing about labeling images made elsewhere. Reddit: no post-level AI label; account-level labels for bots and apps were announced in March 2026, and rules are set per subreddit. Discord: no AI label; its guidelines (effective Sep 29, 2025) mention synthetic media only inside the sexual-content rules (Discord guidelines).
The three kinds of signal
- In the file: C2PA Content Credentials (a signed manifest in a JPEG APP11 segment, PNG caBX chunk or MP4/MOV uuid box), IPTC
DigitalSourceTypevalues such astrainedAlgorithmicMedia("created using generative AI") andcompositeWithTrainedAlgorithmicMedia("edited using generative AI"), and XMP tool signatures (IPTC). These can be removed; see how. - In the pixels: Google's SynthID, "embedded in the pixels of an image" and in Veo video and Lyria audio, which Google says survives when metadata is removed or lost (Google DeepMind). TikTok is testing its own invisible watermark. No metadata tool affects these.
- In the platform: detection models run on the content itself. TikTok, Pinterest and YouTube say they use them. Meta does not say.
The law behind the labels
The EU AI Act's Article 50 applies from August 2, 2026: outputs of AI systems must be "marked in a machine-readable format and detectable as artificially generated or manipulated", and deepfakes must be disclosed (Article 50). Generative systems already on the market before that date have until December 2, 2026 for the marking duty; the disclosure duties were not delayed (Goodwin, Aug 2026). California's AI Transparency Act (SB 942) took effect the same day after AB 853 moved its start date, with obligations for large platforms following on January 1, 2027. There is no general US federal labeling law. Adobe's generative AI guidelines separately prohibit removing Content Credentials from Adobe generative output (Adobe, May 2026). None of this is legal advice; if you publish AI content to an audience in these places, disclosure is your responsibility whatever the file carries.
How we handle this at StripShot
StripShot removes the first kind of signal only: the C2PA manifest, IPTC and XMP fields, and the tool and software metadata AI apps write into images and video, at the binary level with no quality loss and nothing uploaded. It shows you exactly what it found first, and it is honest about the rest: it cannot touch SynthID or a platform's own detector, and a label that is legally required stays your call.
Changelog
| Date | Change |
|---|---|
| 2026-10-30 | Page published; every row checked against the platform's own page on this date. |
| 2026-08-31 | Instagram renames its "AI creator" label "AI-generated profile" and says unlabeled profiles featuring an AI-generated person may have their reach limited. |
| 2026-08-02 | EU AI Act Article 50 transparency obligations and California's AI Transparency Act (SB 942, as amended by AB 853) take effect. |
| 2026-05-27 | YouTube starts auto-applying AI labels from its own tools, C2PA metadata and internal detection; label moves below the player. |
| 2025-11-19 | TikTok announces a Manage-topics control for AI content and tests invisible watermarks on content made with its AI tools. |
| 2025-04-30 | Pinterest launches the "AI modified" label (IPTC metadata plus classifiers). |
| 2024-10 | YouTube's "captured with a camera" C2PA label (requires C2PA 2.1 or higher). |
| 2024-09-12 | Meta moves the label for AI-edited (not generated) content into the post menu. |
| 2024-07-01 | Meta renames "Made with AI" to "AI info". |
| 2024-05-09 | TikTok begins reading Content Credentials on uploads and attaching them to its own AI content. |
| 2024-04-16 | Snapchat adds a watermark to exported Snap AI images. |
| 2024-02-06 | Meta announces labeling from C2PA and IPTC signals across Facebook, Instagram and Threads. |
Last checked
See what a platform will see
Check a file for C2PA, IPTC and AI signals, free
Runs in your browser. Nothing is uploaded.
Videos save best in Safari or Chrome.
Need unlimited strips?
Scans are free. Pro is unlimited.
Images, video, audio, documents, PDF and text in one tool. Pro adds unlimited strips, whole-folder batches and zip download with a JSON report.
Frequently asked questions
Which platforms add an AI label automatically?
Instagram, Facebook and Threads (the "AI info" label, from C2PA or IPTC signals in the file or from self-disclosure), TikTok ("AI-generated", from Content Credentials and its own detection models), YouTube (since May 2026, from its own AI tools, C2PA metadata or internal detection), Pinterest ("AI modified", from IPTC metadata and classifiers) and LinkedIn (a C2PA icon when a signed credential is present).
Can a platform label my post as AI if the file has no metadata?
On TikTok, Pinterest and YouTube, yes: all three say they run their own detection on the content itself. Meta describes its label as driven by industry-standard signals in the file plus self-disclosure. It also says images may show a label if its systems detect AI, without saying how. LinkedIn shows its icon only when a C2PA credential is present.
What is the difference between "Made with AI" and "AI info"?
Same label. Meta renamed "Made with AI" to "AI info" on July 1, 2024, after complaints that real photos edited with AI tools were being labeled as if they were fully generated.
Do I have to disclose AI content myself?
YouTube requires disclosure of realistic altered or synthetic content and says penalties include removal and suspension from the Partner Program. Meta may require a label on photorealistic video and realistic audio, not images, and says there may be penalties for omitting it. TikTok requires labeling of realistic AI content. In the EU, Article 50 of the AI Act applies from August 2, 2026.
Does removing metadata guarantee no label?
No. It removes the signals Meta and LinkedIn describe reading (C2PA, IPTC, XMP). It cannot remove Google's SynthID, which is in the pixels, and it does not affect the detection models TikTok, Pinterest and YouTube say they run on the content itself.