Instagram 'Made with AI' Label Explained: What Triggers It in 2026
Updated April 2026
The short version
Instagram's "Made with AI" label is triggered by C2PA Content Credentials: a binary certificate embedded by AI tools. Adobe Firefly, DALL-E/ChatGPT, and Microsoft Copilot all embed it. Midjourney, Stable Diffusion, and Flux don't. Removing the C2PA data before uploading prevents the label. Instagram may also use visual detection for non-C2PA images.
When Instagram launched its AI content labeling system in early 2024, it said it would use "industry-standard signals" to identify AI content. That standard is C2PA. Understanding exactly how it works tells you exactly how to prevent it.
This guide covers the full mechanism: what C2PA is, which tools embed it, how Instagram reads it, what else Instagram uses, and what actually works to prevent the label.
What C2PA is and why Instagram uses it
C2PA stands for Coalition for Content Provenance and Authenticity. It's an open technical standard co-developed by Adobe, Microsoft, BBC, Intel, Arm, and others. The goal was to create a verifiable chain of custody for digital content: a signed certificate that travels with an image and records how it was created, who created it, and what tools processed it.
When Adobe, Microsoft, and OpenAI built it into their AI tools, they effectively created a self-reporting system. Every image from Adobe Firefly, DALL-E, and Copilot Designer ships with a certificate saying "I was generated by AI." Instagram reads that certificate.
C2PA certificate structure (simplified)
Which AI tools trigger the label
The label is almost always triggered by C2PA. Whether your AI tool embeds C2PA determines whether you'll see the label automatically.
Embed C2PA (will trigger label)
- Adobe FireflyFounding C2PA member
- Photoshop Generative FillVia Adobe
- DALL-E 3 / ChatGPTOpenAI joined C2PA 2023
- Microsoft Copilot DesignerVia DALL-E
- Bing Image CreatorSame Copilot pipeline
- Adobe ExpressAll Adobe AI
No C2PA (visual detection only)
- MidjourneyXMP signatures, no C2PA cert
- Stable DiffusionNo C2PA by default
- FluxNo C2PA by default
- Leonardo AIXMP/EXIF signatures
- RunwayVideo-focused, no C2PA image cert
- KlingNo C2PA image cert
StripShot detects and removes both C2PA certificates and XMP AI signatures. It covers all tools in both columns.
How Instagram reads C2PA when you upload
When you upload an image to Instagram, Meta's servers parse the file at the binary level before storing it. The parser specifically looks for:
- JPEG APP11 segment (0xEB marker) containing a JUMBF box with a C2PA generator assertion
- PNG caBX, caMs, or caSt chunks containing C2PA credential data
- XMP metadata fields identifying known AI tools in the creator or software fields
- Hash matches against a perceptual hash database of known AI-generated images
The C2PA check is deterministic: if the byte sequence is present, the label is applied. The hash and XMP checks are probabilistic. The visual classifier runs in parallel and is independent of metadata.
The visual AI classifier: Instagram's second detection layer
Even with all metadata stripped, Instagram may label an image if its visual classifier recognizes it as AI-generated. This is a deep learning model trained on large datasets of AI-generated and human-photographed images.
The visual classifier is less reliable than C2PA detection. It produces false positives on some real photos and false negatives on some AI images. Midjourney images are particularly likely to be caught by it because of Midjourney's distinctive visual style.
What you can do about the visual classifier
- Anti-fingerprint mode: Applies micro-transforms and pixel noise to defeat perceptual hash matching. Does not defeat semantic visual analysis.
- Image editing before posting: Adding overlays, adjusting exposure, or compositing with real elements can change the visual signature enough to avoid detection.
- Dispute the label: Use the three-dot menu on the post to report an incorrect label. Manual review, takes days.
Check your image
Drop your image. See exactly what Instagram will read.
JPEG, PNG, WebP · Up to 25MB · 3 free strips/day
Drop images here or browse
Select files· JPEG, PNG, WebP · Up to 20 at once
AI label detection across platforms in 2026
Instagram is not the only platform reading C2PA. This is a cross-platform issue:
Instagram / Facebook (Meta)
C2PA + visual classifier
C2PA + visual classifier
YouTube
C2PA + disclosure requirement
C2PA + visual classifier
TikTok
C2PA + visual classifier
X (Twitter)
C2PA (partial)
Removing C2PA and AI signatures before uploading addresses the metadata layer across all platforms. Visual detection is platform-specific and varies in reliability.
Frequently asked questions
How does Instagram know an image was made with AI?
Instagram uses two detection methods. The primary method reads C2PA Content Credentials: a signed binary certificate that AI tools like Adobe Firefly, DALL-E, and Microsoft Copilot embed directly in image files. The secondary method is a visual AI classifier that analyzes the image pixels for patterns typical of AI generation. The C2PA check is deterministic; the visual check is probabilistic.
Can Instagram detect AI images without metadata?
Yes. Instagram runs visual AI classification on images even if no metadata is present. This classifier looks for visual patterns typical of AI-generated images. It is less reliable than C2PA detection and produces more false positives, but it can label images that have had their metadata stripped.
Does the Made with AI label hurt my reach on Instagram?
Meta has stated the label does not directly reduce reach. It is a disclosure label, not a shadowban trigger. However, audience perception varies. Some creators report lower engagement on labeled posts.
Can I dispute an incorrect Made with AI label?
Yes. Instagram provides a way to report incorrect labels through the three-dot menu on posts. The review process is manual and can take several days. Removing C2PA data before posting is more reliable than disputing after the fact.
Does video trigger the Made with AI label on Instagram?
Yes. AI-generated video from tools like Sora, Runway, Kling, and Pika may trigger the label through metadata or visual classification. Video metadata stripping is more complex than image metadata stripping and requires different handling.
What's the difference between the Made with AI label and an InfoLabel on Instagram?
The 'Made with AI' label appears below the post in the feed. The InfoLabel appears as a small information icon on the post. Both can be triggered by C2PA data. The Made with AI label is the most prominent and most commonly encountered.
Related guides
Remove the Made with AI Label
Adobe Firefly
Firefly Instagram Label
Midjourney
Midjourney Instagram Label
DALL-E
DALL-E ChatGPT Instagram Label
Check your image before posting
See exactly what Instagram reads. Then strip it.
3 free strips to try it. No credit card. Pro is $9/month for unlimited.