Meta is reportedly developing a native AI Detector tool integrated directly into the Meta AI interface.
This move comes as Meta faces mounting pressure from its Oversight Board to curb the spread of “AI slop” and deepfakes, which spiked significantly following the regional conflicts in early 2026.
The “AI Detector” Menu Feature
Reports from early testers (including TestingCatalog) have highlighted a new “AI Detector” option appearing in the Meta AI menu. While not yet fully operational for all users, the tool’s intended purpose is to provide real-time verification of content.
- Contextual Analysis: The tool is designed to analyze text, images, and videos shared within Meta’s ecosystem to determine if they are synthetically generated.
- 100% AI Confidence: Early screenshots show the tool providing a percentage-based confidence score (e.g., “100% AI”) to help users identify non-human content.
- Cross-App Integration: While appearing first in the standalone Meta AI interface, the technology is expected to power the automated “AI Info” labels across Facebook, Instagram, and WhatsApp.
Why Now? The “Oversight Board” Pressure
The launch of this tool is a direct response to a critical ruling from the Meta Oversight Board on March 10, 2026:
- The “Haifa” Incident: The board overturned a previous Meta decision that failed to label a viral, AI-generated video of “damaged buildings in Haifa” during the June 2025 conflict. The board called Meta’s current labeling “too slow” and “not robust enough” for times of crisis.
- New Labeling Standards: Meta has agreed to implement “High Risk AI” labels starting in May 2026 for content that could materially deceive the public on matters of national importance.
- Provenance Standards: Meta is being forced to more strictly adhere to C2PA standards (digital watermarking), ensuring that content generated by its own tools—including the delayed “Avocado” and “Mango” models—is identifiable across the web.
Broader AI Safety Rollout (March 2026)
The AI Detector is part of a “sweeping” anti-fraud and safety update Meta announced on March 11, 2026:
| Feature | Platform | Function |
| Celeb-Bait Detection | Facebook / IG | AI that identifies “fake fan sentiment” and misleading bios used in celebrity impersonation scams. |
| Suspicious Friend Alerts | Real-time warnings when a friend request comes from an account with high “suspicious signals.” | |
| AI Scam Review | Messenger | Users can now submit recent chat messages for an instant AI review if they suspect a job offer or crypto pitch is fraudulent. |
| Device Linking Warning | An alert that shows exactly where a device linking request is coming from to prevent account takeovers. |
Technical Setbacks
Despite the push for new detection tools, Meta’s generative capabilities are currently lagging. Internal reports suggest that Meta’s next-generation image/video model, “Avocado,” has been delayed until at least July 2026 after failing to beat Google’s Gemini 3.0 in head-to-head testing. This delay has reportedly led Meta to consider licensing rival models in the short term to keep its “Superintelligence” roadmap on track.
