X Tests AI Labels That Explain Why a Post Was Flagged
X is testing expanded AI transparency labels that show not just whether AI was used, but how and why. Creators and social teams need to prepare for new detection and refute workflows.
What happened
X is developing new AI transparency labels for posts. According to a Social Media Today report, the labels will not only indicate whether AI was used to generate an update, but also how it was applied in each instance.
Michael O’Herlihy, X’s head of brand safety, shared example screens during a presentation at the Brand & AI Safety Summit in New York on Tuesday. The examples were posted on X by Adweek senior tech reporter Kendra Barnett. The new labels include more detailed information on why a post was flagged for AI content, giving creators more opportunities to refute the labels by understanding the rationale behind X’s automated enforcement system.
The company hopes expanded notes will prompt more creators to provide feedback on the specifics, which can then be used to train the detection system. X also said it has increased its reliance on automated tagging of AI violations because it is much faster, and the automation should ensure that more harmful content is refuted in-stream.
Key facts
- X is testing an automated way to detect and tag posts that may have been generated using artificial intelligence.
- The new AI content labels will show not only whether AI was used, but also how it was applied in each post.
- Labels will include more detailed information on why a post was flagged, helping creators understand and refute potential concerns.
- X hopes creator feedback on these labels will refine its AI detection process and train the system.
- O’Herlihy said X increased its reliance on automated tagging because it is much faster, with the goal of refuting more harmful content in-stream.
Our analysis
This moves X beyond a simple binary “AI or not” label toward an explanatory enforcement system. For creators and social media managers, that changes the risk calculation. A post that uses AI for a minor edit may be flagged differently from fully generated text, but the source text suggests the label detail will make that distinction visible. This is likely intended to reduce confusion and give users context before they re-share a flagged post.
The most important shift is the feedback loop. X is explicitly hoping that creators will respond to the labels and provide specifics, which will then train the automated system. That suggests creators who actively refute false flags could have a direct influence on how X’s detection improves. The downside is equally clear: false positives could annoy creators if their original or lightly edited content is unfairly tagged. The source text notes this as a potential problem, and for brand accounts it could create moderation friction if automated labels appear on high-volume campaign content.
What it means for operators
- Audit your AI-assisted content workflow now. Document where AI is used in ideation, drafting, editing, or image generation so you can respond quickly if a post is flagged and you need to explain the process.
- Monitor flags on brand and creator accounts closely. Do not ignore an AI label. Check the detailed rationale X provides and decide whether to refute, especially if the content is original or only lightly edited.
- Use the refute path as a training signal. Provide specific, factual feedback when a label is wrong. X says this feedback will train the system, so consistent, clear refutes may help reduce false positives for your content type.
- Prepare internal guidance for teams. Set a policy on when to label AI use voluntarily and how to respond to automated flags, because users may be less likely to re-share a post that has been flagged for AI content.
Source:X experiments with expanded AI transparency labels — Social Media Today(2026-11-11)
Editor's note: prepared by the NoobClaw newsroom with AI assistance from the public report above. Facts are as reported by the source; the analysis is our view. Spotted an error? Contact us and we will correct it.