July 3, 2024

Meta has flagged one of my images or videos with an ‘AI info’ label, but I didn’t use AI to create it, what do I do?

Don’t panic!

The ‘AI info’ labels are intended to be a neutral signal to viewers and do not affect content distribution or monetization in any way.

This label is for informational purposes only. Meta does not change the content’s distribution based on it and the label is designed to share this AI context neutrally.

 

If you are seeing an image that has “AI Info” included, it is likely because the tool used to create or edit the content is from a company that includes a C2PA and IPTC metadata tag in the file (such as Google, Adobe, Shutterstock, or others). 

It could also be because a person disclosed it as AI content when they uploaded it. 

There is no way to remove the ‘AI info’ label once the metadata tags are detected and the label has been applied. 

However, you can check if a photo you have or are planning to upload to Instagram, Facebook, or Threads contains one of these metadata tags at the links below:

Click here to check if the content contains a C2PA-standard marker of AI usage.

  • Take the image file you want to understand if it has the C2PA marker and upload it to the tool.
  • On the right side, you will see a summary providing information if AI tools were used, what tools and who issues the content credential. 

Click here to check if the content contains an IPTC standard marker of AI usage. 

  • Take the image file you want to understand if it has the IPTC marker and upload to the tool.
  • At the top of the results, tap Full Metadata and scroll down.
  • If a section digitalsourcetype is present and it contains “algorithmicMedia”, then an IPTC AI marker is present. If not, the marker is not present.

When image creation and editing tools from companies like Adobe, Google, OpenAI, Microsoft and Shutterstock use AI as part of the process, they include industry-standard AI signals to the files (e.g. C2PA or PITC metadata). When content is shared on Instagram, Facebook, and Threads, Meta identifies these industry-standard AI signals and uses them to display the Meta AI label on the content. Meta sees the metadata, which signals if AI was used in the creation or editing process – it doesn’t show the level of edits. 

For Adobe specifically, Firefly-powered features in Photoshop will automatically attach Content Credentials to the content. Like a nutrition label for digital content, Content Credentials are tamper-evident metadata that can provide more information about the content, including whether AI was used in the creation or editing process. You can read more about how this works in this blog post by Adobe.

The ‘AI info’ labels are automatically applied based on industry-standard AI signals provided within the file by creative platforms or when people self-disclose AI. Meta is not able to remove an AI label and they do not offer controls on the advertiser side to opt-out of the label being applied.