How to Tell If an Image Is AI Generated
If you want to know how to tell if an image is AI generated, start by looking for visible labels or watermarks. Under the EU AI Act, providers of generative AI systems must mark synthetic images in a machine-readable format so they can be detected as artificially generated. This means many AI images now carry embedded metadata or invisible watermarks that tools can read. You can also check for inconsistencies like unnatural hands, distorted backgrounds, or odd lighting, but these are not reliable on their own.
Check for EU AI Act Labels and Metadata
The EU AI Act’s Article 50 transparency obligations require providers to mark AI-generated images so they can be detected. This marking must be effective, interoperable, robust and reliable as far as technically feasible. The EU provides voluntary icons for labelling: one for fully AI-generated content, one for partially AI-modified content, and one for general AI involvement. If an image carries one of these icons or similar clear disclosure, it is likely AI-generated.
Metadata is another clue. C2PA, also known as Content Credentials, is an open standard that binds cryptographically signed claims about the origin and edits of a file. If an image has a C2PA manifest, you can verify its provenance. However, metadata can be stripped when images are screenshotted or uploaded to platforms that remove it. This is why combining metadata with invisible watermarks is recommended.
Use Free Tools to Check for AI-Generated Images
Free tools can help you test if an image is AI generated. Many of these tools look for invisible watermarks or analyse the image for patterns typical of generative AI. For example, you can use the public validator at laitelier.com/validator to check for C2PA manifests or watermarks in an image. Other free tools may detect inconsistencies or artefacts, but they are not always accurate.
The EU’s Code of Practice on Transparency of AI-Generated Content encourages a multi-layered approach to marking. This includes metadata, invisible watermarks, and clear labels. Signatories to the Code can rely on it to demonstrate compliance with the EU AI Act’s marking duties, so images from these providers are more likely to be properly labelled.
How to Check If a Photo Is AI Generated: Practical Steps
To check if a photo is AI generated, start with a visual inspection. Look for unnatural details like blurry edges, distorted faces, or unusual textures in hair, fabric, or backgrounds. These are common AI artefacts, but they are not foolproof. Next, check for labels or icons indicating AI involvement. If the image is from a platform or provider that follows the EU AI Act, it should be clearly marked.
If there are no visible labels, use a tool to check for embedded metadata or watermarks. Upload the image to a validator that supports C2PA or other provenance standards. If the tool detects a manifest or watermark, the image is likely AI-generated. If no metadata is found, the image may still be AI-generated but not properly labelled.
What the EU AI Act Requires for AI Image Disclosure
The EU AI Act’s Article 50 transparency obligations apply from 2 August 2026. Providers of generative AI systems must mark synthetic images in a machine-readable format so they can be detected as artificially generated. This applies to all AI-generated images, except those used in business-to-business or industrial contexts. Artistic, creative, satirical or fictional works only need disclosure in a way that does not hamper the display or enjoyment of the work.
Deployers must also disclose deepfakes and AI-generated or manipulated text published to inform the public on matters of public interest. The Digital Omnibus on AI gives generative AI systems placed on the market before 2 August 2026 until 2 December 2026 to meet the marking duty. Content generated and published before 2 August 2026 does not need to be labelled retroactively.
How to Determine If a Picture Is AI Generated: Common Pitfalls
One common pitfall is relying solely on visual inspection. AI models are improving rapidly, and many images now look indistinguishable from human-made ones. Another mistake is assuming that the absence of a label means an image is not AI-generated. Some providers may not yet comply with the EU AI Act, or metadata may have been stripped during sharing or editing.
To avoid these pitfalls, use a combination of methods: check for labels, use free tools to detect watermarks or metadata, and look for inconsistencies. If you are a creator or business, consider certifying your files with a tamper-evident provenance system to ensure transparency and compliance.
What Tools and Standards Are Available Now
C2PA is the leading open standard for content provenance. Its manifests, known as Content Credentials, are cryptographically signed and bind claims about the origin and edits of a file. Many platforms and tools now support C2PA, making it easier to verify the authenticity of images. The EU’s Code of Practice on Transparency of AI-Generated Content also recommends using C2PA and invisible watermarks for marking AI-generated content.
For creators and studios, tools like L’AItelier DNA can certify files and issue tamper-evident provenance certificates. These certificates can include C2PA manifests for images and video, as well as invisible watermarks for additional security. Anyone can check a file’s provenance using the public validator at laitelier.com/validator.
Frequently asked questions
How can I check if an image is AI generated for free?
Use free tools like the public validator at laitelier.com/validator to check for C2PA manifests or invisible watermarks. You can also look for EU voluntary icons or other labels indicating AI involvement. Visual inspection alone is not reliable.
Does the EU AI Act require all AI images to be labelled?
Yes, under Article 50, providers of generative AI systems must mark synthetic images in a machine-readable format so they can be detected. There is a narrow exemption for business-to-business or industrial contexts, and artistic works only need disclosure that does not hamper their display.
What is the best way to detect AI-generated images?
The most reliable way is to check for embedded metadata or watermarks using a validator that supports standards like C2PA. Combine this with visual inspection for inconsistencies, but be aware that AI models are improving and may not always show obvious artefacts.
Are there any legal requirements for AI image disclosure in the EU?
Yes, the EU AI Act’s Article 50 requires providers to disclose AI-generated images and mark them in a machine-readable format. Deployers must also disclose deepfakes and certain AI-generated text. These rules apply from 2 August 2026, with a grace period for some systems.
Can AI-generated images be detected without metadata?
It is possible but less reliable. Some tools can analyse images for patterns typical of generative AI, but these methods are not as accurate as checking for embedded metadata or watermarks. Metadata can also be stripped, so combining multiple methods is recommended.