How to prove art is not AI: start with provenance
If you want people to believe your art is human-made, you need evidence they can check. The simplest way is to create a clear, tamper-proof record of how the work was produced. That means saving every stage of the process: sketches, layers, raw files, time-stamped exports and any behind-the-scenes material. When someone asks how to prove art is not AI, the answer is the same as proving any creative work is authentic—show the chain of custody from idea to finished piece.
Platforms, clients and buyers increasingly expect a provenance trail. Without one, even genuine work can be flagged by detectors that look for statistical patterns rather than intent. A strong provenance file removes doubt by letting anyone verify the work’s origin. If you use software that supports it, embedding a signed provenance manifest directly into the file adds an extra layer of trust.
What counts as proof your art is human-made
Raw files are the gold standard for photographs and digital art. A RAW image contains unprocessed sensor data that is almost impossible to replicate with AI. For illustrations, save layered PSD, AI or AF files that show brush strokes, selections and adjustments. For video, keep project files, timelines and unrendered footage. For text, retain drafts, editor comments and version history in tools like Google Docs or Notion.
Next, document the creation process. Screenshots of your workspace, dated WIP images and even screen recordings can all serve as supporting evidence. If you collaborate, keep emails, contracts and shared drives that prove human involvement. The more granular the record, the harder it is for anyone to claim the work is synthetic.
Use open standards to certify your work
Open standards like C2PA (Content Provenance and Authenticity) let you attach a cryptographically signed manifest to your file. This manifest records who created the work, when, and what tools were used. It is stored as metadata inside the file, so it travels with the content wherever it is shared. If the file is edited later, the manifest updates to reflect the new changes, maintaining a clear audit trail.
C2PA manifests are known publicly as Content Credentials. When someone views your work on a supported platform, they can click a small icon to see the provenance information. The standard is designed to be tamper-evident: any attempt to alter the manifest breaks the cryptographic signature, making the tampering obvious. This is why C2PA is one of the most reliable ways to prove art is not AI.
Why metadata and watermarks are not enough on their own
Metadata can be stripped from a file when it is resaved, uploaded to some platforms, or screenshotted. An invisible watermark embedded in the pixels can survive these changes, but it can also be removed by aggressive compression or cropping. That is why the strongest approach combines both: a signed manifest for the origin and edits, plus an invisible watermark that persists even if the metadata is lost.
If you rely only on a visible watermark, a bad actor can crop or clone it out. If you rely only on metadata, it can disappear during sharing. The most robust systems use a multi-layered approach: signed provenance, invisible watermarking and, where possible, platform-level verification. This is the method recommended by the EU’s Code of Practice on Transparency of AI-Generated Content.
How to respond if your art is flagged as AI
If a platform or client accuses your work of being AI-generated, stay calm and provide your provenance evidence. Start with the raw or layered files, then add your C2PA manifest or other signed certificates. If you have a public profile listing your certified works, link to it. Explain that your files carry both metadata and invisible watermarks, and offer to walk the accuser through the verification process.
If the platform uses an AI detector, ask which tool it is and what threshold triggered the flag. Many detectors are trained on a narrow range of AI outputs and can misclassify human work, especially if it has been heavily edited or processed. Your provenance evidence is often more reliable than a detector’s guess.
What the EU AI Act means for human-made art
The EU AI Act does not require human-made art to be labelled, but it does set rules for AI-generated content. From 2 August 2026, providers of generative AI systems must mark synthetic audio, image, video and text output in a machine-readable way so it can be detected as artificially generated or manipulated. This means platforms and buyers are becoming more aware of provenance, and they may start expecting evidence that work is human-made.
For creators, this is an opportunity. If you can show a clear, verifiable provenance trail, your work stands out as authentic. The EU also provides voluntary icons for labelling AI content, and while these are not mandatory for human-made work, using a ‘human-made’ label alongside your provenance can help reassure buyers.
A quick checklist to prove your art is not AI
Start by saving every version of your work: raw files, layers, drafts and exports. Document the process with screenshots, screen recordings and behind-the-scenes material. Use tools that support C2PA to embed a signed provenance manifest into your files. Where possible, also add an invisible watermark for extra resilience.
When sharing your work, include a link to your provenance evidence or a public validator. If someone questions your work, provide the raw files and explain your process. For high-value or commercial work, consider certifying your files with a service that issues tamper-evident provenance certificates. This gives buyers and platforms a simple way to verify authenticity.
Frequently asked questions
Can I prove my art is not AI without raw files?
Raw files are the strongest evidence, but you can still build a case with layered files, version history, screenshots of your workspace and dated WIP images. A signed provenance manifest or certificate adds weight to your claim, even without raw files.
Do I need to label human-made art under the EU AI Act?
No. The EU AI Act only requires disclosure for AI-generated or manipulated content. Human-made art does not need to be labelled, but providing provenance evidence can help avoid false accusations.
What if a platform’s AI detector flags my human-made art?
Provide your provenance evidence: raw files, layered files, signed manifests and any behind-the-scenes documentation. Detectors can misclassify human work, especially if it has been edited or processed, but verifiable proof overrides a false positive.
Is a visible watermark enough to prove my art is not AI?
No. Visible watermarks can be cropped or cloned out. A stronger approach combines a signed provenance manifest, invisible watermarking and, where possible, platform-level verification.
How do I know if my file has a C2PA manifest?
Open the file in a C2PA-compatible viewer or upload it to a public validator. If a manifest exists, you will see the provenance information, including the creator, date and tools used. Some platforms also display a small icon or badge when C2PA data is present.