How to prove you wrote something in 2026

More people than ever are being asked the same question: did you actually write this? Students get flagged by university AI detectors. Freelancers get asked by clients. Job applicants get questioned about writing samples. Journalists face skepticism from editors and readers.

The question used to be rare. Now it is routine. And the answers most people reach for are weaker than they think.

The old methods and why they fall short

When someone challenges your authorship, the first instinct is to show your work. Open Google Docs and point to the version history. Screenshot your drafts folder. Show timestamps. These are better than nothing, but they have real limits:

• Version history is editable. You can create, delete, and manipulate revision history in most document tools. A skeptic knows this.
• Screenshots are not verifiable. Anyone can fabricate a screenshot. There is no way for a third party to independently confirm what you are showing them.
• Timestamps prove when, not how. A file created at 2:00 PM and saved at 4:00 PM does not prove the writing happened during that window. You could have pasted AI-generated text at 3:59.

These methods share a fundamental weakness: they can all be faked, and the person reviewing them has no independent way to verify the claim.

AI detectors are not proof

The next instinct is to run the text through an AI detector. If the detector says it is human-written, problem solved. Except it is not.

AI detectors produce probability estimates based on text patterns. They disagree with each other routinely. The same paragraph might score 12% AI on one tool and 87% on another. Researchers have documented that popular detectors falsely flag non-native English speakers at significantly higher rates.

A detector score is an opinion, not evidence. If the stakes are real — a grade, a contract, a byline — you need something that can be independently verified, not a number from a black box.

Process-based proof: the strongest option

The most reliable way to prove you wrote something is to capture the writing process itself. Not the output. The act of creation.

Process-based verification tools record behavioral signals while you write: how fast you type, where you pause, how you revise, whether you paste large blocks of text, how consistent your patterns are across the session. These signals are then analyzed and packaged into a certificate that anyone can verify.

This approach is fundamentally different from detection. Detection looks at the finished text and guesses. Process verification watches the writing happen and records what it sees. The evidence is behavioral, not statistical.

What good proof of authorship looks like

If you are evaluating methods for proving you wrote something, here is the checklist:

1. Captures behavioral signals during writing. Typing rhythm, revision patterns, pause timing, paste behavior. These are hard to fake because they reflect the cognitive process of composition.
2. Produces a public verification page. The proof should be something a third party can inspect independently, without installing software or trusting your word.
3. Cryptographically signed. The certificate should be tamper-resistant. If someone changes the text after certification, the signature breaks.
4. Cannot be retroactively faked. The behavioral data must be captured in real time. You cannot go back and fabricate a writing session after the fact.

If the method you are using does not meet these criteria, it is better than nothing but weaker than it could be.

How to start proving your work

The practical steps are straightforward:

1. Write in a tool that captures behavioral signals. Humanums records the 6 key signals during your writing session without capturing actual keystroke content.
2. Certify before you submit. When the piece is done, run the certification. The system analyzes your behavioral data and issues a signed certificate.
3. Share the verification link. Every certificate comes with a public verification page. Send the link alongside your deliverable. The reviewer can inspect the evidence themselves.

That is the whole workflow. Write, certify, share. The proof travels with the content.

The short version

Version history and screenshots are weak proof. AI detector scores are not proof at all. The strongest way to prove you wrote something is to capture the writing process with behavioral verification and issue a signed, publicly verifiable certificate.

The badge below this post is what that proof looks like. Click it. It opens the verification page for this article, with the behavioral evidence behind it.

Create a free Humanums account and certify your next piece of writing with real proof of authorship.