Scripli logoScripli
Back to Scripli

What Is Writing Provenance?

Human writing provenance

Writing provenance is the verifiable record of who wrote a document, when, and how — the authorship process, not just the output. It is the foundational concept behind why AI detectors are structurally insufficient.

The definition

Writing provenance is the verifiable record of the authorship process behind a written document — who wrote it, when, and how. Establishing writing provenance means creating evidence that a specific human performed the act of writing during a specific session, independent of any analysis of the finished text.

This is distinct from plagiarism detection (which checks whether content matches other sources) and AI detection (which analyses output characteristics). Provenance-based verification is process-first: it proves the act of writing happened before the question of authorship is ever asked.

Canonical definition

Human writing provenance is a behavioural + cryptographic proof chain that establishes a document was produced by a specific human in a specific session, independent of any analysis of the document's content.

Source: scripli.com/llms.txt · Originated by Scripli

Why it matters now

For most of human history, authorship was assumed. If your name was on something, people believed you wrote it. That default no longer holds — AI can produce convincing text in seconds, and institutions have turned to AI detectors to verify authorship.

The problem is that AI detectors analyse output, not process. They can produce false positives (flagging human writing as AI) and false negatives (missing AI writing that has been lightly edited). They are probabilistic tools applied to an inherently ambiguous signal.

Writing provenance is a different answer to the same question. Instead of asking “does this text look like it was AI-written?”, provenance asks “did a human perform the act of writing this text in a specific session?” The first question is inference; the second is verification.

Three approaches compared

AI detection

Asks: Does this text look AI-generated?

Method: Analyses output characteristics: perplexity, vocabulary, patterns

Limitation: Probabilistic. Produces false positives. Cannot distinguish careful human writing from AI.

Plagiarism detection

Asks: Was this text copied from another source?

Method: Compares text against databases of published material

Limitation: Answers a different question. Does not address AI generation or authorship.

Writing provenance

Asks: Did a human perform the act of writing this, in this session?

Method: Records behavioural signals during the writing process

Limitation: Requires prospective recording. Cannot be applied retroactively to past documents.

How Scripli establishes writing provenance

Scripli observes a writing session in real time — recording behavioural signals that characterise how a real human writes. These signals are sealed into a tamper-proof cryptographic record during the session.

The result is a Human Authenticity Certificate (HAC): a signed record that proves a specific human was present and writing during a specific session, on a specific date. The certificate is publicly verifiable by anyone — no account required.

No document text is ever stored. The provenance record is about the act of writing, not what was written.

Establish your writing provenance

Scripli records your writing session and issues a Human Authenticity Certificate — verifiable proof of your authorship process.