Add context to an AI review.
Bring the document, review its available history and add a statement about AI use. Share those details alongside the detector reading. The signature protects the issued record; it does not independently verify the history or prove authorship.
Google will ask you to let Nota see the file you pick — that one document, nothing else in your Drive. Nota reads its history once and shows you what it found. It works after the fact: the history is already in the document.
Three steps, two of them yours.
Bring the document
Upload the .docx, or pick the document in Google’s own file picker. One file. It works after the fact: the history is already in the document.
Read the ledger
What the history recorded, said as a receipt: sessions, editing time, what arrived at a typing pace and what arrived faster, what was cut, the final size. Nothing has been stored yet.
Issue the signed record
Add a declaration if you want. The record brings submitted history and check results together. A reader can verify the signature and compare their document with the recorded fingerprint.
A ledger, not a grade.
Dates, durations and counts provide context for a review. They are supplied by the browser, with their reported source shown. The AI reading is separate and uses up to the first 1,024 tokens; a measured false-positive rate appears when available.
Typed ≈4,100 words over 3 sessions (2 h 40 min, 12–15 May 2026) and cut ≈1,900. 310 words arrived in 2 blocks — faster than typing between saves. The final document is 2,650 words: a 1.7 : 1 written-to-final ratio.
Illustrative submitted history. Its origin is not independently verified.
- Submitted
- The history above; origin and connection to this text unverified
- Observed
- The character check on the submitted text
- Inferred
- The model’s reading, with the rate at which it is wrong about real human writing
- Declared
- Your statement, if you make one — signed as yours, not verified
- Signed
- Ed25519, by Know Nota; verifiable at the link against the document’s fingerprint
Who brings one.
A writer answering a question
Bring available history and explain how you worked. A reviewer can consider that context alongside the text and other evidence.
An assignment that allows AI
State what tools you used and how. The record preserves that statement without claiming to verify it or decide whether it meets the assignment’s policy.
An editor or a client
Read the detector result, matching sources, submitted history and writer’s declaration together before deciding what needs a closer look.
Teachers: how to read a Writing Record. Writers who work in Google Docs every day: the Chrome extension builds the record from inside the document.
Questions people ask first.
What does Nota read, exactly?
For a Google Doc: that one document’s revision history — when it was edited, how much text arrived per save and at what pace, what was cut, how many editors. For a .docx: the properties the editing application wrote into the file — created, last saved, editing time, save count. For a PDF or text file: the words only. In every case Nota also reads the final text and puts its own reading in the record, with the measured error rate.
Does this read everything I do?
No. Nothing is installed and nothing runs in the background. You bring one document; Nota reads that document once, shows you what it found, and stores nothing until you choose to issue the record. With Google Docs the permission you grant covers the file you pick and nothing else in your Drive.
Is a Writing Record proof that I wrote it?
No. History is supplied to the record by your browser, and its origin and connection to the text are not independently verified. The signature lets a reader check that the issued record has not changed; it does not authenticate the history or prove authorship. Typing can be scripted, file properties can be edited, and a declaration is your own statement.
What does it never say?
It never says “human-written” or “AI-written”, never names a generator, and never scores you. The ledger is counts, dates and durations. Nota’s reading of the text appears as one line with its measured error rate, never as a verdict.
What does a Word file include?
A .docx can carry file properties such as editing time, save count and dates. This flow reads those properties, not a complete Word revision history. Google Docs can provide revision snapshots when access is available. Both arrive in the record through the browser and are labelled as unverified supporting context.
What does it cost?
A Writing Record — one with the document’s history in it — is free to issue, share and download on every plan. Records for documents that carry no history (PDF, text) are on Individual and up, like any other Nota Record.
What if I used AI and was allowed to?
Add a declaration such as “AI drafted from my notes” or “AI edited my draft”. It is preserved as your statement beside the submitted history and the detector reading, so a reviewer can consider it against the policy for that work.
Before you share your history, read what Google Docs version history shows, and what it does not. For the detector side, see how often Nota is wrong.
Ready when you are.
Nothing is stored until you issue the record. Free.