Skip to content
IFAI

AI tools

Grammarly tracks how you write to prove it's not AI

Published Reach 36

Grammarly just introduced Authorship, a feature that monitors the writing process in real time to distinguish between human typing, AI generation, and copy-pasted text. Instead of relying on flawed post-hoc detection algorithms, it creates a replayable audit trail to protect students from false accusations. A beta launches in Google Docs next month before expanding to Word and Pages.

Grammarly tracks how you write to prove it's not AI
Image: zdnet.com

Turning in a paper you spent all night writing, only to have a teacher claim an AI detector flagged it as machine-written, has become an everyday nightmare for students.

OpenAI tried to solve this and failed, shutting down its own detection classifier in 2023 because it was embarrassingly inaccurate. Once words are on the page, distinguishing human prose from AI output is basically a guessing game.

Grammarly is taking a different approach. The company just announced Grammarly Authorship, a new tool designed to stop guessing after the fact and track the writing process as it happens.

A dashcam for your document

Most AI detectors analyze a finished piece of text and calculate probabilities. Authorship takes a radically different route by staying active while you write across 500,000 apps and websites.

Because it watches the session in real time, it knows whether you manually typed a sentence or pasted a block of text from an external tab.

The system automatically tags every section of your document. It separates human-typed text, AI-generated drafts, AI-assisted rewrites, pasted snippets, and basic grammar or spell checks.

It also compiles a full analytics report showing total time spent and the number of active writing sessions. There is even an authoring replay feature that plays back how the draft was built from a blank page.

Authorship hits Google Docs in beta next month for both free and paid users. Support for Microsoft Word across desktop and web, as well as Apple Pages, is slated to arrive by the end of the year.

Fixing the classroom standoff

Grammarly is squarely targeting the education market. False positives have turned grading into an adversarial mess between educators and students.

Jenny Maxwell, head of Grammarly for Education, noted that half of young people aged 14 to 22 have used generative AI, yet schools still lack clear guidelines, leading to an overreliance on broken detection tools.

Authorship gives students an objective paper trail to prove their work is authentic. Early next year, the tool will also actively prompt users to properly cite any external material they paste in.

Access will not be locked behind an expensive tier. It will roll out to Grammarly's free plan, its $12 monthly Premium tier, $15 Business plans, and educational accounts.

Why it matters

For anyone who writes for school or work, this is an insurance policy. If an instructor or client accuses you of using ChatGPT, an authoring replay showing your keystrokes and time spent settles the argument immediately. Given that it works on the free tier, students should turn it on the moment it lands in Google Docs next month.

For the AI industry, Grammarly just shifted the battlefield. Trying to identify AI text purely by reading the final draft is an impossible cat-and-mouse game that even OpenAI gave up on. Replacing predictive statistical guesses with an immutable process log is a remarkably practical pivot.

That said, the approach carries real trade-offs. Normalizing session recorders that track every paste, pause, and keystroke creates an unsettling level of surveillance. If institutions start mandating process logs to accept assignments, the line between academic honesty and intrusive tracking will blur quickly.

Sources (1)

  1. [1]
    zdnet.com(opens in a new tab)Originalzdnet.com

Best of AI, daily