Does Grammarly detect AI? Its detector and Authorship explained

Grammarly's AI Detector page offers a paste box with no sign-up and returns a percentage described as how much "of this text appears to be AI-generated", read 9 November 2026. The page's accuracy claim is specific: the detector "achieves 99% detection accuracy and ranks #1 on RAID's independent benchmark", a public benchmark for telling human and AI text apart (Grammarly AI Detector, read 9 November 2026).

The same page carries Grammarly's caveats, and they are worth quoting because the vendor wrote them. "No AI detector is 100% accurate." Detectors "do not track authorship, which means even human-written text can be flagged as AI-generated", while "lightly edited AI-generated text may evade detection". Grammarly's own description of its method is that the model "segments the text and analyzes each section for language patterns that are often linked to AI-generated writing", trained on tens of thousands of human and AI texts, read 9 November 2026.

Grammarly's detector is not one of the five in this site's run, which uses GPTZero, Originality.ai, Copyleaks, Pangram and ZeroGPT. For comparison, GPTZero publishes a false positive rate of no more than 1 percent for its document classifier (GPTZero technology page, read 9 November 2026); Grammarly's page publishes a detection accuracy figure and a benchmark rank but no false positive figure. Those are different measures, and a reader comparing detectors should notice which one each vendor chose to publish. The guide to whether AI detectors work sets every vendor claim against the independent studies.

Authorship is Grammarly's second answer to the question, and the more unusual one. Grammarly describes it as an opt-in feature that "automatically tracks your writing process in Microsoft Word, Google Docs, or the Grammarly Editor", labelling text as typed, pasted or AI-assisted, and generating a report that colour-codes each sentence by source (Grammarly Authorship, read 9 November 2026). The page states it can also be enabled for Canvas LMS submissions, that it is available on Grammarly Free and Grammarly Pro, and that Grammarly "doesn't auto-send it to your professor or anyone else": the user creates the report and shares the link.

Grammarly draws the contrast itself: "AI detectors only guess from the final draft. Authorship records your actual writing process." Its FAQ adds that detectors "don't keep track of document history; they just analyze your text in a moment in time", while Authorship "helps you keep track of version history and where text in your writing came from", read 9 November 2026. The report can replay the document's creation from first word to last.

For a student, the difference is the difference between a guess and a record. A detector can say a paragraph looks AI-written; Authorship can show it was typed over forty minutes on a Tuesday. The guide can Google Docs detect AI covers the same idea using Google's own version history, which needs no extra product.

This is the question behind many searches, and Turnitin has answered it in its own FAQ. Turnitin states that its detector "is not tuned to target Grammarly-generated spelling, grammar, and punctuation modifications", and that in tests on human-written documents, "in most cases, changes made by Grammarly (free & premium) and/or other grammar-checking tools were not flagged as AI-written" (Turnitin, AI writing detection FAQs, read 9 November 2026).

The same answer draws a line. Turnitin adds that this "excludes content generated by Grammarly's generative AI-powered features, including draft generation, paraphrasing, summarizing, and other features", and that "content produced using these features will likely be flagged as AI-generated by our detector", read 9 November 2026. Grammarly's AI Humanizer is a generative rewrite, so by Turnitin's own statement it sits on the flagged side of that line, which matches what Grammarly says about it: the humanizer "is not intended to bypass AI detectors" (Grammarly AI Humanizer, read 9 November 2026).

Yes, and the two products come from the same account. Grammarly's humanizer offers four preset voices and a custom voice built from a writing sample, works in six languages, and is free with basic humanization, with the full version on Pro, read 9 November 2026. Grammarly prices Pro in "AI prompts", 100 a month on Free and 2,000 a member a month on Pro, never in humanizer words (Grammarly plans, read 9 November 2026). The Grammarly AI Humanizer review covers plans, voices and the main limitation on the same fields as every other tool, and Grammarly vs QuillBot AI Humanizer sets it against its closest rival. Whether the rewrite does its job is the question of does Grammarly humanizer work.

One practical point follows from Grammarly running both a detector and a humanizer. Grammarly's AI Detector page suggests pairing the detector with its rewriter to "refine flagged language", read 9 November 2026. A reader should expect Grammarly's detector to be the detector most aligned with Grammarly's rewrites, and should check any text that matters on a detector from a different company, because detectors disagree.

Treat it as Grammarly tells you to: one input, not a verdict. The vendor's page says no detector is fully accurate, that human text can be flagged, and that edited AI text can pass. Independent work points the same way; in a 2023 test of 14 detectors, none reached 80 percent accuracy and paraphrased AI text was mostly missed, as the AI detectors guide sets out with the citation.

No detector score should be the only evidence in a misconduct case. Students should check their institution's AI policy before using any writing tool, including Grammarly's generative features; a humanizer does not make undisclosed AI use acceptable. The rest of this series is in the guides hub.