Humanums vs Grammarly Authorship: where the text came from, or how it was written.

Turn Authorship on in a blank document and Grammarly starts sorting text by how it got there. Typing counts as typed. Pasting is labeled as copied from a website or from an unknown source. Text produced or rewritten by Grammarly's AI features, or by other generative AI, is labeled as such. When you are done you generate a report, make it shareable, and send the link to whoever asked, usually an instructor. A replay shows the document being built.

For a student whose teacher already uses Grammarly, this is close to frictionless. It costs nothing, it lives where they already write, and it hands the teacher something more useful than a detector percentage. That is a good product decision and worth saying plainly.

The categories answer a provenance question: did this sentence arrive through the keyboard, the clipboard, or an AI feature? They do not answer whether the typing looks like composition. Someone who reads AI output in one window and types it into another produces text that arrives through the keyboard, sentence by sentence.

Humanums is built around that distinction. Composing and transcribing produce different behavior: composition stalls before hard sentences, goes back to rework earlier paragraphs, and varies its rhythm with the difficulty of the thought. Transcription runs at a steady clip and rarely revisits anything. The scoring weights exactly those differences, and it never stores the characters you type.

In November 2025, Plagiarism Today tested Authorship against Grammarly's own Humanize feature. A Gemini-generated essay that the report had labeled as unknown came back with a green header reading 75% typed by a human after being run through Humanize, even though the fine print attributed that 75% to text rephrased with Grammarly's AI. Grammarly's parent company replied that it was changing how humanized text is displayed.

The specific display has likely been fixed, and we are not claiming otherwise. The structural point remains: when the company labeling the text also sells the AI features being labeled, labeling decisions are product decisions. Humanums sells no writing assistant, so it has nothing to protect in how AI-assisted text is described.

An Authorship report is a document you share with a recipient. A Humanums certificate is a public artefact: signed, bound to a hash of the exact certified text, and reachable at a stable URL with a badge you can embed beside the work. If the text changes after certification, the certificate no longer matches it.

That difference matters most outside the classroom. A freelancer handing a deliverable to a client, a journalist publishing under a byline, or a newsletter writer whose readers are suspicious of AI wants evidence that travels with the piece and that a stranger can check without being sent a link.

If your instructor asks for an Authorship report, give them one. If you write in Word and want a zero-cost record of where your text came from for a single assignment, Authorship does that well. If what you need is a source breakdown, for instance to show which passages were quoted, it does something Humanums does not.

Humanums is the better fit when the question is specifically whether a person composed the words, when the evidence needs to be public and durable, or when the writing is going somewhere other than a gradebook.