Grammarly’s ‘Expert Review’ Lets Stephen King Criticize Your Writing (Without His Permission)

Remember being a teacher’s pet? Or that one professor whose margin notes actually made you a better writer? Grammarly is betting you do, and that you’ll pay for an AI that simulates exactly that experience—except the “professor” might be dead and definitely didn’t sign up for this.

The company, which recently rebranded its parent as Superhuman while keeping the Grammarly name, has quietly rolled out an “Expert Review” feature. Pick from a list of real writers and academics—Stephen King, Neil deGrasse Tyson, the late Carl Sagan—and the AI will produce a critique of your text in their style. The disclaimer is buried but clear: “References to experts in this product are for informational purposes only and do not indicate any affiliation with Grammarly or endorsement.” In other words, these people have nothing to do with it.

I’ve been watching Grammarly’s slow creep from spell-checker to full AI writing suite for years. The latest additions include a chatbot that answers questions as you write, a paraphraser, a “humanizer” that tweaks your voice, and even an AI grader that scores your document like a college professor would. There’s also a tool that flags phrases that sound too AI-generated—which feels ironic given that the entire platform now runs on LLMs. But the Expert Review is where things get weird.

Vanessa Heggie, a history professor at the University of Birmingham, posted a screenshot on LinkedIn showing the AI offering analysis from a model based on David Abulafia, an English historian who died in January. She called it “obscene.” C.E. Aubin, a historian at Yale, told WIRED that the feature “validates the profound mistrust so many scholars in the humanities have for AI and its seemingly constant use in fundamentally unethical ways.” Her point sticks: these aren’t expert reviews because no experts are involved. It’s just an LLM trained on scraped work, wearing a dead person’s name as a costume.

Superhuman’s senior communications manager Jen Dakin says the feature “doesn’t claim endorsement or direct participation from those experts” and that it “provides suggestions inspired by works of experts.” That’s a careful phrasing, but it doesn’t address the core problem. If I write a paper and the Grammarly AI tells me “Carl Sagan would suggest adding more citations,” I’m trading on his credibility without his consent. And if he’s dead, he can’t even object.

The legal landscape here is murky. Multiple copyright lawsuits are already winding through courts over similar uses of copyrighted material to train AI. Grammarly’s approach feels like it’s daring someone to sue, banking on the fact that most authors don’t have the resources to fight a company with Superhuman’s backing. Stephen King and Neil deGrasse Tyson didn’t respond to requests for comment, which I read as either “we’re too busy” or “we’re consulting lawyers.”

An independent review by WIRED tested the feature and got recommendations from AI agents modeled on Steven Pinker, Gary Marcus, William Strunk Jr., Pierre Bourdieu, Margaret Mitchell, and Virginia Tufte—the last of whom died in 2020. The AI’s advice was generic: “Replace repetition with vivid, varied sentence patterns.” That’s not an expert review, that’s a fortune cookie written by a machine.

Aubin also pointed out something that bothers me more than the legal issues: the reduction of personhood. “The issue of ‘reanimating’ the dead so cynically” is awful enough on its own, she said, but there’s also the implication that scholars are reducible to their work, that the human can be removed from the equation entirely. That’s a dangerous idea to normalize, especially when the humanities are already under attack from every direction.

And then there’s the practical question: does any of this actually help you write better? I’ve been using Grammarly’s standard features for years, and they’re fine for catching typos and passive voice. But the Expert Review feels like a gimmick designed to justify a higher subscription tier. The AI doesn’t have insight into your argument, your audience, or your intent. It just pattern-matches against a corpus and spits out generic advice that sounds like the author it’s mimicking. If you want real feedback, ask a colleague. If you want to feel like Stephen King is reading your work, this is cheaper than a séance.

Grammarly’s parent company Superhuman is betting big on AI features to differentiate itself in a crowded market. Microsoft, Google, and a dozen startups are all offering similar writing tools. But this particular feature crosses a line that most have been careful to avoid. Using real people’s names and reputations to sell a subscription service, without their permission, is a choice. It’s a choice that says “we’ll figure out the ethics later.” I suspect the courts will have something to say about that.