Copy editor Benjamin Dreyer pasted several paragraphs of lorem ipsum into Grammarly's Expert Review tool. The AI returned writing advice attributed to Stephen King. Tips on improving placeholder text that isn't even English.
That test captures everything wrong with the feature Grammarly (now operating under the Superhuman brand) just pulled from its product.
What Expert Review Actually Did
Launched in August 2025 as part of Grammarly's $12/month Pro subscription, Expert Review let users click a button to receive AI-generated editing suggestions presented as coming from specific real-world writers, journalists, and academics. The system chose which "expert" to display. Users couldn't pick.
The tool used large language models (LLMs, the same technology behind ChatGPT) to generate feedback, then slapped a real person's name and a description of their editorial style on top. Journalist Casey Newton tested it on a colleague's article about OpenAI protests. Advice attributed to investigative reporter John Carreyrou suggested: "Try starting with sensory imagery, a chant echoing off glass towers or chalk dust in the air, to immerse the audience instantly." Advice attributed to Kara Swisher suggested inserting a narrative digression "inspired" by her podcast.
None of these people were asked. None were paid. A small disclaimer buried in the interface stated the references "do not indicate any affiliation with Grammarly or endorsement by those individuals."
The list of impersonated experts included Stephen King, Neil deGrasse Tyson, the late Carl Sagan, and dozens of working journalists and academics. Investigative journalist Julia Angwin called the imitations a "slopperganger," a combination of AI slop and doppelganger. She noted the actual editing suggestions were poor: "The edits were not good... were making the sentences worse, more complex."
From Opt-Out to Shutdown to Lawsuit
Grammarly's initial response was to offer an opt-out email address. Writer and former Verge Creative Director James Bareham summarized the reaction: "I think 'We're going to keep stealing your stuff until you tell us you don't want us to steal your stuff' isn't quite the defense Grammarly thinks it is."
Kara Swisher was more direct: "You rapacious information and identity thieves better get ready for me to go full McConaughey on you. Also, you suck."
The opt-out approach lasted about a day. On March 11, CEO Shishir Mehrotra posted on LinkedIn that the feature was being disabled, writing: "We hear the feedback and recognize we fell short on this."
By March 12, Julia Angwin had filed a class-action lawsuit in the Southern District of New York against both Superhuman and Grammarly, alleging misappropriation of identity and privacy rights violations. The suit seeks damages exceeding $5 million. Within 24 hours of filing, the plaintiff's lawyer received over 40 inquiries from other affected individuals.
Product Director Ailian Gan issued a fuller statement: "Based on the feedback we've received, we clearly missed the mark. We are sorry and will do things differently going forward."
What This Means for AI Writing Tools
The core issue isn't that AI generated bad writing advice. It's that a company with 40 million daily users across 500,000 organizations commercially exploited real people's identities through a subscription product without consent.
Functionally, Expert Review did what anyone can do by prompting ChatGPT to "edit in the style of [writer]." The difference: Grammarly curated a list of real people, monetized their names at $12/month, and presented AI output as if it carried those experts' professional authority. That's a business model built on someone else's reputation.
This lawsuit could set precedent for other AI writing tools regarding consent and identity use. Microsoft Editor, ProWritingAid, and similar products will be watching closely. The legal question of whether AI-generated content can be commercially attributed to real people without permission hasn't been tested at this scale before.