Grammarly Pulls AI Tool After Backlash From Writers

Grammarly just hit pause on a feature that crossed a line many writers weren't willing to accept.

The writing assistant launched "Expert Review" in August, a tool that generated feedback appearing to come from famous authors, scientists, and journalists. The AI would analyze your writing and offer suggestions supposedly from people like Carl Sagan or bestselling novelists, all without asking those writers first.

The feature pulled information from "publicly available information from third-party LLMs," which raised serious questions about consent and legality. Living writers discovered their names and expertise were being used to sell an AI product they never agreed to support.

Grammarly tried to smooth things over by adding a disclaimer saying the experts didn't actually endorse the platform. That didn't satisfy many writers who found themselves unwittingly branded as Grammarly collaborators.

The company's next move was offering writers an opt-out option, which ignored a fundamental problem. How would writers know they needed to opt out if they didn't follow AI news closely? And what about deceased authors who couldn't object even if they wanted to?

The Bright Side

Today, Grammarly CEO Shishir Mehrotra announced the company is disabling Expert Review entirely while they reconsider the feature. The decision came after mounting criticism and a pending class action lawsuit against Superhuman, Grammarly's parent company.

Mehrotra said the tool was meant to help users discover relevant perspectives while helping experts connect with fans. But the execution missed the mark by treating writers' identities as raw material for AI training without basic respect for consent.

The reversal shows that companies can still listen and change course when they get it wrong. Writer advocacy worked, and Grammarly recognized that innovation shouldn't come at the cost of people's professional reputations.

This moment matters because it sets a precedent for how AI tools should handle real people's expertise and identities in the future.