Grammarly disabled an AI review feature that cloned editors' style without their knowledge

Superhuman disabled the Expert Review feature in Grammarly — the one that presented AI-powered edits as advice 'inspired' by the style of specific real-life journalists and editors. The authors whose names were used were never asked. The scandal erupted after it became clear that the feature referenced The Verge's Editor-in-Chief and other staff members of the publication, as if they had personally reviewed users' texts.

In reality, there was no involvement from real people — their names merely served as marketing wrappers for an ordinary language model. Superhuman's Product Director Ailian Gan issued a public apology: 'After careful analysis, we have decided to disable Expert Review and rethink the feature. We want to make it truly useful for users while giving experts real control over how they are presented — or not presented at all.

Based on the feedback we received, we clearly fell short. We're sorry, and we'll do things differently going forward.' The Grammarly case fits into a broader discussion about how AI companies use the names and reputations of real people to promote their products.

Using names without consent — even if the texts themselves are not copied verbatim — already raises legal and ethical questions in several jurisdictions. For journalists, the situation is particularly painful: their professional authority is directly tied to what they actually recommend. It remains unclear in what form Expert Review will return.

The company promises to give authors the ability to decide for themselves whether to participate in similar features. This sounds reasonable — but the question arises: why was such a basic requirement not built in from the start?