Grammarly Sued Over Expert Review Feature

Over the past week, Grammarly and its parent company Superhuman have been drawing a great deal of unwanted attention to their “Expert Review” feature. Though the feature has been out since August 2025, many of the “experts” have only recently become aware of it and are upset.

The feature is part of their “agent-powered” editor. Though the editor provides the proofreading and plagiarism detection tools it has always had, it added an AI grader, a “reader reactions” agent and more.

However, it’s the expert review feature that has come under the most fire recently. Last week, Miles Klee at Wired wrote an article highlighting the feature, noting that it gives “expert” advice generated by an AI system. However, none of the experts involved gave permission for their names to be used, and many have long since passed away.

Over the next few days, backlash against the feature grew. However, things came to a head yesterday when Grammarly announced it was disabling the feature and launching an option for experts to opt out moving forward.

According to Ailian Gan, Superhuman’s director of product management, “After careful consideration, we have decided to disable Expert Review as we reimagine the feature to make it more useful for users, while giving experts real control over how they want to be represented — or not represented at all.”

But even as Grammarly was pressing pause on the feature, journalist Julia Angwin filed a class action lawsuit against Superhuman. The lawsuit alleges that the feature violates the privacy and publicity rights of the various experts included in the product. The lawsuit alleges that the damages to the plaintiff class exceed $5 million. 

So what exactly is happening, and what takes place from here? To answer that, we need to first look at the feature itself and understand why it’s become so controversial.

What is (was) Grammarly’s Expert Review?

As I mentioned before, Grammarly Expert Review was a feature first introduced in beta in August 2025. It was part of an overhaul of their editor to introduce more AI features such as an AI grader, reader review and a citation finder.

The idea behind the expert review feature was to parse what your document is about and highlight issues/concerns that an expert in that field might raise. 

For example, when I was testing the feature on an article about academic integrity, I would get “recommendations” from experts such as Sarah Elaine Eaton and Tricia Bertram Gallant. However, the recommendations weren’t actually from them. It was an AI’s interpretation of what they might say based on the AI’s interpretation of books, articles and other material they had written.

Grammarly, for its part, stated this in a footer that read, “References to experts in this product are for informational purposes only and do not indicate any affiliation with Grammarly or endorsement by those individuals or entities.”

However, that disclosure was in the fine print and was easily missed. At the time, I reached out to several people I knew who were listed as experts, and none were aware that their names were being used.

From a purely attribution standpoint, this is incredibly bizarre. Grammarly was taking AI suggestions and attributing them to experts solely because the AI was trained on that expert’s content. Instead of passing off expert content as human-written, it’s passing it off as being written by very specific experts, unless one very closely reads the fine print.

It’s easy to see why Angwin and so many others are upset. 

So What’s Next?

To be clear, I am not an expert on privacy or publicity rights, so I won’t comment on the likelihood of the lawsuit succeeding. 

That said, I would encourage those who are impacted by this and wish to take legal action to discuss it with their lawyer or, at the very least, reach out to the lawyers who filed the lawsuit.

But this story raises a difficult question. Though we’ve discussed that failing to disclose the use of AI makes one a plagiarist, this is something different. Yes, the AI is being disclosed, but, at the same time, it’s being presented as guidance from experts in the field.

However, is this much different from how AI is being presented? People already treat AI as a general expert. People tweet to Grok or ask ChatGPT to verify facts or answer complex questions, placing their faith in those answers. 

Yet, we already know that AI systems are prone to hallucinations, contain biases and generally make mistakes. While that is true of humans, this system sought to increase trust in AI by tying AI’s answers to those of actual human experts.

While the AI suggestions were based on the writings of the human experts it cited, they were still just AI suggestions. And they were AI suggestions that Grammarly was claiming those experts would likely make.

But how accurate were they? We don’t have a way of knowing. But for Grammarly, it’s likely a moot point. If they were accurate, they were only accurate because of the use of copyrighted material created by the experts. If they were inaccurate, they misrepresented what the expert believes or says. Neither is a good outcome.

And the copyright question is especially interesting. AI companies have claimed that their work is fair use and that their answers are based on billions of works. But with these AI “experts”, they are acknowledging that the answer is based only on that one person’s materials. That fair use argument becomes even more shaky.

In short, this feature was ill-advised from the outset, and Grammarly’s decision to shelve it was the right move. But, given that the lawsuit has already been filed, it may be too little, too late.

Bottom Line 

I’ve been incredibly fortunate that, in the 20-plus years I’ve run this site, I don’t have many regrets.

However, I do regret not focusing more on this “agent” back in August. Though I told several friends and colleagues that their names were being used, I never followed up with a full article. I definitely should have.

The reason I didn’t was that there were other Grammarly-related issues at the time, and, as I said above, I’m not an expert in privacy or publicity rights. Still, one doesn’t need to be an expert to know that this is wrong, and I missed a crucial opportunity to shine a light on it.

Moving forward, I’m going to try to be more willing to step outside my comfort zone and expertise, especially when something is this clearly problematic.

My hope is that this leads to a broader conversation about how AI is used and the “guidance” that it provides. Regardless of how I feel about AI, it clearly isn’t going anywhere, and it’s even more clear that many people are putting far more faith into it than it deserves.

So what happens when an AI claims to offer the same or similar advice as human experts? Companies like Character.ai and PolyBuzz have been doing this with celebrities and other experts for some time. Though Grammarly muddied the waters by making its lack of affiliation less clear, what they did isn’t particularly special.

It’s an AI issue that hasn’t been getting much attention. Admittedly, of all the threats and problems AI has created, it may not be the biggest. But it is still extremely important.