The writing was not the first thing that caught my attention. It was the quiet. Late at night, in a dimly lit apartment, a draft was open on a laptop screen—the kind of quiet time when sentences typically start to breathe a little. Coffee is cooling next to the keyboard. The cursor is slowly blinking. Grammarly then awoke.
It behaved as it always had at first. Here’s a green underline. There was a courteous suggestion. Use “enormous” in place of “very big.” A comma should be moved. Not very dramatic. For many years, Grammarly was like a courteous proofreader that sat over the shoulder.
Then the “Expert Review” button emerged.
It felt harmless to click. An additional tool in an expanding array of digital writing aids. However, what followed had a strangely unsettling quality, akin to hearing your own voice slightly distorted.
The program started giving advice purportedly inspired by well-known authors. The names of journalists, scientists, and best-selling authors were displayed in blue next to the recommendations. Though subtle, the implication was clear. The draft was being shaped by these individuals. However, they weren’t.
| Category | Details |
|---|---|
| Company | Grammarly |
| Founded | 2009 |
| Headquarters | San Francisco, California |
| CEO | Shishir Mehrotra |
| Core Product | AI-powered writing assistant |
| Daily Users | Over 40 million users worldwide |
| New Feature | “Expert Review” AI editing suggestions |
| Controversy | Suggestions attributed to well-known writers without consent |
| Legal Issue | Class-action lawsuit related to the AI “Expert Review” feature |
| Industry Context | Competition from ChatGPT, Claude, Gemini |
| Reference Sources | WIRED coverage of the Grammarly AI controversy |
| Platformer analysis of the Expert Review feature |
A few of those names belonged to authors who had never consented to take part. In fact, a number of them later claimed they were completely unaware that their identities were being used. Product managers may have thought this was a clever feature in a conference room somewhere. a method for giving artificial intelligence a more human feel.
Rather, it had an oddly dramatic quality.
One recommendation was to begin a story with “sensory imagery.” Another suggested increasing the tension in the story. The advice wasn’t too bad. However, it was strangely generic, akin to suggestions taken from a workshop on creative writing and diluted into something more secure.
There was also the peculiar tone. The software spoke with the assured courtesy of someone who has never had trouble finishing a challenging paragraph.
There’s a feeling that something subtle starts to change as you watch the program rewrite passages of text. Sentences start to flow more naturally. more tidy. Moreover, it’s flatter.
These things are often noticed by writers.
Once having a somewhat awkward rhythm, a phrase now appears polished but unfamiliar. A more common word takes the place of an uncommon one. The draft loses the minor flaws that gave it character while improving technically.
It’s difficult to ignore how fast the voice starts to stray.
Grammarly established its reputation for years as a grammar checker, which is more akin to spell checking than authorship. However, the goal of such tools has changed with the introduction of large language models. Your writing is no longer being corrected by the software. It seeks to direct it.
Or maybe guide it.
Some detractors believe the business is merely responding to pressure from more recent AI platforms. ChatGPT, Claude, and Gemini are already remarkably adept at editing tasks. Suddenly, a stand-alone grammar checker seems like a thing of the past.
Grammarly grew as a result.
The company gradually repositioned itself as an AI productivity ecosystem rather than a grammar tool in recent years by acquiring the email tool Superhuman and the collaboration platform Coda. The decision makes sense from a business standpoint. Platforms are adored by investors.
However, the experience of writing conveys a different message.
When a rough draft was run through the system late one evening, the software recommended tightening a paragraph that purposefully strayed a little. Sometimes writers stretch sentences in the same way that musicians stretch notes, so the meandering was intentional. The AI didn’t agree.
It made the passage more effective and well-balanced.
Better technically. worse on an emotional level.
These kinds of tools seem to misinterpret what writing is. Language is more than just well-organized information on a page. It conveys personality, hesitancy, and occasionally even stubbornness. It is possible for a sentence to feel alive even if it is awkward.
Symmetry is preferred by algorithms.
The notion that these recommendations were influenced by particular experts—investigative journalists, science communicators, and technology critics—was also presented by Grammarly at some point. The action caused controversy in the journalism community.
A number of authors whose names were included in the article claimed they had never been contacted.
One well-known tech columnist joked that he would at least like to get paid if the AI version of himself was editing articles. A deeper uneasiness was concealed by the humor. As it happens, identity is evolving into yet another raw material in the AI economy.
As the controversy develops, it’s hard to ignore a broader trend emerging in the tech sector. Businesses that create language models analyze vast volumes of writing, including books, articles, and essays, and then create tools that can imitate those styles. Seldom do the original authors get a call.
In the end, Grammarly accepted the criticism and gave experts the choice to withdraw. That response implies that the business understands how delicate the situation is. It’s still unclear if it really comprehends the deeper cultural tension. Because attribution might not be the true problem. It could be authorship.
Something delicate vanishes when software silently rewrites a paragraph, standardizing every phrase and smoothing every edge. Millions of AI-edited emails, blog entries, and reports start to sound like the same composed, predictable voice. The writing is effective. However, nobody is surprised by it anymore.
People who write for a living are experiencing an odd feeling as they watch this develop online. Not quite panic. It’s more akin to a subdued suspicion that language is being subtly condensed into a more secure form.
Grammarly and other similar tools were designed to help people communicate more effectively.
