Grammarly Disables AI 'Expert Review' Feature After Backlash Over Unauthorized Use of Writers' Names
Key Facts
- Grammarly has disabled its Expert Review feature, which offered generative AI feedback presented as coming from real writers, academics, and experts.
- The feature was launched in August by Superhuman, Grammarly's parent company, and used publicly available information from third-party large language models.
- Experts' names appeared without their permission or knowledge; living writers expressed strong disapproval upon discovering the tool.
- Superhuman CEO Shishir Mehrotra announced the disablement in a LinkedIn post, citing valid critical feedback from concerned experts.
- The company initially responded to complaints by offering an opt-out for living writers, but has now paused the feature for reassessment.
Grammarly has disabled its controversial Expert Review feature that generated AI-powered writing feedback credited to real authors, academics, and journalists without their consent, the company's parent Superhuman confirmed today.
The tool, launched in August, allowed users to receive editing suggestions framed as coming from notable figures based on the subject matter of their writing. These ranged from scientific luminaries and bestselling fiction authors to tech bloggers. The feature relied on "publicly available information from third-party LLMs," raising questions about data sourcing practices. A disclaimer noted that references to experts were for informational purposes only and did not indicate affiliation or endorsement.
Superhuman CEO Shishir Mehrotra addressed the issue directly in a LinkedIn post, stating the company is disabling Expert Review while it reimagines the feature. "The agent was designed to help users discover influential perspectives and scholarship relevant to their work, while also providing meaningful ways for experts to build deeper relationships with their fans," Mehrotra wrote. He acknowledged receiving "valid critical feedback from experts who are concerned that the agent misrepresented their voices."
The backlash was swift once writers became aware of the feature. Prominent tech journalist Kara Swisher was among those who publicly criticized it, with the tool claiming to offer advice in her style. Multiple reports indicate that staff members from publications like The Verge, including editor-in-chief Nilay Patel, were also referenced by the AI system.
How the Feature Worked
According to descriptions from multiple sources, Expert Review analyzed a user's writing and suggested feedback presented as originating from selected experts. The system would match subject matter to appropriate figures — living or deceased — drawing on their publicly known writing styles, opinions, and expertise.
Users could access the generative AI features through several methods, including highlighting text for inline suggestions, clicking a green lightbulb icon, or using the side panel in the Grammarly Editor. The Expert Review component stood out for its use of real individuals' identities rather than generic AI personas.
The approach immediately sparked ethical concerns about consent, identity usage, and the implications of AI systems impersonating public figures. Critics argued that using someone's name and perceived voice without permission constituted a form of digital impersonation, even if the output was generated by large language models trained on public data.
Initial Response and Further Criticism
When complaints first surfaced, Superhuman offered an opt-out mechanism for living experts. However, this solution drew additional criticism for failing to address deceased figures — who obviously could not opt out — and for placing the burden on writers to monitor AI news and actively protect their identities.
The feature's use of third-party LLMs also raised questions about the legality and ethics of web crawling practices used to gather training data. While companies increasingly rely on publicly available information to train AI models, the direct attribution to specific individuals crossed a line for many in the writing and journalism communities.
Reports from The Verge, Futurism, and others highlighted the intensity of the backlash. One article described the feature as "explosively controversial" due to its unauthorized use of writers' identities and mannerisms. Tech journalist Kara Swisher's strongly worded response, which included colorful language criticizing the company, was widely circulated.
Industry Context
This incident occurs amid growing scrutiny of AI companies' use of personal data, likenesses, and creative output. As generative AI tools become more sophisticated in mimicking individual writing styles, questions about consent, attribution, and intellectual property have moved to the forefront of industry discussions.
Grammarly, long known for its grammar and style checking tools, expanded aggressively into generative AI features in recent years. The company was acquired by Superhuman, which has positioned itself at the intersection of productivity software and artificial intelligence. Expert Review represented an ambitious attempt to add personality and expertise to AI writing assistance, but appears to have underestimated the importance of obtaining consent from the experts being emulated.
The controversy also highlights tensions between innovation in AI applications and respect for individual rights. While companies argue that publicly available information can be used to create new tools, affected individuals often view the commercial use of their identities and voices as exploitative.
Impact on Users and Developers
For Grammarly users, the disablement of Expert Review means the loss of a feature that promised uniquely insightful feedback from domain experts. Many had found value in receiving suggestions framed through the perspectives of influential thinkers, even if the implementation proved problematic.
Developers and product teams across the AI industry are likely watching this situation closely. The rapid backlash and subsequent disablement demonstrate the reputational risks of launching features that appropriate individuals' identities without clear consent mechanisms. Companies may need to invest more heavily in ethical review processes and stakeholder consultation before releasing similar tools.
The incident could also affect user trust in Grammarly's other generative AI offerings. While the core grammar and writing assistance tools remain available, the Expert Review controversy may cause some users to approach the company's AI features with greater skepticism.
Writers and journalists, particularly those whose names were used, have gained a platform to voice concerns about AI's impact on their profession. The episode reinforces ongoing debates about how creative professionals should be compensated or credited when their work influences AI systems.
What's Next
Mehrotra indicated that Superhuman plans to "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."
This suggests potential future iterations might include:
- Explicit consent and participation programs for experts
- Clearer disclosure about AI-generated nature of all suggestions
- Options for experts to customize or approve how their perspectives are represented
- Alternative approaches that avoid direct attribution to real individuals
However, no specific timeline has been provided for when or if a revised version of Expert Review might return. The company appears focused on addressing the core issues of consent and representation before proceeding.
The broader AI industry continues to grapple with similar challenges as tools become more capable of emulating human voices and expertise. This case may serve as a precedent for how companies handle the use of public figures' identities in generative AI applications.
For now, Grammarly users will need to rely on the service's other AI writing tools, which do not attribute suggestions to specific real-world experts. The company has not indicated any changes to its general generative AI features beyond the Expert Review disablement.