Grammarly’s “Expert Review” was not a review by a famous author, journalist, scientist, or academic. It was an AI feature that analyzed a draft and generated suggestions inspired by those people’s publicly available work. The named individuals were not the reviewers, and reporting found that some had not given Grammarly permission to represent them this way.
What Grammarly’s Expert Review actually did
Launched in August 2025, Expert Review appeared inside Grammarly’s writing assistant and sidebar. It promised feedback through the perspective of influential writers, scholars, scientists, and journalists—including figures such as Stephen King, Neil deGrasse Tyson, Carl Sagan, and William Zinsser.
As an Amazon Associate I earn from qualifying purchases.
A user submitted a draft, and Grammarly’s system produced automated revision suggestions. The company said it selected names according to the substance of the writing and the relevance or prominence of each person’s published work. The suggestions were described as being “inspired by” those works.
Free tools Windows power users keep installed
One-click scans. No signup required.
That wording matters. “Inspired by” means the system generated text based on patterns, ideas, or characteristics associated with published material. It does not mean that the named person read the document, approved the recommendation, or would have made the same editorial decision.
#1 Best Overall
WIRED reported on the feature’s use of recognizable names and the disclaimer that the references did not indicate affiliation or endorsement. (WIRED)
Were real experts involved?
No evidence in the reporting indicates that the named individuals reviewed users’ documents, helped produce the feedback, or endorsed the feature.
The Verge reported that several of its journalists and editors appeared in Expert Review despite not giving Grammarly permission. WIRED similarly reported that the authors and scholars named in the feature were not involved in the process. (The Verge)
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsGrammarly’s parent company, Superhuman, said the names appeared because the individuals’ published works were publicly available and widely cited. Its disclaimer stated that the references did not indicate affiliation or endorsement.
That is a defense based on attribution and discovery: the company said it was surfacing influential perspectives rather than claiming to arrange a personal consultation. But public availability is not the same as consent. A person’s work can be publicly readable without that person agreeing to be used as a selectable AI feedback persona in a commercial product.
Why the presentation created a problem
The controversy was not simply that an algorithm gave writing advice. Ordinary AI feedback can be clearly labeled as automated. Expert Review wrapped that feedback in the names and reputations of real people.
Rank #2
The distinction can be illustrated this way:
- Transparent automated feedback: “AI critique: Your introduction needs a clearer thesis.”
- Persona-based feedback: “Stephen King’s review: Start with a stronger hook.”
The second formulation may cause users to infer a connection to Stephen King’s actual judgment, even if a disclaimer says otherwise. Prominent names, visual markers, check marks, and comment-like layouts can carry more practical weight than fine-print qualification.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →The Verge reported that, in Google Docs, some suggestions could resemble comments from real users. It also found source links that were inaccurate, unrelated, spammy, or merely archived copies. (The Verge)
A disclaimer may be technically clear while the overall product experience remains easy to misunderstand. The relevant question for users is not only what the disclaimer literally says, but what the interface leads a reasonable person to believe.
Style imitation is not expert judgment
An AI system may analyze an author’s published prose, but that does not establish how the author would edit someone else’s document. Published writing generally does not reveal:
- How the person would review a student essay, news article, business memo, or scientific paper.
- Which changes they would prioritize in a specific context.
- What facts they would verify or what sources they would reject.
- How they would respond to the writer’s goals, audience, or assignment.
- Their private editorial standards or tolerance for ambiguity, humor, jargon, repetition, and unconventional style.
- Whether they would recommend changing the passage at all.
A recognizable voice is not the same thing as subject-matter expertise. It also does not prove factual accuracy, sound reasoning, current knowledge, or agreement with the named person’s actual views.
This distinction is especially important for students and professionals. A suggestion that sounds like a famous writer may still be generic, contextually wrong, or inconsistent with that person’s known editorial preferences.
Rank #3
The feature reportedly had practical reliability problems
The reported problems went beyond consent and representation. The Verge described crashes, inaccurate or outdated descriptions of experts, unreliable source links, and suggestions that were generic or questionable in context. (The Verge)
Those failures weaken the strongest possible defense of the product. Even if a company believes it may study public writing, a useful expert-style tool would still need to identify its sources accurately, explain the limits of its simulation, and produce advice that can be checked.
Users should never treat a named expert or linked source as proof that a recommendation is authoritative. Open every citation, confirm the author and title, check that the source is genuine, and verify that it actually supports the advice.
Was it plagiarism, impersonation, or copyright infringement?
These are separate questions, and the available reporting does not establish a final legal answer to all of them.
- Plagiarism: The reporting does not establish that every generated suggestion copied protected text. It would be inaccurate to label the entire feature plagiarism without examining particular outputs.
- Identity or persona misuse: Living people were named and represented in the product without permission, according to reporting from The Verge and WIRED. Critics characterized this as simulated or fabricated expert authority.
- Copyright: The feature appears to have drawn on publicly available published work, but the legal treatment of using and modeling that work depends on factors such as jurisdiction, licensing, purpose, and model behavior. Publicly available does not automatically answer every copyright question.
- Consumer protection: The central product-language concern is whether users could misunderstand “expert review” and believe that a real expert had supplied the feedback.
- Academic integrity: Students could mistake the feature for legitimate expert or instructor-like evaluation, even though it was only automated assistance.
The Verge reported that journalist Julia Angwin filed a class-action lawsuit after learning that she and others were represented in the feature. That report describes an allegation, not a final finding that Grammarly was legally liable. (The Verge)
The use of deceased figures raises additional questions. They cannot consent or opt out, and their estates may or may not control relevant rights. More fundamentally, attributing modern advice to someone who never encountered a user’s context is inherently speculative. That does not prove every such use is unlawful, but it makes the presentation less defensible as genuine expertise.
What Grammarly and Superhuman said
Superhuman’s position was that the feature did not claim direct participation or endorsement. The company said the named people were included because their published work was publicly available and widely cited, and that the product was intended to surface influential perspectives and relevant work.
That explanation fairly describes the company’s stated intent. It does not resolve the product-design criticism: a commercial service could use a person’s name and reputation to make automated recommendations more attractive even when that person had no relationship with the service.
Superhuman CEO Shishir Mehrotra later acknowledged that Expert Review “was not a good feature” and said it was not good for either experts or users. The Verge reported that the company first offered an email-based opt-out and then removed the feature entirely. (The Verge Decoder interview)
Is Grammarly’s Expert Review still available?
According to the reporting reviewed, no. As of August 18, 2026, Expert Review had been removed rather than merely left in a controversial but active state. Grammarly’s public homepage currently highlights other functions, including proofreading, rewriting, tone suggestions, Humanizer, AI Detector, and reader reactions, but does not list Expert Review. (Grammarly)
The timing and cause should not be overstated. The feature drew public criticism in March 2026 after launching in August 2025. Reporting covered the opt-out and subsequent removal, while accounts differed over whether the lawsuit caused the removal or followed a decision already made. The safe conclusion is that the feature is no longer presented as a current Grammarly offering.
Recommended Free Tools
What should you use instead?
For grammar and clarity
A clearly labeled automated writing assistant can be useful for spelling, grammar, readability, and basic rewriting. Options include Grammarly, LanguageTool, ProWritingAid, and Hemingway Editor.
Best Value
These tools have different features and may use AI or other automated language technology. The relevant distinction is transparency: automated suggestions are not presented as feedback personally supplied by a named writer or scholar.
Grammarly’s public pricing display showed a Free plan at $0 and Pro at $12 per member per month when billed annually or $30 monthly on August 18, 2026. Prices and plan features can change, so check the company’s current page before paying.
For substantive editorial judgment
Use an identifiable human when you need someone to understand the document’s purpose, audience, evidence, and consequences. Depending on the project, that could be:
- A developmental editor or copyeditor with relevant experience.
- A subject-matter specialist who can verify technical claims.
- A university writing center or qualified tutor.
- An instructor who understands the assignment rubric.
- A professional editor found through a trusted network or a marketplace such as Reedsy.
- A human editing service such as Scribendi, after confirming the editor’s qualifications for specialized work.
A genuine expert review should identify the reviewer, explain their qualifications and scope, confirm that they directly handled the document, and provide a way to ask follow-up questions. It should also make clear whether the reviewer is being paid and how conflicts of interest are handled.
For academic work
Follow the instructor’s rubric and academic-integrity rules first. A writing center, qualified tutor, citation-verification tool, or human subject specialist may help, but an AI assistant should not be treated as an instructor, peer reviewer, or authority on the subject.
What this means for Grammarly customers
The Expert Review controversy does not mean Grammarly’s ordinary proofreading and rewriting tools are identical to the discontinued feature. A user may still find automated grammar or clarity suggestions worthwhile, particularly for low-stakes writing.
But paying for Grammarly does not buy accountable human editorial judgment. No automated writing assistant should substitute for a qualified reviewer when the document involves high-stakes academic, legal, medical, financial, scientific, or professional claims.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →The key consumer distinction is simple: AI-generated feedback can be useful when it is honestly labeled as AI-generated feedback. The risk rises when that feedback borrows the authority of a real person who was not there.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
