Grammarly has been the default writing assistant for so long that it is easy to forget how good the core is. Catching genuine grammatical errors, spotting the comma that changes the meaning, flagging the sentence you rewrote twice and left broken in the middle. On that job it remains more reliable than asking a general purpose model to proofread, because the task is narrow, the model is tuned for it, and it runs inline everywhere you type rather than requiring you to paste text somewhere else. For someone writing professional email all day, that is worth the subscription on its own.
The writing handbook is an underrated free resource. Clear explanations of grammar and usage questions, written for people who want an answer rather than a lecture, and useful regardless of whether you use the product. I would happily recommend it to someone who never installs the software. Clarity suggestions are where judgement is required and where the material offers least help.
Grammarly will tell you a sentence is wordy and offer a shorter one, and it is right perhaps two thirds of the time. The other third, the length was doing something, the rhythm mattered, the qualification was load bearing, or the construction was the thing that made the sentence sound like a person wrote it. Accept everything and your writing becomes clean, correct and interchangeable. The single most valuable piece of guidance the company could publish is on when to click dismiss, and it does not exist, which is unsurprising since the suggestion count is the product's visible value.
The generative features are the awkward part. Rewriting, drafting and replying are now available free and better inside the tools people already have open, and Grammarly's versions are competent without being a reason to choose it. Adding them has diluted a very clear identity, which was the thing that catches your mistakes, into a vaguer one, which is another AI writing assistant among many. I would rather they had defended the narrow position, and the material's attempt to present the bundle as coherent is not convincing.
Business features are genuinely useful and get less attention than the consumer story. Enforcing a house style across a team, catching the terminology that should be capitalised, keeping tone consistent across people who have never met. That is a real organisational problem with few good solutions and the material handles it competently. The corporate rebrand, where the parent company became Superhuman in late 2025 while the Grammarly product kept its name, adds a layer of confusion that the help material does not clear up well.
Users encountering Superhuman Go and the associated suite alongside familiar Grammarly branding get little help understanding what they have or what they are being upsold. Three point four. The core remains excellent and is still the right recommendation for most people, marked down for a generative expansion that adds little, a corporate identity the documentation has not caught up with, and complete silence on the one skill that separates good use from bad, which is knowing which suggestions to refuse.