The best AI writing tools compared: tools, with what each one does, what it costs and who it suits.
Writing was one of the first jobs people handed to AI, and it remains one of the most common. Students use AI to outline essays and check grammar, professionals use it to draft emails and reports, marketers use it for campaign copy and researchers use it to tighten abstracts. Some of that use saves hours. Some of it produces bland, error filled text that takes longer to fix than to write from scratch. And some of it crosses lines that schools, employers and publishers care about.
This section looks at AI writing tools without the hype: what they do well, what they do badly, what research says about their impact, how to use them without losing your voice and how to stay within the rules. Our guide to AI writing tools compares types of tool in more detail. Descriptions of products are based on companies' own materials, and features change frequently.
The most cited experiment on AI and writing comes from Shakked Noy and Whitney Zhang, published in Science in 2023. They gave 453 college educated professionals occupation specific writing tasks and randomly gave half access to ChatGPT. Average time taken fell by 40 percent and rated quality rose by 18 percent, with the biggest gains for weaker writers. The tasks, such as cover letters and short reports, did not demand precise factual accuracy or deep knowledge of a specific context, which is exactly where AI writing tends to struggle.
The broader research on AI productivity adds a warning. In a large experiment with Boston Consulting Group consultants, Fabrizio Dell'Acqua and colleagues found that AI helped greatly on tasks within its capabilities but made people less likely to get the right answer on a task outside them. In writing, that means AI is strong on structure, tone and fluency, and weaker on accuracy, original argument and specialist knowledge.
| Task | How AI tends to do | Why |
|---|---|---|
| Routine emails and messages | Very well | Familiar formats, low stakes |
| Outlines and structure | Very well | Patterns are common across texts |
| Editing for clarity and grammar | Well | Surface level improvements are reliable |
| Summaries of provided text | Well, with errors | Can omit or distort key points |
| Persuasive or creative drafts | Fluent but generic | Tends toward safe, familiar phrasing |
| Factual or research writing | Risky | Invented facts and citations |
| Original arguments | Weak | Repeats common positions |
| Kind of help | What it looks like | Best practice |
|---|---|---|
| Brainstorming | Ideas, angles, questions, counterarguments | Use as prompts for your own thinking |
| Outlining | Structure for an essay, report or post | Adapt the outline to your argument |
| Drafting | First versions of routine text | Rewrite in your own voice |
| Editing | Grammar, clarity, concision, tone | Review each change |
| Rewriting | Changing audience, length or format | Check meaning is preserved |
| Feedback | Comments on structure, logic, clarity | Treat as one reader's view |
| Summarising | Shorter versions of long texts | Check against the original |
The biggest complaint about AI writing is sameness. Text written largely by AI tends to use similar structures, safe vocabulary and smooth but empty phrasing. Readers increasingly notice. To keep your voice:
AI writing tools can invent facts, statistics, quotes and references that look real. This is one of the most serious risks, especially in academic and professional writing. Never include a fact or citation you have not checked against the original source. When you need research support, use tools that search and link to sources, and read those sources yourself. If an AI suggests a reference, find the actual paper or page before citing it.
Universities and schools have very different AI policies, and they vary by course and assignment. Some ban AI for graded work, some allow it for brainstorming and editing with disclosure, and some build it into assignments. Before using AI on coursework, read the syllabus and ask your instructor if anything is unclear. Keep notes of how you used AI, such as prompts and drafts, in case you need to explain your process. Using AI to produce work you submit as entirely your own, where that is not allowed, is academic misconduct.
Many institutions have tried AI detection tools, and many students worry about being falsely accused. Research gives reason for caution. A 2023 study by Weixin Liang and colleagues at Stanford, published in Patterns, tested seven widely used detectors on essays written by non native English speakers for the TOEFL exam and found that more than half were wrongly labelled as AI generated, an average false positive rate of about 61 percent, while essays by U.S. eighth graders were classified almost perfectly. The authors also showed that simple prompting could help AI text evade detection. Detectors should not be used as the sole basis for accusations, and writers should keep drafts and notes that document their process.
Employers and publishers increasingly have their own rules. Some require disclosure of AI assistance, some prohibit AI for certain content, and many forbid pasting confidential information into public tools. Journalists, lawyers, scientists and others work under professional standards that AI does not change. Know the rules where you work, and when in doubt, disclose and verify.
AI is useful for adapting the same content to different readers. A researcher can ask for a plain language summary of a paper for the public, a short version for a grant report and a technical abstract for specialists, then edit each. A manager can turn detailed notes into an executive summary and a team update. The key is to start from accurate source material you wrote or checked, and to review each version, because simplification can introduce errors and nuance can disappear.
| Step | You | AI |
|---|---|---|
| Purpose and audience | Decide what you want to say and to whom | Ask clarifying questions |
| Ideas | Note your key points and evidence | Suggest angles and counterarguments |
| Structure | Choose the outline | Propose alternative structures |
| Draft | Write the core argument in your words | Help with stuck sections or routine parts |
| Revise | Rewrite for voice and accuracy | Flag unclear passages, suggest cuts |
| Check | Verify every fact and citation | Never the final authority on facts |
| Polish | Final read aloud | Grammar and consistency check |
AI writing tools can be especially helpful for people writing in a language that is not their first. They can explain grammar, suggest more natural phrasing and help with formal conventions. They can also smooth away a writer's own voice and, in academic settings, cross lines about original work. The Liang study is a reminder that non native writers are particularly vulnerable to false accusations from detectors. Use AI to learn and improve, keep drafts and follow your institution's rules.
Vague requests produce vague writing. Better requests give the AI a role, the audience, the purpose and constraints. Instead of asking it to write an email about the deadline, explain that you need a polite but firm email to a client who has missed two deadlines, that you want to keep the relationship, that it should be under 150 words and that it must propose a new date. For feedback, ask specific questions: which paragraph is weakest and why, where would a sceptical reader push back, which sentences could be cut without losing meaning. Specific questions produce useful answers; general requests to improve this produce generic rewrites.
Some writing is better done without AI. Personal statements and reflective essays are valued precisely because they are personal. Condolence messages, apologies and other emotionally significant notes lose meaning when they sound generated. Writing that requires confidential information should not go into tools that are not approved for it. And when the purpose of an assignment is to practise a skill, outsourcing it defeats the point. Knowing when to close the AI tool is part of using it well.
These illustrative examples are not real people.
A sophomore writes her essay draft herself, then asks an AI assistant which paragraphs are unclear and checks grammar, as allowed by her course policy.
A program manager drafts routine status updates with AI from bullet points, then edits for accuracy and tone.
A PhD student writing in English as a second language asks AI to explain why certain phrases sound unnatural and keeps every draft as a record.
These are recurring themes from user discussions, not quotations from individuals.
The Review NYU is an independent publication. We base these guides on hands-on testing, companies' own documentation and published research. We label company claims as such, do not accept payment for placement and never invent statistics, reviews or credentials. Read about our editorial standards.
On routine tasks, research has found faster writing and higher rated quality. On factual, specialist or original work, AI is less reliable and needs careful checking.
Only within your course and university policies. Many allow brainstorming or editing with disclosure; some prohibit AI for graded work.
Research has found them unreliable and biased, particularly against non native English writers. They should not be the sole basis for accusations.
Start with your own ideas and draft, ask for feedback rather than rewrites, give samples of your style and edit until it sounds like you.
Language models generate plausible text, not verified facts. Always find and read the original source before citing.
Where your school, employer or publisher requires it, yes. When in doubt, disclosure is the safer choice.
Yes, it is generally reliable for grammar and clarity, but review each change to make sure meaning is preserved.
Yes, by explaining grammar and suggesting natural phrasing, but follow rules on original work and keep your drafts.
For personal statements, emotionally significant messages, confidential material in unapproved tools and assignments meant to build your own writing skills.
Ask specific questions, such as which paragraph is weakest and why, where a sceptical reader would push back and which sentences could be cut. Specific questions produce more useful feedback than a general request to improve the text.
Yes. It can turn the same content into a plain language summary, an executive summary or a technical version. Start from accurate material and check each version, because simplifying can introduce errors.
This guide is independent and editorially reviewed. Company claims are labelled as such, and we never invent statistics, reviews or credentials. Read about The Review, our editorial standards, or contact us.