I’d been staring at the same blank document for almost an hour.
The essay was supposed to be 1,500 words. The topic wasn’t even that difficult. I had the assignment instructions open, a few notes scattered around, and exactly zero sentences I was happy with.
That’s the point where AI starts looking very tempting.
Type the topic into a chatbot, wait ten seconds, and suddenly there’s a polished essay sitting in front of you.
Except there’s a problem.
A complete AI-generated essay can look impressive while being completely useless to you. It can miss the professor’s actual question, invent a citation, make an argument you don’t agree with, or sound so generic that it feels like it was assembled from every school essay ever written.
I’ve found that the better use of AI is much less glamorous. Use it to get unstuck. Use it to challenge your argument. Use it to explain something you don't understand. Use it to clean up a rough paragraph you actually wrote.
After trying different approaches, ChatGPT is the AI writing tool I’d start with for most essay work, particularly if I wanted one tool that could help with brainstorming, outlining, research, drafting, revision, and working with uploaded files.
But I wouldn't automatically tell everyone to pay for it. And I definitely wouldn't tell anyone to paste an assignment into an AI tool and submit whatever comes back.
There’s a much better way to use these tools.
The real comparison
There are several AI assistants worth considering for essay writing right now. The names you'll hear most often are ChatGPT, Claude, and Gemini.
They overlap a lot. All three can help brainstorm ideas, explain difficult subjects, rewrite paragraphs, create outlines, and critique drafts. The differences become more obvious when you give them a real assignment and ask them to work through several stages rather than simply saying, “Write my essay.”
Here’s how I’d think about them:
| Tool | Where I’d use it | Cost consideration | What stands out |
| --------------------------- | ---------------------------------------------------------------- | ------------------------------ | --------------------------------------------------------------- |
| ChatGPT | General essay work, brainstorming, outlining, revision, research | Free option; Plus is $20/month | Broad set of writing, file, research, and reasoning tools |
| Claude | Draft refinement, long documents, natural-sounding prose | Free and paid options vary | Strong long-form writing and instruction following |
| Gemini | Research-heavy work and Google ecosystem users | Free and paid options vary | Useful if your workflow already revolves around Google services |
| Dedicated grammar tools | Final proofreading | Usually separate subscription | Grammar, clarity, spelling, and style cleanup |
ChatGPT currently has a free version, while ChatGPT Plus is listed at $20 per month. OpenAI says Plus provides broader model and tool access, including higher limits, file uploads and analysis, and expanded research capabilities where available.
That doesn't mean Plus automatically makes your essays better.
The quality of your instructions still matters enormously.
Why I pick ChatGPT
The biggest advantage isn't simply that ChatGPT can write.
Pretty much every serious AI assistant can write.
What I like more is that it can stay involved through the whole writing process.
Let's say the assignment is:
“Discuss the effects of social media on political communication.”
I wouldn't start with:
“Write a 2,000-word essay about social media and political communication.”
That's asking for trouble.
I'd start by giving it the assignment requirements and asking it to help break down what the professor is actually asking.
Then I'd ask for possible arguments.
Then I'd pick one.
Then I'd ask for an outline.
Then I'd find sources.
Then I'd write at least some of the actual material myself.
After that, AI becomes useful again.
Give it the draft and ask:
“Where is my argument weak?”
That's a much more interesting question than “Make this essay better.”
ChatGPT is particularly useful for that back-and-forth process because it can handle different kinds of tasks in the same conversation. Its current capabilities include drafting, rewriting, summarizing, explaining concepts, reasoning through problems, and analyzing uploaded files. Its web-search capability can also be used for current or source-backed information when available.
That combination is what makes it practical for essay work.
Don't start with writing
This was the mistake I’d avoid first.
People often think the first step is generating an introduction.
It isn't.
The first step is figuring out what you're actually trying to argue.
A weak essay usually doesn't become strong because the AI gave it prettier sentences. If the central argument is vague, the entire paper will wander around regardless of how polished the prose sounds.
Instead, give the AI your assignment prompt and ask it to identify:
* the actual question being asked
* the required subject
* important limitations
* possible interpretations
* terms that need defining
* evidence you'll probably need
* questions your professor might expect you to answer
That gives you something to work with.
For example, an assignment asking you to “evaluate” something is different from one asking you to “describe” it.
An assignment asking you to “compare” two theories isn't necessarily asking which theory is better.
Those little distinctions can completely change the structure of the essay.
AI is surprisingly useful at spotting them.
Build the argument first
Once you understand the assignment, I’d move to the thesis.
Not the final polished thesis. Just the basic idea.
Suppose you're writing about remote work.
Instead of asking:
“Give me a thesis about remote work.”
Try something like:
“My argument is that remote work improves flexibility for employees but can create communication problems in organizations that haven't changed their management practices. Tell me where this argument is too broad or unsupported.”
That gives the AI something to attack.
And that's what you want.
Ask it to disagree with you.
Ask:
“What would a professor object to here?”
Or:
“What evidence would I need to prove this?”
Or:
“What is the strongest argument against my position?”
Those prompts are far more useful than endlessly asking for more paragraphs.
A good essay needs tension. There needs to be something to discuss, compare, question, explain, or defend.
AI can help you find that tension before you've spent three hours writing around it.
Where Claude fits
I wouldn't ignore Claude.
Claude is particularly worth trying if your biggest problem is long-form editing. Anthropic describes its current models as capable of long-context reasoning and knowledge work, and Claude Sonnet 4.6 has a 1-million-token context window in beta.
That matters when you're dealing with a large collection of notes or a long draft.
For example, imagine you've written 8,000 words of notes for a research project but your final assignment needs to be 2,500 words.
That's not really a “write an essay” problem.
It's an organization problem.
You need to figure out:
Which ideas belong?
Which ones are duplicates?
Which evidence supports the thesis?
Which sections wander off-topic?
Where are the transitions weak?
A large-context model can be useful for this kind of editing because you're able to give it more material and ask it to work across the document rather than feeding it one paragraph at a time.
Claude's model lineup and pricing change, so I'd check Anthropic's current plan information before paying specifically for essay work. Its official documentation currently lists free and paid access depending on the model and plan.
I wouldn't subscribe just because someone online says one model is “the smartest.”
Your actual assignment matters more.
Where Gemini makes sense
Gemini is another option I'd keep around, especially for students already living inside Google's ecosystem.
If your research notes, documents, email, and other school work are already organized around Google products, having an AI assistant that fits naturally into that environment can be convenient.
But convenience and writing quality aren't exactly the same thing.
That's why I wouldn't pick an AI based on a leaderboard alone.
Try your own assignment.
Give each tool the same prompt.
Give each one the same rough thesis.
Then compare the results.
You'll notice differences in how they structure arguments, handle instructions, explain concepts, and rewrite your prose.
That's a more useful test than somebody else's “Top 10 AI tools” article.
The prompt that changed things
The biggest improvement doesn't actually require a better AI model.
It requires better instructions.
Here's the sort of prompt I'd use when revising an essay:
“I'm going to give you my assignment instructions and my draft. Don't rewrite the essay yet. First identify the three weakest parts of my argument, explain why they're weak, and tell me what evidence or reasoning would strengthen each one. Keep my original position unless there is a clear logical problem.”
Notice what's missing.
“Make it sound smarter.”
That's one of my least favorite AI writing prompts.
It often produces exactly what you don't want: inflated language.
Suddenly “use” becomes “utilize.”
“Shows” becomes “demonstrates.”
Every sentence gets longer.
The essay sounds academic in the same way a rental apartment looks fancy after somebody puts three fake plants in it.
Clear writing is better.
The mistake I wouldn't repeat
Here's the failure mode that catches people.
You ask an AI to produce an essay with sources.
It gives you eight beautifully formatted references.
Everything looks legitimate.
Then you check them.
One source doesn't exist.
Another source exists but doesn't say what the paragraph claims.
A third citation is real, but the publication date or title is wrong.
That's not a minor problem.
It's the kind of mistake that can undermine an otherwise decent assignment.
AI systems can produce incorrect information, and even AI-generated text that sounds extremely confident still needs verification. OpenAI itself warns that model outputs can be inaccurate or misleading.
So I wouldn't treat AI-generated citations as automatically trustworthy.
If the assignment requires academic sources, find the actual sources and read them.
Use AI to help you understand a paper.
Use it to compare two arguments.
Use it to explain complicated terminology.
Use it to organize notes.
But verify the source yourself.
That's especially important for research papers where your grade depends partly on the quality of your evidence.
The better workflow
Here's the workflow I'd use if I had an essay due this week.
First, paste the assignment instructions into the AI.
Don't summarize them yourself. Give it the actual wording.
Ask it to explain what the assignment requires and identify anything ambiguous.
Second, brainstorm three or four possible arguments.
Don't choose the first suggestion automatically.
Third, pick one argument and try to break it.
Ask the AI for the strongest objections.
Fourth, create an outline.
Keep the outline relatively simple:
Introduction.
Main argument.
Evidence.
Counterargument.
Response.
Additional evidence or analysis.
Conclusion.
The exact structure will depend on the assignment, of course.
Fifth, research the actual evidence.
This is where you should leave the AI bubble and look at legitimate sources.
Sixth, write your first draft.
It can be ugly.
That's fine.
Seventh, give the draft back to the AI and ask for criticism rather than a rewrite.
This is the stage where AI becomes genuinely useful.
Ask it to identify:
* unsupported claims
* repetitive points
* unclear sentences
* weak transitions
* logical jumps
* places where evidence doesn't actually support the claim
* sections that don't answer the assignment
Then fix those issues yourself.
Finally, use AI for proofreading.
Grammar.
Clarity.
Awkward sentences.
Wordiness.
That's where I think it earns its keep.
What about $20?
If you're a student, $20 a month isn't nothing.
And I don't think everyone needs ChatGPT Plus.
OpenAI currently lists ChatGPT Plus at $20 per month, while also offering a free ChatGPT tier. OpenAI says the paid plan provides broader access and higher limits, along with additional capabilities such as file uploads, analysis, and research tools.
I'd stay with free access if you're writing one or two ordinary assignments.
I'd consider paying for a month if you're doing a lot of research, working with long documents, uploading multiple files, or using AI heavily throughout a semester.
And I'd cancel when you don't need it.
There's no prize for maintaining five AI subscriptions.
If you're comparing services, remember that advertised prices and included limits can change. OpenAI also says some pricing is localized by country and payment method, so the amount displayed at checkout may differ from the headline U.S. price.
The “AI detector” problem
This is another area where I'd be careful.
Students sometimes become obsessed with making AI-written text “undetectable.”
I wouldn't build your workflow around that.
First, an essay should actually represent your own work and comply with your school's rules.
Second, AI systems themselves aren't reliable judges of whether a particular piece of text was generated by AI. OpenAI explicitly says ChatGPT cannot reliably determine whether it generated a particular piece of writing.
So instead of playing games with detection software, write something you understand.
Keep your notes.
Keep your sources.
Keep earlier drafts.
Be able to explain your argument.
That's a much healthier way to approach academic writing.
What if AI makes writing lazy?
Yeah. This is the obvious objection, and it's a fair one.
If you let AI do every step, you'll probably get worse at some of those steps.
There's a difference between using a calculator and never learning basic arithmetic.
The same thing applies here.
If AI always creates your thesis, you'll get less practice developing arguments.
If it always writes your first draft, you'll get less practice organizing ideas.
If it always rewrites your paragraphs, you'll never develop much of an instinct for good sentences.
That's why I like using AI as a demanding editor rather than a ghostwriter.
Make your argument.
Then make the AI fight with it.
Ask it where you're wrong.
Ask what you're missing.
Ask why a reader might reject your conclusion.
That still requires you to think.
And honestly, that's the part of essay writing that matters.
A simple test
If you're unsure which AI to use, don't spend the afternoon reading comparison articles.
Run a test.
Take one real assignment.
Remove any private information you shouldn't upload.
Give the same instructions to ChatGPT, Claude, and Gemini.
Ask each one to:
Then compare the responses.
Don't ask which one “sounds smartest.”
Ask which one helps you think.
That's the distinction I care about.
A beautiful paragraph that doesn't help you understand your subject isn't much use.
A tool that catches a weakness in your argument before your professor does? That's useful.
The one thing I'd change
If I were starting an essay from scratch today, I wouldn't ask an AI to write the essay.
I'd give it the assignment sheet.
I'd make it explain the task.
I'd brainstorm with it.
I'd develop my own thesis.
I'd research real sources.
I'd write the rough draft.
Then I'd use AI aggressively as an editor.
That sounds slower than “write my essay for me,” but it usually produces something much more valuable: a paper you actually understand.
And that's ultimately what you're paying for in an education.
Not 1,500 words.
The ability to take an idea, investigate it, defend it, recognize its weaknesses, and explain it clearly.
If you want one AI to start with, I'd start with ChatGPT. Its combination of conversational writing help, reasoning, file analysis, and research-related tools makes it a practical all-purpose choice for essay work.
But don't marry yourself to one tool.
Try Claude when you're dealing with a large draft or want another perspective on your prose. Try Gemini if its ecosystem fits the way you already work. And keep your actual brain involved throughout the process.
The best AI for writing essays isn't the one that produces the most impressive essay with the least effort.
It's the one that helps you produce your best thinking with less wasted effort.
That's a much better deal.