It’s 11:40 p.m., the portal closes at midnight, and you’ve just read a post saying your university runs every essay through an AI detector. You wrote this one yourself. You also used a grammar tool on the introduction and asked a chatbot to explain a theory you didn’t understand in week four. Now you’re wondering what a detector will make of all that.
I’ve helped a lot of students through this exact moment, and the good news is that a calm fifteen-minute check, done a day or two earlier, removes most of the panic. The steps below are the ones I’d follow for any essay that matters.
Start with your course’s AI policy
Before you check anything, reread the rules. Policies vary a lot between institutions, departments and even individual modules. Some courses ban generative AI entirely. Some allow it for brainstorming or grammar but not drafting. Others allow it if you disclose it in a short note at the end.
Look for three answers:
- What uses are allowed (outlining, grammar, feedback, drafting)?
- Do you need to disclose AI use, and in what format?
- Does the course use a detection tool, and does it say how results are handled?
If the policy is unclear, email your tutor. A one-line question like “Is it okay to use a grammar checker on my final draft, or should I disclose it?” protects you far more than any scan.
Run a check on the full essay
Once you know the rules, paste the whole essay into an ai Checker rather than testing one paragraph at a time. Detectors look for patterns across a text, and short samples produce erratic scores. A 150-word chunk might read as 80% AI in one run and 20% in the next. Full essays give steadier results.
I usually point students to Quillbot’s detector for this step because it separates writing into AI-generated, human-written, and human-written but AI-refined. That third category is useful for students. If you used a grammar tool or a paraphraser on a couple of sentences, you’ll often see them land there rather than being lumped in with fully generated text. The tool highlights sentences, so you can see exactly which lines caught its attention.
Do the check at least a day before the deadline. At 11:40 p.m. you don’t have time to do anything useful with the result.
Read the highlights, not just the number
A headline percentage is the least interesting part of the report. What you want are the flagged sentences. Go through them one by one and ask what they have in common.
In my experience, flagged student writing tends to share a few habits:
- Generic openers (“Throughout history, humans have always…”)
- Balanced summary sentences that say nothing specific
- Lists of three abstract nouns (“efficiency, sustainability and innovation”)
- Transitions that sound like a template (“Furthermore,” “Moreover,” “Additionally,” one after another)
None of these mean you used AI. They mean your writing, in those spots, is predictable. And predictable writing is also weaker writing, so fixing it improves the essay whatever the detector says.
What the score actually tells you
Students often read a detector result as “the percentage of my essay written by AI.” That isn’t how most tools work. The number is usually an estimate of how likely the text is to be machine-generated, based on patterns the tool was trained to recognise. A result of 30% doesn’t mean 30% of your sentences came from ChatGPT. It means the tool is somewhat unsure.
A few practical rules help:
- A low score isn’t a guarantee. Your institution may use a different tool with different settings.
- A high score isn’t a verdict. It’s a reason to look closely at the flagged lines.
- Mixed results are normal. If you drafted yourself and used a grammar tool on some sentences, a tool that separates AI-refined text from fully generated text will often show exactly that mix.
Save a copy or screenshot of the report with the date. If questions come up later, a record showing you checked your work in good faith and revised it is useful context.
Revise flagged passages in your own voice
Take a typical sentence from a sociology essay on social media, the kind that shows up in almost every first draft on the topic.
Before:
Social media has had a significant impact on young people in many different ways, affecting their mental health, relationships and overall wellbeing.
After:
Most of the people I interviewed checked Instagram before getting out of bed, and four of the six said the first thing they felt was comparison, not connection.
The second version uses the student’s own research, names a specific behaviour and makes a claim someone could argue with. A detector has almost nothing predictable to latch onto, and the marker gets a sharper point.
When you revise, go back to your notes, your sources and your own reasoning. Add the detail only you would know: the specific case study, the page number you’re responding to, the reason you disagree with an author. Resist the urge to run the sentence through a rewriter and hope for a lower score, because that tends to produce stiffer text and can itself get flagged as AI-refined.
Here’s a second example, this time from a history essay’s conclusion:
Overall, the causes of the First World War were complex and involved a combination of political, economic and social factors.
That sentence could close any essay on the topic, which is exactly why it reads as generic. A revision that pulls from the student’s own argument might look like this:
The alliance system didn’t make war inevitable, but it turned a regional crisis in the Balkans into a continental one within about a week, and that speed is what the essay has tried to explain.
It’s longer, it commits to a position and it points back to the argument the student actually made. It also gives the marker something to agree or disagree with, which is what a conclusion is for. If you can’t write a sentence like that yet, the problem is usually the argument rather than the wording, and that’s worth fixing before any detector sees the essay.
Keep proof of your process
This is the step most students skip, and it’s the one that saves you if something goes wrong. Detection tools make mistakes. Formal writing, writing by non-native English speakers and heavily structured assignments are all more likely to be misread. Your best defence is being able to show how the essay came together.
Practical ways to build that record:
- Write in Google Docs or Word with version history on. Google Docs saves revisions automatically. In Word, saving to OneDrive keeps a version history you can open later.
- Keep your notes and outlines. Photos of handwritten notes count. So does a messy bullet list in your notes app.
- Save your sources. A folder of PDFs with your highlights shows you read them.
- Don’t paste the finished essay in from elsewhere. If you drafted in one app and moved it to another, keep the original file too, because a single giant paste in your version history looks odd even when it’s innocent.
If you did use AI in an allowed way, keep that record as well. A screenshot of the chat where you asked for an explanation of a concept, plus a line in your disclosure note, is much easier to discuss than a vague memory.
Write a clear disclosure note (if your course asks for one)
Plenty of courses now allow limited AI use as long as you say what you did. A disclosure note doesn’t need to be long, and a specific one is much better than a vague one. Something like this works for most assignments:
AI use statement: I used ChatGPT to explain the difference between structural and cultural theories of deviance while revising my notes in week 4. I used a grammar checker on the final draft for spelling and punctuation. All arguments, structure, analysis and interview data are my own. No AI-generated text was pasted into this essay.
Adjust it to match what actually happened and whatever format your department asks for. If your course has its own template or requires a specific citation style for AI tools, use that instead. The point is that anyone reading your work can see exactly where a tool helped and where it didn’t.
Check the rest of the essay while you’re at it
An AI check is only one part of a pre-submission review. While the essay is open, spend ten minutes on the things markers notice first. Make sure every quotation and paraphrase has a citation and every citation appears in your reference list. Run a plagiarism check if your institution gives you access to one, since accidental close paraphrasing of a source is a far more common problem than AI misuse. Confirm the word count sits inside the limit, and check the file format and naming rules in the assignment brief. None of this is glamorous, but it’s where easy marks disappear.
What to do if your own writing gets flagged
First, breathe. A detection score on its own isn’t proof of anything, and most institutions say so in their own guidance.
Then:
- Ask to see the report. You’re entitled to know which passages were flagged and what tool was used.
- Bring your evidence. Version history, notes, outlines, source files. Walk through when you wrote each section.
- Offer to talk through the argument. Explaining why you structured the essay a certain way, or why you chose a particular source, shows ownership far better than any document.
- Stay factual. Anger is understandable, but a calm timeline is far more persuasive, and the person reviewing your case usually wants to get it right.
- Ask about the appeal process if the first conversation doesn’t resolve it. Student unions and academic advisers can help you prepare.
A student who walks into that meeting, opens a laptop, clicks on version history and scrolls through three weeks of edits, deleted paragraphs and scribbled comments to themselves (the “fix this tomorrow” notes everyone leaves) is in a far stronger position than the report alone suggests, because that kind of messy trail is very hard to fake.
Mistakes that cause trouble
Most problems I’ve seen around AI checks come from a handful of avoidable habits.
Running a rewriter over flagged sentences. It feels like a quick fix, but paraphrasing tools can make writing stiffer, change your meaning and sometimes land the text in the “AI-refined” category anyway. Rewrite by hand, using your own notes.
Checking only the introduction. Introductions and conclusions get the most generic phrasing, but body paragraphs matter too. Check the whole thing.
Drafting somewhere without history. Writing in a plain text editor or an app that doesn’t keep versions leaves you with no record of your process. Switch to a tool that saves revisions before you start, not the night before the deadline.
Leaving the disclosure out because it feels awkward. An honest one-paragraph note is almost always treated better than undisclosed use that comes to light later.
Assuming every detector agrees. A clean result from one free tool says little about what your institution’s tool will show. Focus on good writing and good records, not on chasing a particular number.
A short pre-submission routine
Two days before the deadline:
- Reread the AI policy and note anything you need to disclose.
- Run the full essay through an ai Checker and save a copy of the report.
- Revise any flagged sentences using your own notes and examples.
- Check citations and references.
- Export a PDF of your version history, or at least confirm it’s there.
Submission day:
- Read the essay aloud once. You’ll catch clunky sentences a detector would also notice.
- Add your disclosure note if needed.
- Submit early.
FAQ
Can I see the report my university’s detector produces? Often not before you submit, since tools like Turnitin are usually set up for instructors. You can ask your tutor what they see and how they use it, and you can request the report if your work is questioned.
Does using a grammar checker count as AI use? It depends on your course. Many policies allow spelling and grammar tools, but some ask you to mention them. When in doubt, disclose it.
Should I change my writing style to avoid being flagged? Only in ways that make your writing better: more specific examples, clearer claims, fewer filler transitions. Don’t make your essay worse to please a tool.
One last thing before you hit submit
Detectors are one input, and a fairly blunt one. What protects you is honest use of whatever tools your course allows, a clear record of your drafts, and writing that sounds like someone who actually did the reading. If your essay has your examples and your arguments in it, the score matters a lot less than you fear at 11:40 p.m.
