A few weeks into most semesters, someone emails a lecturer at 11:40 p.m. with a version of the same message: “I didn’t use AI, but the checker says 68%. What do I do?” Sometimes they did use it a bit, to tidy a paragraph or reword an awkward introduction, and they genuinely didn’t think that counted. Sometimes they used nothing at all. Either way, they’re now trying to prove a negative at midnight with a deadline behind them.

Workplaces are having smaller versions of the same moment. A manager reads a client report that feels oddly smooth and wonders whether anyone thought about it. A junior analyst uses a chatbot to summarise a confidential contract and doesn’t realise they’ve just pasted it into a third-party service.

Using AI tools well is now a writing skill in its own right, alongside structuring an argument or editing for clarity. Like those skills, it can be learned. Knowing roughly what an AI detector looks for in your text is part of it, but most of it comes down to a few habits.

Know the Rules Before You Open the Tool

The first habit is boring and saves the most trouble. Find out what your institution, course, or employer allows.

For students, this usually lives in three places: the university’s academic integrity policy, the course handbook, and the specific assignment brief. They can differ. A university might permit AI for brainstorming in general while a particular module bans it entirely for a reflective essay, because the point of that essay is your own thinking. If the brief is silent, ask. An email to a tutor saying “Is it acceptable to use an AI tool to check grammar and suggest structure on this assignment?” takes two minutes and may protect you later.

For professionals, look for an acceptable use policy, an AI policy, or guidance from IT or legal. Many organisations now specify which tools are approved, what data can be entered into them, and whether AI-assisted work needs to be labelled for clients. If none exists, a sensible default is to treat any public AI tool like a public website: don’t paste anything you wouldn’t post online.

A Starting Point for Assessing AI Use

Not all AI use is equal. Asking a tool to fix comma splices is very different from asking it to write your argument. This rough guide works in many contexts, though your own rules always take priority.

Task	Usually fine	Use with care	Usually not acceptable without permission
Spelling, grammar, punctuation	✓
Brainstorming topics or angles	✓
Suggesting an outline you then rework		✓
Summarising your own sources to check understanding		✓
Rewording sentences you wrote		✓
Generating paragraphs you submit as your own			✓
Writing a reflective or personal piece			✓
Producing citations or references			✓ (and they’re often wrong)

The middle column is where judgment lives. Rewording one clumsy sentence is closer to using a thesaurus. Rewording every sentence in an essay until it no longer sounds like you is closer to having someone else write it.

A small example shows where the line sits. Say your draft contains: “The survey results kind of show that people who work from home talk to their managers less, which is bad for learning.” You ask a tool to tighten it and get back: “Survey findings indicate that remote employees engage in significantly reduced managerial communication, with detrimental implications for professional development.” That version is longer, vaguer, and has quietly added “significantly,” a claim your data may not support. A better result comes from fixing it yourself with the tool’s suggestion as a nudge: “In the survey, people working from home reported fewer conversations with their managers, which may limit how much they learn on the job.” Still your sentence, just clearer.

Use AI to Think, Then Write It Yourself

The most useful way to work with AI, in my experience, is as a sparring partner before and after drafting, not as the drafter. You keep the thinking, which is the part being assessed (at university) or the part you’re paid for (at work).

Some prompts that work well at the thinking stage:

I’m writing a 1,500-word essay arguing that remote work has weakened informal mentoring for junior staff. List the three strongest counterarguments someone could make, and for each one, what evidence they’d likely cite.

Here’s my outline for a project update to senior management. What questions would a sceptical finance director ask after reading it?

Explain the difference between qualitative and quantitative research as if I’m a first-year student, then give me two questions to test whether I’ve understood it.

None of these produce text you’d paste into your work. They produce angles, objections, and gaps for you to deal with in your own words.

After you’ve drafted, AI can act as a reader:

Read this paragraph and tell me where the argument is unclear. Don’t rewrite it. Just point to the sentences and explain the problem.

That instruction, “don’t rewrite it,” is the difference between getting feedback and getting a replacement.

Check Every Fact and Every Source

AI tools produce fluent text whether or not it’s true. They are particularly unreliable with references. Ask a chatbot for five academic sources on a topic and you may get a list where two are real, one has the right authors but the wrong year, and two simply don’t exist, complete with plausible journal names and page numbers.

The rule is simple: never cite anything you haven’t opened and read. Find sources through your library database, Google Scholar, or the reading list, and use AI only to help you understand them.

The same goes for facts, dates, figures, legal points, and quotations at work. If a tool tells you a regulation changed last year, find the regulation. Chatbots will happily quote a policy clause your organisation never had, in confident and entirely invented wording.

Protect Privacy and Confidentiality

This one matters more for professionals, though students working with interview data or placement information should pay attention too.

Before pasting anything into an AI tool, ask:

• Does it include personal data about clients, patients, students, or colleagues?
• Is it covered by a confidentiality agreement or marked internal?
• Would I be comfortable if this text appeared in a news story about a data leak?

If the answer to any of these gives you pause, don’t paste it. Many organisations offer enterprise versions of AI tools with stricter data handling, and those are the ones to use for work material. Anonymising text (replacing names, removing figures) can help, but it’s easy to miss identifying details.

Keep a Record of How You Worked

Documenting your process feels unnecessary right up until someone questions your work. Then it’s the most useful thing you have.

A few low-effort habits:

• Write in a tool with version history. Google Docs and Microsoft Word (with OneDrive) both keep a record of edits over time. A document that grew over several sessions, with deletions and rewrites, looks very different from one that appeared fully formed in a single paste.
• Save your prompts. If you used AI for brainstorming or feedback, keep the conversation or copy the prompts into a notes file.
• Keep your notes and drafts. Handwritten notes, rough outlines, and early drafts are evidence of your thinking.

Some universities now ask students to submit an AI use statement alongside assignments. Even when they don’t, writing one for yourself, a few lines on what tools you used and for what, makes it easy to answer questions later.

Understanding AI Detectors (And Their Limits)

Many universities and some employers use AI detection software, so it helps to understand what it does. A detection tool analyses patterns in text, mostly how predictable the word choices are and how much the sentence structure varies, and estimates the likelihood that some or all of it was machine-generated. Some tools also highlight which sentences look AI-written, and Quillbot’s separates fully AI text from human writing with some AI refinement, which is useful for anyone who used a tool for light editing.

What detectors produce is an estimate, not proof. They can flag human writing, particularly short pieces, very formal or formulaic writing, and writing by people using English as a second language, because all of those can be quite predictable. They can also miss AI text that has been heavily edited. Most universities that use them say a detector score alone shouldn’t be the basis for a misconduct finding, and that it should prompt a conversation instead.

Running your own work through a checker before you submit can be a reasonable step, for two reasons. If you used AI for some editing within the rules, you’ll see whether that editing changed the style more than you intended. And if a section you wrote yourself gets flagged, you have time to look at why. Often it’s a passage where you leaned on stock phrases or very even sentence lengths, and revising it in your own voice improves the writing anyway.

What you shouldn’t do is treat the checker as the goal. Rewriting text purely to get a lower score, especially text an AI generated, misses the point. The question your tutor or manager cares about is whether the work represents your understanding and effort.

If You’re Accused of Using AI

This happens to people who did nothing wrong, so it’s worth knowing how to respond calmly.

• Don’t panic or get defensive. Ask what evidence prompted the concern and what the process is.
• Gather your evidence. Version history, notes, drafts, browser history showing your research, and any prompts if you used AI within the rules.
• Offer to talk through the work. Being able to explain your argument, why you chose certain sources, and how your thinking developed is very persuasive. People who wrote something can usually discuss it in depth.
• Know your rights. Universities have formal procedures, and students usually have access to an advice service through the students’ union.

At work, the stakes and process are usually informal, but the same approach applies. Explain how you produced the work and show your drafts.

Disclosing AI Use

When AI use is permitted, disclose it. The major citation styles now include guidance on this. APA and MLA have both published advice on citing generative AI tools, and Harvard-style guidance varies by institution, so check your library’s referencing guide.

A brief disclosure might read:

I used ChatGPT to brainstorm possible counterarguments during planning and Quillbot’s grammar checker to proofread the final draft. All analysis and writing are my own.

At work, disclosure norms are still forming. Some teams add a line to reports; others mention it only when asked. If you’re unsure, telling your manager how you used the tool is rarely a mistake.

Habits Worth Building

Responsible use is mostly a set of small, repeatable habits rather than a single decision:

• Read the rules for each task, not just once a year.
• Draft in your own words, then use AI for feedback rather than replacement.
• Verify every fact and source independently.
• Keep sensitive information out of public tools.
• Save your drafts, notes, and prompts.
• Use an AI detector as a self-check if it’s useful to you, and understand that its result is a probability.
• Disclose your AI use when it’s allowed and when it’s asked for.

There’s also a broader question about which skills you want to keep sharp. If you always let a tool structure your reports, you may find in two years that you can’t do it under pressure in an exam room or a meeting. Some people deliberately write first drafts with no AI at all, for that reason.

FAQ

Is using a grammar checker the same as using AI?

Many modern grammar checkers use AI, but most institutions treat basic spelling and grammar correction differently from content generation. Check your specific policy, since some courses (language courses especially) restrict even that.

Can I use AI to help me understand a difficult reading?

Usually yes, as long as you’re using it to understand rather than to replace reading. Ask it to explain concepts, then go back to the original text. Don’t quote or paraphrase the AI’s explanation in your work as if it were the source.

What should I do if my policy is unclear?

Ask in writing, keep the reply, and follow the stricter interpretation until you hear back.

---

Conclusion: Before Your Next Assignment or Report

Pick one piece of writing due in the next fortnight. Before you start, write three lines in a notes file: what the rules say about AI for this task, which (if any) AI you will use, and where you’ll keep your drafts. Then write it. When you finish, look back at those three lines and see whether you stuck to them. It’s a small exercise, and it turns “using AI responsibly” from a vague intention into something you can check.

---