How do I make AI-assisted writing sound like me?
A practical tip from the detection side: keep your drafts in a single document with version history switched on (Word on OneDrive or Google Docs both do it). If your voice ever gets questioned, a timeline of your own edits is the clearest evidence there is.
Cite
humanise.ai Forum, "How do I make AI-assisted writing sound like me?", https://humaniseai.ai/forum/read/how-do-i-make-ai-assisted-writing-sound-like-me/, accessed 2026-11-20.- Harriet PellStaffForum editor, voice and editing ·
Write the parts that carry your thinking yourself, and use AI for the narrower jobs around them: questions, outlines, feedback, checking. When AI does produce wording, treat it as a rough suggestion you rewrite in your own sentences. Your voice lives in your choices, examples and rhythm, not in polish.
Why AI drafts all sound alike
A language model predicts likely next words. That makes its default prose smooth, balanced and safe: three-part lists, tidy summaries, words like "crucial" and "landscape", every paragraph the same length. None of that is wrong. It is just nobody in particular. Your markers and managers have read hundreds of pieces like it, and they notice when your writing suddenly stops sounding like the person they know.
The fix is not a cleverer prompt. It is deciding which parts of the work are yours.
Decide what AI is for before you open it
Start with the rules (the thread on whether you are allowed to use ChatGPT at university explains where to find them). If your assessment allows AI in an assistive role, check what it lists. UCL's assessment guidance, for example, gives drafting and structuring content, limited support with writing, feedback and proofreading as possible assistive uses, and asks students to take a critical approach because tools "can generate superficial, inaccurate and unhelpful outputs". Then pick jobs that leave the voice with you:
- Questioning. Ask it to quiz you on your topic, or to list objections to your argument.
- Outlining. Ask for three possible structures, then build your own from them.
- Feedback. Paste your paragraph and ask where a reader might get lost. Do not ask it to rewrite.
- Checking. Ask it to flag inconsistent spelling or tense, then fix them yourself.
Five habits that keep your voice
- Draft the hard paragraph first, by hand. The paragraph where you make your main claim is the one that most needs to sound like you. Write it before you ask any tool anything.
- Use your own examples. A specific case from your placement, your reading or your job does more for voice than any phrasing. AI cannot supply the detail only you know.
- Read it aloud. Anything you would never say out loud, cut or rephrase. If you trip over a sentence, a reader will too.
- Vary the rhythm. Mix short sentences with longer ones. Let some paragraphs be two sentences long.
- Keep a "words I do not use" list. Note the phrases you keep deleting from AI output and search for them before you submit.
A quick before and after
AI suggestion: "In today's rapidly evolving healthcare landscape, effective communication plays a crucial role in ensuring positive patient outcomes."
Rewritten by the student: "On my second placement, a patient refused her medication until someone explained what it was for. That ten minute conversation did more than the care plan."
The second version is shorter, specific and unmistakably someone's. It is also a better opening for a reflective essay.
Be honest about what you used
The APA Style team's current guidance on citing generative AI puts it well for academic work: using AI to edit or refine your own writing is a legitimate use, but because you are the human author you must check the output, disclose that you used it and cite the tool. Your institution's declaration form may ask for less or more, so follow it. Voice and honesty go together: the more of the thinking and wording is genuinely yours, the easier both the writing and the declaration become.
When English is not your first language
If you write in English as an additional language, it is tempting to let AI smooth everything. Use it to check grammar and to explain idioms, but keep your own sentence structure where it is clear. Markers value clear and accurate over elaborate. There will be a separate thread on detector bias against non-native writers. In the meantime, the thread on what a Turnitin AI writing score means explains why plain, careful prose can sometimes score as AI, and why a score on its own is not a finding against you.
Write something first.
Give people something to work with: at least 30 words on what happened and what you tried.
That is too long. Keep it under 6,000 characters.
Write the question as the title, 15 to 140 characters, no links.
Pick a category.
Add a name (2 to 40 characters, no links).
That email address does not look right.
That was quick. Read the thread, then try again.
The form expired. Reload the page and post again.
Something went wrong with the form. Reload and try again.
Please complete the check and post again.
Limit reached for now. Try again later.
This thread is closed to new replies.
This thread no longer accepts replies.
Something went wrong with the form. Reload and try again.
Post a reply or question first (name and email), then this browser can vote, edit and accept answers.
You cannot vote on your own post.
Only the author (within 30 days) or the forum team can do that.
That email already has an account. Use your sign-in link instead.
New accounts are paused for the moment. Try again later.
That was already posted.
1 answer
- Bronwyn AshbyStaffDetection and policy editor ·
A practical tip from the detection side: keep your drafts in a single document with version history switched on (Word on OneDrive or Google Docs both do it). If your voice ever gets questioned, a timeline of your own edits is the clearest evidence there is.