How to Show AI Use in an Essay When Your Teacher Allows AI

If your teacher allows AI but wants to know what you used, give them three things: a short statement naming each tool and what you did with it, a copy of the essay with AI-written text visibly marked, and your prompts or chat links. That's how to show AI use in an essay in a way your teacher can check.

Most AI transparency advice for students stops at the statement. That covers the "which tool" question, but teachers who allow AI usually also want to know which words came from it. This guide covers both, plus the awkward cases (grammar checkers, AI feedback, paragraphs you half rewrote) that the usual templates skip.

What does your teacher actually want to know about your AI use?

When a course permits AI "with disclosure", the teacher is usually trying to answer three questions about your submission:

1. Which tools did you use? ChatGPT, Claude, Gemini, Copilot, Grammarly, a translation tool.

2. What did you use them for? Brainstorming, outlining, explaining a concept, drafting, editing, translating.

3. How much of the final text is theirs, and how much is yours? This is the one students most often leave out.

University templates ask almost exactly this. The University of Cambridge's School of the Humanities and Social Sciences publishes a sample declaration that asks which permitted use you are acknowledging, which tool and version you used, what prompt you gave, what you used it for, and how you used or changed the output. The University of Melbourne tells students to write a declaration under a clear "Declaration" heading that names the specific tools and explains how their outputs were used. The University of Queensland offers an AI acknowledgement coversheet for courses that haven't supplied their own.

The Modern Language Association put the goal plainly in its August 2025 guidance on describing AI use: acknowledge substantive use so readers can see "where the AI ends and the human input begins." A statement alone tells your teacher that AI was involved. It rarely tells them where.

Disclosure versus citation

These are two different jobs, and many students mix them up.

• Disclosure (or acknowledgement) describes your process: "I used Claude to suggest an outline and ChatGPT to check my grammar."

• Citation points to a specific piece of AI output that appears in your essay, the same way you'd cite a book you quoted.

The University of Queensland's library draws the same line: acknowledgement explains how you used AI, while referencing follows a citation style with in-text citations and a reference list. Some courses want one, some want both. If your instructions only say "disclose AI use", a statement is the minimum, and marking the text is what makes it believable.

How to show AI use in an essay: five methods compared

There is no single official format. Here are the five methods teachers commonly accept, with what each one shows and where it falls short.

The strongest submissions combine two or three: a statement, a marked copy, and an appendix if the teacher asks for prompts. Marked text gets its own section further down, because it's the method most guides skip.

The disclosure statement

A good statement is specific and short. It names the tool (not just "AI"), the task, and what you did with the output. Tulane University's library guide makes the same point in its disclosure builder: name the tool, and say explicitly what AI did not do.

Weak: "I used AI to help with this essay."

Better: "I used ChatGPT to brainstorm three possible arguments about the causes of the 1929 crash and chose the second. Claude suggested the order of my body paragraphs. Two sentences in the introduction and one in paragraph four are AI wording that I kept, and they are marked in blue in the attached copy. All analysis of the primary sources is my own."

The second version answers all three of your teacher's questions in four sentences.

Citing AI in APA or MLA style

If you quote or paraphrase AI output, most style guides now want a proper citation.

APA. In September 2025 the APA Style team updated its 2023 advice. Because most chat tools now let you share a conversation via a unique link, APA recommends citing the specific chat when that helps readers. The template is: AI company name. (Year, Month Day). Title of chat [Generative AI chat]. Tool name or model. URL of the shared chat. In text, that becomes (OpenAI, 2025) or (Anthropic, 2025). APA treats the company as the author because AI itself cannot be one.

MLA. The MLA's revised guidance from August 2025 says not to treat the AI tool as an author. Describe the prompt in the title element, name the tool as the container, give the model name in the version element, then the company, the date, and a stable shareable link if the tool offers one. A paraphrase gets a short in-text reference to the prompt description.

The catch. Both styles assume AI output is a discrete thing you quote, like a source. That works when ChatGPT gave you one useful definition. It works badly when AI helped write half a paragraph that you then edited. That's where marking the text does a better job.

One more rule both the MLA and the University of Queensland stress: if an AI tool mentions a source, go to that source, read it and cite it directly. Don't cite ChatGPT as your evidence for a historical date.

An appendix of prompts and chat links

APA's 2025 guidance suggests keeping your prompts for your own records and, where it helps readers, including them in an appendix or similar section, as agreed with your instructor. The MLA points to the approach of educator Anna Mills, who links to or appends chat transcripts.

Practical tip: copy the share link from each chat as soon as you finish with it. Shared links depend on the chat still existing, so if you delete the conversation later, the link may stop working.

Process records

Some schools give teachers tools that record the writing process. Turnitin launched Turnitin Clarity in 2025, a writing space inside Turnitin where students draft, instructors can switch on an AI assistant, and teachers can review the writing process. Grammarly Authorship, which Grammarly made generally available in 2025 and extended to Microsoft Word that year, records where text came from in Google Docs and Word and can replay how a document was written.

Both are useful if your teacher has set them up. The trade-off is that you don't choose: Clarity runs inside your institution's Turnitin setup, and an Authorship report is built around Grammarly's tracking. Google Docs version history is a free fallback, but it shows edits, not which pasted text came from an AI.

How to show teacher which parts are AI, line by line

This is the part most guides skip. Your statement says "some AI wording remains." Your teacher reads the essay and has no idea which sentences you mean. Marking the text closes that gap.

There are two ways to do it: by hand in your usual word processor, or in an editor that marks AI text as it enters the document.

How to highlight AI text in essay drafts by hand

In Google Docs or Microsoft Word, the manual method looks like this:

1. Pick one highlight colour and use it only for AI text. Write the key at the top of the document: "Yellow = wording generated by ChatGPT."

2. Every time you paste AI output, highlight it straight away, before you do anything else.

3. When you edit a highlighted sentence, decide whether it is still AI text. A common rule: if you changed more than a few words, use a second colour for "AI, edited by me."

4. Before submitting, scroll through and check that every pasted section has a colour.

5. Keep an unhighlighted copy if your teacher wants a clean version to grade.

It works. It also breaks in predictable ways:

• You forget. Paste at 11 pm, plan to highlight later, and you've lost track by morning.

• Editing blurs the lines. You rewrite half of an AI sentence. Is it yellow, orange or nothing? Hand highlighting forces a yes-or-no answer to a question that's really a matter of degree.

• Highlights get copied around. Move a highlighted sentence and type into the middle of it, and your own new words often pick up the highlight too.

• No totals. If your teacher's policy caps AI text at, say, a fifth of the essay, you'd have to count words by hand to show you're under it.

• Portals strip formatting. Pasting your essay into a learning platform's text box often removes highlights. Upload a file instead.

None of this makes manual highlighting dishonest. It just makes it fragile, and a fragile record helps nobody when a teacher asks a follow-up question.

How do you mark AI text automatically with Assist Writer?

Assist Writer is a browser-based editor built for this exact problem. It records which text came from AI and which you typed, based on how the text entered the document, not on guessing from writing style. It doesn't generate text; you keep using ChatGPT, Claude or Gemini in another tab and bring the output in.

Here's the full workflow, from blank page to submission.

1. Open the editor. Go to the free Assist Writer editor. You can start without an account in demo mode, which keeps documents in your browser. A free account stores up to five documents in the cloud, which is safer if you switch computers.

2. Choose your editing mode. There's a rich-text editor that behaves like a normal word processor, and a Markdown editor with live preview and a linter. For a typical essay, rich text is the easier choice. You can switch a Markdown document to rich text, but not back again, so decide before you start.

3. Type your own writing normally. Everything you type counts as your text. A normal paste with Ctrl+V (Cmd+V on Mac) also counts as yours, which is right for your own notes or earlier drafts.

4. Paste AI output with Ctrl+Shift+V (Cmd+Shift+V on Mac). Text pasted this way is marked as AI text and shown in colour. If you prefer the mouse, right-click and choose "Paste as AI Text". There's also "Paste as Human Text" for your own material.

5. Edit the AI text into your own words. As you rework an AI passage, its colour gradually shifts toward normal text. A sentence you rewrote heavily looks different from one you pasted and left alone, which answers the "half rewritten" problem that hand highlighting can't.

6. Watch the stats bar. It shows the human versus AI split of the whole document as you go. If your course sets a limit on AI wording, you can see whether you're inside it before you submit, not after.

7. Pick a colour theme. AI text can show as a five-colour spectrum or as solid blue. Use whichever is easier for your teacher to read.

8. Export the marked version (Pro). With Pro, you can export to PDF, Word (.docx), RTF or Markdown with the author colours and the human/AI stats kept. That file is your marked copy.

The quick guide to shortcuts and editor modes has the full list of keyboard shortcuts if you want to check one.

Two smaller features help with a long essay: find and replace, and autosave about a second after you stop typing, so a crashed laptop doesn't take your marked draft with it.

Already started the essay somewhere else?

If you've drafted part of the essay in Word, Pro lets you import existing files such as.docx. When you import, you say whether the imported text is your own or AI. That works best for text with one clear origin, like a draft you typed yourself. If a file mixes AI and your own writing, it's cleaner to bring the AI passages in separately with the AI paste, and to note the mix in your statement.

What's free and what needs Pro?

Pro costs $4.99 a month or $52.99 a year; the Pro plan and pricing page has the details. If you only need one essay, a single month is enough. On the free plan, the marked document lives in the editor, so you'd need to show your teacher the screen or a screenshot instead of attaching a coloured file.

One honest limitation

Assist Writer records what you tell it. If you paste AI text with a normal Ctrl+V, or retype a ChatGPT paragraph by hand, it counts as your writing. The tool isn't an AI detector and doesn't police you. Its record is only as honest as the person using the shortcuts, which is also true of a disclosure statement. The difference is that once you build the habit, the marking takes no extra effort and doesn't depend on remembering to highlight later.

How to submit work showing AI assistance

You've written the essay and marked the AI parts. Now you need to hand it in so your teacher can actually use what you've given them.

Check what format your teacher wants first

Read the assignment page and the syllabus before you assemble anything. Look for:

• A required template or coversheet (several universities, including Queensland and Cambridge, supply one)

• Whether they want a clean copy for grading plus a marked copy, or one document

• Whether prompts or chat logs are required, optional or not wanted

• A citation style for any quoted AI output

• Any limit on how much AI text is allowed

If the instructions only say "tell me what you used", one well-organised file is usually best.

A submission package that covers everything

For most essays, this is a structure that answers every reasonable follow-up question:

1. The essay itself, with in-text citations for any AI output you quoted or paraphrased, in your course's style.

2. An AI use statement under a clear heading such as "Declaration" or "AI use statement", placed at the end before the reference list, or wherever your course template puts it.

3. The marked copy: the same essay with AI text coloured and the human/AI percentage shown. Exported as PDF if your teacher will read it on screen, Word if they want to comment in it.

4. An appendix with the prompts you used and the shared chat links, if asked.

If the learning platform only accepts one file, put the marked copy after the reference list as an appendix ("Appendix A: AI-marked version of this essay"). If you're printing, print the marked copy in colour; I'd pick the solid-blue AI theme for printing, because a single colour is the simplest thing to explain in a one-line key.

A worked example

Say you're writing a 1,500-word essay on whether social media has changed political protest. Your course allows AI for planning and limited drafting, with disclosure.

You use Claude to suggest counterarguments, ChatGPT to explain a sociology term you didn't understand, and ChatGPT again to draft a transition paragraph you then edit. You write everything else yourself.

Your AI use statement might read:

AI use statement. I used Claude (Anthropic) to list possible counterarguments to my thesis; I developed two of them in my own words in paragraph five. I used ChatGPT (OpenAI) to explain the term "connective action", then read and cited Bennett and Segerberg's original work rather than the chatbot's summary. ChatGPT drafted a transition between paragraphs three and four, which I edited; the remaining AI wording is shown in colour in Appendix A. Overall, about 6% of the final text is AI wording, according to the stats in the attached marked copy. Prompts and chat links are in Appendix B.

Notice what this does. It names each tool and company. It separates help (explanations, counterarguments) from text (the transition). It shows the original source was checked. It points to evidence instead of asking your teacher to take your word for it. And the 6% isn't a guess; it comes straight from the stats bar.

What counts as AI use you should disclose?

The obvious case is pasting ChatGPT paragraphs into your essay. The less obvious cases are where students get caught out, usually by under-disclosing.

AI help that never becomes text

Brainstorming, outlining and asking for explanations don't put AI words in your essay, but they do shape it. Tulane's library guide lists AI-generated theses, outlines or core argument structure as things to always disclose. Mention them in your statement even though nothing in the marked copy will be coloured.

Grammar and spelling tools

Policies vary most here. Some courses ignore spellcheck; others want any AI editing mentioned. Tools such as Grammarly now include generative features that can rewrite whole sentences, which is a different thing from fixing a comma. If a tool rewrote a sentence for you, treat that sentence as AI text. If in doubt, a one-line mention costs nothing.

Feedback from AI on your own draft

You paste your paragraph into Claude and ask what's weak about it. It tells you your evidence is thin. You go and find better evidence and rewrite the paragraph yourself. The words are yours, so they shouldn't be marked as AI, but the feedback should be in your statement ("I asked Claude for feedback on my draft argument").

If instead you ask the AI to rewrite your paragraph and paste its version back, that's AI text, even though it started as yours. Paste it with Ctrl/Cmd+Shift+V.

Translation

If you wrote in another language and machine-translated it, many universities treat that as something to acknowledge too. The University of Queensland keeps a separate guide just for citing machine translation, which tells you how seriously some institutions take it.

AI-suggested sources

If ChatGPT pointed you to a book, find the book, check it says what the chatbot claimed, and cite the book. The AI's role goes in your statement. Chatbots sometimes describe sources inaccurately or invent them, so the check isn't optional.

Common mistakes when you disclose AI use

These are the patterns that turn an honest submission into an awkward conversation.

• "I used AI." Too vague to be useful. Name the tool and the task.

• Disclosing the tool but not the extent. "I used ChatGPT for editing" when it rewrote three paragraphs undersells it. Your teacher will notice the change in voice.

• A statement that doesn't match the essay. If you say "no AI wording remains" and your marked copy shows coloured sentences, the mismatch is worse than either alone. Write the statement last, after checking the stats.

• Citing the chatbot as evidence. Facts need real sources. Cite the AI for the AI's words, not for the facts.

• Losing the chat links. Save share links as you go, in a note or at the bottom of your draft.

• Pasting AI text the normal way. In Assist Writer, Ctrl+V counts as your own text. Make Ctrl+Shift+V a habit for anything from a chatbot.

• Submitting a marked copy without a key. Add one line: "Coloured text was pasted from ChatGPT or Claude; colour fades as I edited it."

What if your teacher's AI rules are unclear?

Ask, and ask early. A short email before you start is far better than a disagreement after grading:

"The assignment says AI is allowed with disclosure. I plan to use ChatGPT for brainstorming and possibly some drafting. Would you like a statement only, or also a copy with the AI text marked? Do you want my prompts?"

Some teachers work from published frameworks. One widely used example is the AI Assessment Scale developed by Perkins, Roe and Furze, revised in 2025 in the Journal of University Teaching and Learning Practice. It has five levels running from "No AI" to "AI Exploration". If your assignment says something like "Level 3", look up what that level allows before deciding how much to disclose.

When rules are vague, over-disclosing is the safer mistake. In my view, a teacher who allows AI is far more likely to be annoyed by a vague statement than by a detailed one.

Does showing your AI use protect you from AI detectors?

Not directly. AI detectors guess from writing style, and they can flag human writing too. A marked copy and a clear statement don't change what a detector outputs. What they change is the conversation afterwards: if a detector flags a paragraph you've already marked as AI, there's nothing to explain. If it flags a paragraph you wrote, you have a record of how the rest of the essay came together, plus the drafts and chat links to back up your account.

That's the real argument for doing this properly. Disclosure isn't a confession. It's a record that lets your teacher grade what you actually did, and it keeps you out of arguments about what you didn't. For more on detectors, authorship and honest AI workflows, browse the Assist Writer blog on AI transparency.

Frequently asked questions

Do I need to disclose AI if I only used it for brainstorming?

Usually yes. Brainstorming puts no AI words in your essay, but it can shape your thesis or structure, and several university guides, including Tulane's, list AI-generated outlines and core arguments as things to always disclose. A single sentence in your AI use statement is enough, for example: "I used ChatGPT to generate possible essay angles and chose one; all wording is my own." Check your syllabus in case it says otherwise.

Should I cite ChatGPT or write a disclosure statement?

Often both. A disclosure statement explains your process: which tools you used and what for. A citation in APA or MLA style points to specific AI output you quoted or paraphrased. If AI only helped you plan or edit, a statement may be enough. If its wording appears in your essay, cite it as your style guide describes and mark that text so your teacher can see exactly where it is.

Can I mark AI text automatically instead of highlighting by hand?

Yes, with an editor that marks AI text as it enters the document. In Assist Writer, pasting with Ctrl+Shift+V (Cmd+Shift+V on Mac) marks the text as AI and shows it in colour, and the colour fades as you rewrite it. A stats bar shows the human and AI split. Marking and stats are free; exporting a coloured PDF or Word file needs Pro.

If I rewrite AI text completely, is it still AI text?

If you rewrote it in your own words, most teachers would treat the result as your writing, but the AI's role in shaping it still belongs in your statement. Rewriting is rarely all-or-nothing, which is why a gradual marker helps. In Assist Writer, AI text shifts toward normal colour as you edit, so a lightly touched sentence and a heavily rewritten one look different.

Does Assist Writer prove I wrote the rest of my essay?

No tool can prove that on its own. Assist Writer is not an AI detector; it records how text entered the document. Text pasted with the AI shortcut is marked as AI, while typing and normal paste count as yours. That makes it a clear, consistent record of your own labelling, which is useful alongside your statement, drafts and saved chat links, but it depends on you using the shortcuts honestly.