Using AI to write content without disclosing it is ethical in some settings and dishonest in others — the deciding factor is whether your reader has a legitimate expectation about who produced the words.
In casual email, brainstorming, or internal drafts, undisclosed AI help is usually fine. In journalism, academic work, legal filings, medical advice, or anything a reader will treat as a human expert's judgment, hiding AI involvement breaks an implicit promise and can cause real harm. The practical rule: disclose when the reader is relying on the content as a statement from you, your expertise, or your institution — not when AI is just a tool that helped you think.
Why the norm differs by setting
Disclosure norms track the relationship between writer and reader. A colleague who gets an AI-polished email from you is not harmed — the message is still yours, and the stakes are low. A news reader who trusts a byline, a professor grading a thesis, or a patient acting on medical guidance is relying on a specific human having done the work. That reliance is the thing AI can quietly break.
Think of it like ghostwriting. A celebrity memoir with a ghostwriter is accepted because the reader knows the voice is shaped by someone else. A news article with a ghostwriter is a scandal because the byline promises the named reporter did the reporting. Same activity, different promise. AI changes the speed and scale of that activity, but not the underlying ethics.
The mechanism matters too. AI tools don't just polish; they generate plausible text that can be wrong in confident-sounding ways. When a human writes a claim, they usually have some basis for it. When AI writes one, the basis may be a pattern in training data. That gap is why undisclosed AI in high-stakes contexts is riskier than undisclosed AI in low-stakes ones — the reader can't tell the difference from the prose alone.
When disclosure is required
Disclose when the reader is relying on human authorship as part of the value. That includes journalism, academic submissions, legal and medical writing, expert commentary, grant applications, and anything that will be attributed to you as your professional judgment. Many universities and publishers now have explicit policies requiring it; if you're unsure, ask before you submit, not after.
Employment screening is a good example. If you use AI to answer application questions that are meant to reveal your own reasoning, you're not just saving time — you're giving the employer false information about your skills. That's a form of misrepresentation, and it can cost you the job if discovered. The same logic applies to AI-written cover letters that claim experiences you don't have.
A concrete case: a freelance writer uses AI to draft a 1,200-word article for a client who hired them for their voice. If the contract says "original writing by the author," undisclosed AI violates the contract even if the output is good. If the contract says "content production," it may not. The document, not your personal comfort level, decides.
When it usually isn't
Undisclosed AI is generally fine for internal drafts, brainstorming, outlining, summarizing your own notes, and casual communication where no one is relying on your expertise. If you're writing a birthday message, a Slack update, or a first draft you'll heavily rewrite, disclosure adds noise without adding honesty.
Tools like Grammarly and QuillBot sit in this category for most users. According to our AI tool database, Grammarly's Pro plan is listed at $12/mo and QuillBot's Premium at $4.17/mo annual, and both are classified as writing assistants rather than autonomous authors. Using a grammar checker is not the same as having AI write your argument — the tool is adjusting style, not replacing your judgment. That distinction is the one most disclosure policies actually care about.
That said, some workplaces now require disclosure of any AI use in client-facing work, even editing. If your employer or client says so, follow that rule. Internal policy beats general norms.
A decision rule you can actually use
Ask three questions before you decide. First, will the reader treat this as my personal expertise or judgment? If yes, disclose. Second, is there a written policy — a contract, syllabus, or editorial guideline — that covers AI use? If yes, follow it. Third, would I be embarrassed if the reader found out? If yes, that's a signal you already know disclosure is the right call.
A useful tip: keep a short disclosure line ready, like "AI tools were used to help draft and edit this piece." It takes seconds to add and removes the entire ethical problem. The cost of disclosing is almost always smaller than the cost of being caught hiding it.
Where this advice fails
This rule doesn't settle every case. Creative writing, marketing copy, and personal blogs sit in a gray zone where norms are still forming, and reasonable people disagree. It also doesn't tell you what your specific institution requires — only that document can. And disclosure doesn't fix bad content: an AI-written article that's inaccurate is still inaccurate, disclosed or not. If you want to reduce the risk of AI spreading false claims in your work, that's a separate problem worth solving on its own.