AI-Assisted Work: When to Disclose and What to Say

Disclose AI assistance when a client, employer, school, publisher, or platform requires it, or when leaving it unmentioned could mislead someone about how the work was produced. If no rule applies, use judgment: be especially transparent when AI materially shaped the result, handled sensitive information, or affected a decision others rely on—and describe specifically what you used and what you checked.

Why disclosure depends on context

There is no single disclosure rule for every kind of AI-assisted work. Requirements can come from an assignment brief, workplace policy, contract, publication guidelines, platform terms, or a regulator. Those rules change over time and differ by setting, so check the instructions that apply to the work rather than assuming one general custom governs it.

Disclosure is also a question of trust. A reader may reasonably care whether a person wrote a passage, an automated tool drafted it, or a system generated an image. That does not mean every spelling correction needs a prominent label. The useful distinction is between routine assistance and a contribution that changes the substance, appearance, or reliability of the work.

Ask: Would someone assess this differently if they knew how AI was involved? If the answer is yes—or if a rule says to disclose—say so. If the answer is no and no requirement applies, disclosure may still be courteous, but it need not become a lengthy explanation.

When to disclose

Disclose plainly in situations such as these:

• A rule requires it. Follow the specified method, format, and level of detail. An assignment may distinguish proofreading from generating or rewriting content.
• You are presenting work as evidence of your own unaided ability. This matters in assessments, hiring exercises, or other evaluations where the method is part of what is being judged.
• AI made a substantial contribution. Examples include drafting sections, generating an image used in a campaign, summarizing source material, writing code, or shaping analysis.
• Someone else will rely on the result. In reports, recommendations, customer-facing material, or decision support, recipients may need to know how the work was produced and reviewed.
• AI handled information that affects privacy or confidentiality. Follow the organization’s data-handling rules. Disclosure to a recipient does not make an unauthorized upload acceptable.
• The output could be mistaken for a real person, event, photograph, or recording. Clear labeling can prevent a misleading impression.

Disclosure may be unnecessary or minimal for ordinary spell-checking, formatting suggestions, or grammar corrections when these are permitted and do not alter meaning. But do not treat that as a universal exemption: a school, client, or employer can define even modest assistance as reportable.

What to say: a useful disclosure

A good disclosure is brief, concrete, and proportionate. Name the kind of assistance, identify the parts affected, and state what you did to review the result. Avoid vague claims such as “AI was used” if the reader needs to understand whether it drafted text, generated an image, or checked code.

For example: “I used an AI writing tool to suggest an outline and improve sentence clarity. I wrote and verified the final text.” If the tool drafted material, say that instead: “An AI tool produced an initial draft of the introduction; I revised it and checked the factual claims against the cited sources.”

Other concise examples include:

• Image: “The background illustration was generated with an AI tool; the layout and captions were edited by the design team.”
• Code: “An AI assistant suggested parts of this code. I reviewed the changes and ran the project’s available tests.”
• Meeting notes: “An automated transcription tool produced the first transcript. I corrected names and checked the summary against the recording.”
• Schoolwork: “I used the permitted tool to brainstorm possible research questions; the argument and submitted writing are my own.”

Only claim review or verification that actually happened. If you have not independently checked a statement, do not imply that you have. Where a form asks for the tool name, dates, prompts, or affected sections, provide those details as requested rather than substituting a generic note.

How to decide and disclose responsibly

1. Check the applicable instructions. Look at the assignment, contract, editorial policy, workplace guidance, or platform requirements. Ask the responsible person if the wording is unclear.
2. Describe the role of the tool accurately. Separate brainstorming, editing, translation, summarization, generation, and analysis; these are not interchangeable.
3. Review the output yourself. AI systems can produce incorrect claims, fabricated references, flawed calculations, biased wording, or code that does not work. Check against reliable sources and test where appropriate.
4. Protect information. Do not enter confidential, personal, or restricted material into a tool unless its use is allowed and appropriate safeguards are in place.
5. Put the note where it will be seen. Use the required disclosure field, a methods note, caption, footnote, or message to the recipient. A hidden or ambiguous note may not meet the purpose.

If the policy is silent, tell the person who owns the decision what you plan to do and ask before submission when the stakes are meaningful. Save a brief record of the tool’s role and your review if you may need to explain the process later. This is practical documentation, not a substitute for required disclosure.

Common mistakes and misconceptions

“AI did not write every word, so it does not count.” A substantial outline, rewrite, translation, or generated visual can shape the final work even when a person edits it extensively. Describe the contribution rather than arguing over a label.

“I disclosed it, so the work must be acceptable.” Disclosure does not override rules against particular uses, cure inaccurate content, grant permission to share protected data, or establish ownership or permission to reuse material. Those questions depend on the applicable policies and circumstances.

“A disclosure proves the result is reliable.” It does not. Readers still need sound evidence, citations, review, and appropriate human judgment. Equally, saying that a tool was used does not mean the result is automatically poor; the relevant questions include the task, the review, and the evidence.

Another common problem is overstating certainty. “The tool verified every fact” is not a meaningful assurance unless a person checked the claims using suitable sources. Prefer a specific, honest account of what was checked and what remains uncertain.

Frequently Asked Questions

Do I have to disclose AI-assisted work?
Follow the rules for the setting. If no rule applies, disclose when AI materially contributed or when someone could be misled about authorship, process, or reliability.

Should I disclose AI used for grammar or spelling?
Often a brief note is unnecessary for routine corrections, but some organizations require disclosure of all AI use. Check the relevant instructions, especially for assessed work.

What should an AI disclosure include?
State what kind of assistance was used, which parts it affected, and what review you performed. Add tool or process details when the recipient’s rules request them.

Do I need to name the AI tool?
Not always. Name it if a policy, form, client, or reader needs that information; otherwise, describing the assistance may be enough. Do not invent details you cannot confirm.

Can I use AI at work if I disclose it?
Disclosure alone does not grant permission. Check workplace rules, confidentiality requirements, and any approval process before using a tool, particularly with internal or personal information.

Is AI-assisted work still my work?
That depends on the task and the rules for authorship or evaluation. Be accurate about your contribution, follow applicable requirements, and do not present generated material as unaided work when that distinction matters.

Related

• Integrated vs RTX Workstation GPUs for CAD & Field Work
• AI Detection in Cybersecurity: How It Works
• Are Signal Messages Secure? Facts & Limits