Norms around disclosing AI involvement in creative or professional work — to a client receiving a deliverable, an employer evaluating output, or an audience consuming published content — are still actively developing rather than settled, and specific requirements vary by industry, platform, and jurisdiction, some of which have introduced or are introducing formal disclosure requirements in specific contexts (certain advertising or political content categories, for instance). This page is general orientation, not legal advice, and specific, binding requirements should be checked against current rules for your specific industry and location.
For a broader risk, privacy, or evaluation perspective, ISO artificial intelligence resources provides useful external guidance.
Why a deliberate policy beats a case-by-case decision
Deciding on a personal or organizational disclosure policy once, deliberately, and applying it consistently tends to produce better outcomes than deciding case by case in the moment, which risks inconsistency that itself can look worse if noticed than a consistently-applied policy in either direction would. A consistent policy also removes a recurring, small decision-fatigue cost, echoing the tool-fatigue guide elsewhere in this section, and gives you a ready, considered answer if the question comes up directly rather than having to improvise a justification on the spot.
Factors worth weighing in setting your own policy
The specific context matters considerably: a client relationship where AI-assisted work is being represented as your own final output tends to warrant more transparency than internal drafting tools nobody outside your own process would reasonably expect to be disclosed, the same way most people don't disclose which specific word processor or research method they used unless directly asked. The degree of AI involvement matters too — AI-assisted editing or structuring of your own genuine thinking and expertise, discussed in the collaborative-writing-workflow guide on this site's Writing section, is a meaningfully different situation from content whose substance was largely AI-generated from a prompt with minimal human input, and a reasonable disclosure policy might treat these differently rather than applying one blanket rule to both.
The same discussion also raises questions about transparency and workplace data; this resource provides related context for evaluating those trade-offs.
- Set a deliberate personal or organizational disclosure policy in advance, rather than deciding ad hoc each time the question comes up — consistency itself has value beyond whatever the specific policy says.
- Weigh the specific relationship context — client, employer, audience — since reasonable expectations differ across these, the same way expectations about process transparency differ in non-AI contexts too.
- Distinguish AI-assisted work (your own substance, AI-assisted structuring or editing) from largely AI-generated work (AI-produced substance with minimal human input) — a reasonable policy may treat these differently rather than as one undifferentiated category.
- Check specific, binding disclosure requirements for your industry, platform, or jurisdiction directly — some contexts (certain advertising, political content, academic work) increasingly have explicit rules rather than leaving it to personal judgment.
- When in doubt, err toward more disclosure rather than less, particularly in a new client or professional relationship where trust is still being established — the downside of a client learning later, unprompted, that AI was involved without being told tends to be worse than the downside of volunteering the information upfront.
- Revisit your policy periodically as norms in your specific industry continue to develop — this is a genuinely moving target, and a policy set once may need updating as expectations shift.
Why this is worth deciding proactively rather than reactively
A policy decided proactively, before the question comes up in a specific, potentially awkward moment, tends to be more consistent and more genuinely considered than a policy improvised defensively after someone asks directly — the second scenario tends to produce an answer shaped by the discomfort of the moment rather than a genuinely thought-through position, which is a worse outcome regardless of which specific disclosure stance you ultimately land on.
This is a good example of a broader pattern worth applying to AI adoption generally: proactive, deliberate decisions made in advance tend to hold up better than reactive ones improvised under the specific pressure of a real, in-the-moment situation.