Using AI to write is not the trust problem. Hiding it is. Readers, clients, and search engines are increasingly fine with AI-assisted work; what they punish is finding out later that nobody told them. A written policy fixes that before it's an issue, and it's the core document behind the Trust pillar. For the human review step this policy assumes, see the Red-Pen Feedback Template.
This template covers three things: what you disclose and where, the language you use to say it, and the internal rules for what AI is never allowed to write alone. Fill it in once, publish the parts your audience should see, keep the rest as your team's working reference.
Part 1: What to disclose (10 minutes)
Not every use of AI needs a public disclosure. A rule of thumb: disclose when AI touched the words the reader is trusting, skip it for research or logistics.
Disclose:
- Blog posts, articles, or long-form content drafted with AI assistance
- Marketing copy, emails, and social posts where AI wrote a full draft
- Customer-facing responses generated or drafted by AI, even if a human edited them
No disclosure needed:
- Using AI to brainstorm topics, outline structure, or research background facts
- Using AI for internal work: meeting notes, data summaries, scheduling
- Light AI-assisted grammar or spelling checks on human-written text
Your fill-in: List anything specific to your business that falls in either bucket.
Always discloses: _________________________________ Never needs disclosure: _________________________________ Gray area you'll decide case by case: _________________________________
Part 2: Where disclosure lives (5 minutes)
Pick every place that applies. A policy that lives in one forgotten page isn't a policy.
| Location | What goes there |
|---|---|
| Website footer | One-line policy statement, linked to the full page |
| Blog post byline or footer | Per-post disclosure line |
| Email footer | Short version, one sentence |
| Support/chat intro | A line at first contact if AI drafts replies |
| Standalone policy page | The full policy, linkable from anywhere above |
Your fill-in: Check the boxes that apply to your business, then note the actual URL or location once it's live.
☐ Footer: ___________ ☐ Blog posts: ___________ ☐ Email: ___________ ☐ Support: ___________ ☐ Full policy page: ___________
Part 3: Sample disclosure language
Use these as starting points, not as-is copy. Swap in your own voice.
Footer line:
AI-assisted, human-vetted. We use AI to draft, and a real person reviews and edits everything before it goes out.
Blog post footer:
This post was drafted with AI assistance and edited, fact-checked, and approved by [name/team] before publishing.
Support chat opener:
Some of my replies are drafted with AI help. I read and approve every one before you see it, and I'm here if you need a real conversation.
Full policy page, opening line:
We use AI as a drafting tool, not a replacement for judgment. Here's exactly how, and what a human always does before you see the result.
Part 4: What AI may never write alone (10 minutes)
This is the internal half of the policy, the part your team follows even though readers never see it. Without this list, "AI-assisted" quietly becomes "AI-only."
AI never writes alone, no exceptions:
- Legal, medical, financial, or safety claims
- Pricing, guarantees, or anything contractual
- Apologies to a specific upset customer
- Anything citing a statistic, study, or number you haven't verified
- Testimonials, reviews, or anything presented as another person's words
Your fill-in: Add the categories specific to your work.
Never without a human first: _________________________________
The human-in-the-loop rule. Write down what "reviewed" actually means at your business, in one sentence, so it's not a vague promise:
Reviewed means: _________________________________ (e.g., "a named person reads every draft against our Voice Charter and fact-checks every claim before it publishes.")
Worked example: a solo consultant's policy
Public version (website footer, linked to a full page):
AI-assisted, human-vetted. I use AI to draft blog posts and email newsletters. I write every client deliverable, proposal, and piece of advice myself. Every AI draft is edited and fact-checked by me before it publishes. Questions: [contact link].
Internal rules (team doc, not published):
- AI drafts blog posts, newsletter copy, and social captions only.
- AI never writes: client-facing advice, proposals, pricing language, or anything with a number in it I haven't checked myself.
- "Reviewed" means I read the full draft against my Voice Charter, verify every fact, and rewrite any sentence that sounds like it could belong to someone else.
- Any AI draft that goes out without this review is a policy violation, not a shortcut.
Publish it, then use it
Once the public version is live, link it from your footer and any content page where it's relevant. Once the internal version is written, put it where new hires and freelancers see it on day one; it pairs directly with the Freelancer Voice Brief if you bring in outside writers.
A policy nobody can find is not a policy. Link it in three places minimum, and revisit it every time your AI use changes, not just once a year.