Your marketing team is almost certainly using AI already. In the 2026 State of Nonprofit AI report from NTEN and the Bridgespan Group, 98% of the 917 respondents said AI is used in some capacity at their organization, and 53% said they had used AI for work in ways that were informal or not clearly covered by guidance. Only about four in ten executives said written guidance on safe and responsible use was fully in place.
A separate December 2025 survey of 346 nonprofits for Virtuous’s 2026 Nonprofit AI Adoption Report found that 47% had no AI governance policy at all. Its more useful finding is about the kind of policy that helps: organizations with “enabling” policies, which spell out what is encouraged, what needs approval, and what is off limits, reported major impact at higher rates than organizations with restrictive policies or none. The report also notes that the groups seeing results often describe “simple, one-page guidelines created in a single meeting.”
That is the goal here: a policy that makes the time AI saves safe to keep, so nobody has to guess.
Why marketing needs its own page
An organization-wide AI policy usually covers data security, HR, and procurement. Marketing has a different exposure: it is where AI output becomes public with your name on it. Donor appeals, admissions emails, social posts, and alt text can all carry a wrong number, an invented detail, or a voice that does not sound like you.
Google’s own guidance on generative AI content notes that generative models predict likely words rather than retrieve facts, and says AI-assisted content should be fact-checked before publishing, including titles, meta descriptions, and alt text.
So the marketing page answers one question for every task: can I just do this, does someone check it first, or is it off the table?
The structure: six short sections
- Purpose. One sentence on what you want AI to make possible, not just what you fear.
- Three tiers of use. Go, Review, and No.
- The data rule. What never goes into an AI tool.
- The review rule. Who checks what, and what they check for.
- Disclosure. When you tell your audience AI was involved.
- Owner and review date. One named person, and a date on the calendar.
Tier the work by risk, not by tool
Tools change every few months. Risk categories do not. The Center for Nonprofit AI’s policy template makes this case directly: tier uses by risk rather than banning tools by name, and keep the list of approved tools somewhere it is easy to update.
Here is a starting version for a marketing and communications team:
| Tier | What it means | Marketing examples |
|---|---|---|
| Go | Use approved tools freely. No sign-off. | Brainstorming subject lines and headlines; outlining a newsletter; summarizing a public report; reformatting copy that is already approved; drafting internal briefs |
| Review | Allowed, but a named person approves before anything goes out | Any public copy; donor and alumni appeals; admissions and enrollment emails; anything with a statistic, name, or title; alt text and meta descriptions; AI-assisted images |
| No | Not allowed without written executive approval | Invented quotes or testimonials; synthetic photos presented as real students, clients, or events; crisis, grief, or safety communications; pasting donor, student, or client records into unapproved tools; AI as the final approver of anything public |
Two notes on the “No” row. The photo rule matters more for mission-driven organizations than for most brands, because your credibility rests on real people and real outcomes. The crisis rule follows Candid’s guidance for nonprofits, which lists moments of crisis, grief, and trauma response among the uses too sensitive for AI drafting.
Write the data rule in plain sentences
Make this the easiest paragraph to read. For example:
Do not enter names, contact details, gift amounts, student or client records, or anything confidential into an AI tool unless that tool is on our approved list for that kind of data. When in doubt, remove the identifying details first or ask the policy owner.
Colleges should name student records specifically, since those are often covered by FERPA and campus data rules. Nonprofits should name donor data and anything about the people they serve.
Make the review rule specific
“A human reviews everything” fails in practice because nobody knows which human, or what they are checking. Name a reviewer for each content type (for example, the development director for appeals, the admissions communications lead for enrollment email) and give them a short checklist:
- Every number and claim traces to a source you could show a board member.
- Names, titles, program names, and dates are correct.
- Any story about a real person is true and has their consent.
- It sounds like you, not like a generic nonprofit.
- Nothing reads as a promise you cannot keep.
- Alt text, page titles, and meta descriptions got the same check as the body copy.
If you already have audience personas or a voice guide, attach them to the brief you give the AI tool and to the reviewer’s checklist. The same reference keeps both honest.
A worked example: a spring appeal
Say a small community nonprofit is preparing its spring email appeal. Here is how the tiers play out:
- Go. The development director asks an approved AI tool for fifteen subject line options and three ways to structure the letter, using last year’s appeal as a reference. Nobody needs to approve brainstorming.
- Review. She drafts the letter with AI from a brief that includes one real story from the program team, with the person’s consent on file. The executive director reviews it, checks the meal count against the latest program report, and confirms the story details with the program lead.
- No. The tool offers to generate a photo of a smiling family at a distribution site. That stays out. The team uses a real photo with a signed release, or no photo.
The time saved on outlines and drafts goes into the part AI cannot do: the conversation that produces a true, specific story.
A college version works the same way. AI drafts five versions of a yield email for admitted students; a named admissions lead checks deadlines, aid language, and program names before anything is scheduled; and no tool sees an applicant’s record unless it is approved for that data.
Decide disclosure once
Pick a disclosure standard once instead of debating it case by case. A reasonable starting point: label AI-generated images; do not label text that a human drafted or substantially edited and approved; and never present AI output as a real person’s words. Google notes that sharing how content was created can give readers useful context, so consistency matters more than any one choice.
Name an owner and a review date
One person owns the page. The Center for Nonprofit AI template suggests putting the first review within six months, which is about right given how fast tools change. At each review, ask what people avoided because the rules were unclear, what slipped through, and which tools should move on or off the approved list.
If your marketing lead is part-time or fractional, the review rule needs backups and response times so it does not become a bottleneck. In my own practice, I take a limited number of engagements and keep a fixed meeting window, so the review step is what keeps the work safe: nothing public goes out until a person has signed off. More on that model is on the services page.
Copy this outline
[Organization] marketing AI policy · Owner: [name] · Next review: [date]
Purpose: We use AI to [give staff time back for storytelling and relationships] while protecting the trust of the people we serve.
Go: [brainstorming, outlines, internal drafts, summaries of public material]
Review (named reviewer approves): [all public copy, appeals, enrollment email, statistics, alt text, AI-assisted images]
No: [invented quotes, synthetic photos of real-seeming people or events, crisis communications, unapproved tools for personal data, AI as final approver]
Data rule: [plain sentences, see above]
Disclosure: [your standard]
Approved tools: see [link to the living list]
Fill it in with your team in one meeting, then use it for a month before refining it. For “Go” row ideas, this earlier post on ways to use AI in nonprofit marketing is a reasonable list to start from.
This is a practical starting point, not legal advice. Check data rules (FERPA, donor privacy, state law) with your counsel.
Sources
- NTEN and the Bridgespan Group, 2026 State of Nonprofit AI: Adoption and Governance Report
- Virtuous, The 2026 Nonprofit AI Adoption Report
- Center for Nonprofit AI, Nonprofit AI Policy Template
- Candid, Getting started on a responsible AI use policy for nonprofits
- Google Search Central, Google Search’s guidance on using generative AI content on your website
If you would like a second set of eyes on your team’s version, Pogona Creative can help.
Drafted with AI assistance; reviewed and edited by David May.