AI Video for Nonprofits: Five Defaults That Cost Donor Trust

In October 2025, The Guardian reported that aid agencies were running AI-generated images of extreme poverty in fundraising campaigns. The story wasn't about a broken law. What those groups risked was something harder to get back, the benefit of the doubt on every real photo they'd ever posted.

That's the actual risk in AI video for nonprofits. The tools are cheap and good, and a volunteer can now make an appeal in an afternoon. The trouble is five assumptions that come along with them, because donors and regulators still apply the old rules to the new footage.

Myth 1: AI-generated beneficiaries get a charity around consent

This is the one behind the Guardian story. Arsenii Alenichev, a researcher at the Institute of Tropical Medicine in Antwerp, had collected more than 100 of them and called the trend poverty porn 2.0. His explanation for why groups use them was blunt: they're cheap, and nobody has to ask for consent.

The reasoning sounds tidy. If the child in the frame doesn't exist, there's no one whose dignity or privacy you can harm. In practice the opposite happens. A prompt for a hungry child returns the most stereotyped picture the model has seen, and it stands in for a whole community that never agreed to be portrayed that way. When donors find out it's synthetic, they stop trusting the real photos too.

Our rule is narrow. Real beneficiaries appear only with informed consent, and generated people never stand in for suffering. AI earns its place in the scenes around the story, such as an animated map of the delivery route, a diagram of where a gift goes, or a dubbed version of a real interview.

Myth 2: a stylized scene in an AI fundraising video doesn't need a label

Some teams argue that a stylized AI scene is obviously illustrative, so a label would only break the mood. The UK's Fundraising Regulator disagrees. Guidance it published in December 2025 asks charities to make clear what is AI-generated, particularly where content could be mistaken by donors for real situations or people, and to publish an AI policy on their website.

That guidance only binds UK fundraising, but it's the clearest statement any regulator has made, and it's the test a skeptical donor will apply anywhere. If a reasonable viewer could think a scene was filmed, label it. That covers any photoreal person who isn't real, any cloned or synthetic voice of a real staff member, and any composite story told as one person's. A plainly animated map or chart doesn't need one.

Here's on-screen wording we'd use, short enough to read on a phone:

• For a synthetic presenter: "AI avatar of Dana Ruiz, executive director, used with her permission."
• For an illustrative scene: "Illustration: AI-generated. Not a real family."
• For a composite story: "Composite story based on several families we serve. Names changed, voiced by a staff member."

Put the label on the scene itself, in the first second it appears. A line in the YouTube description doesn't reach the many viewers who never open it.

Myth 3: a donation video skips the disclosures a letter carries

State charitable solicitation laws were written around mail and phone, but most apply to any solicitation. New York's Executive Law 174-b is a good example. A registered charity's solicitation has to say that its last financial report is available on request from the organization or the attorney general, describe the programs the money funds or offer that description on request, and point people to the attorney general's charities website and phone number. For written materials the statement must be conspicuous, in print no smaller than ten-point bold.

The statute doesn't carve out video. Our practice is to treat a fundraising video like a written solicitation, so the disclosure goes on the end card at a readable size and stays on screen long enough to read, with the full text on the donation page the video links to. Other states have their own wording, so check the ones where you're registered.

This is the myth that costs nothing to fix. Build the disclosure into the end-card template once and every AI-generated version inherits it.

Is it legal for a nonprofit to send AI voice calls or voicemails?

Partly, and the split matters. In February 2024 the FCC ruled that AI-generated and cloned voices are artificial voices under the Telephone Consumer Protection Act, so every robocall rule applies to them.

Tax-exempt nonprofits do get a carve-out, but it's narrow. Under 47 CFR 64.1200(a)(3)(iv), a nonprofit can make prerecorded or artificial-voice calls to a residential landline without consent if it makes no more than three such calls in any 30 days and honors opt-out requests. The exemption doesn't reach cell phones. Calls to a mobile number with an artificial voice need the person's prior express consent, and many donor lists are mostly mobile numbers.

So the cloned-director voicemail drop to the whole list is the one format we'd tell a charity not to run. A personalized video thank-you sent by email or text to donors who gave you their number for that purpose gets the same warmth without the robocall exposure. If you're sending texts, check those consent rules too, because the TCPA covers automated texts as well.

This page explains rule text and isn't legal advice. Charity regulation varies by state and country, so your counsel has the final say.

Myth 4: AI video for nonprofits is mainly for the big appeal

Teams reach for AI to make the year-end emotional appeal, which is the one video where real footage of real people matters most. The tools are better at the unglamorous work around it.

The jobs we'd hand to AI first are personalized thank-you videos to every first-time donor within a day of the gift, dubbed or translated versions of a real interview for the communities you serve (our AI video translation tools roundup compares the options), volunteer orientation and training clips that change every season, and a board or annual report recap built from your own numbers.

None of those need a synthetic beneficiary, and all of them used to go unmade because nobody had the hours.

Here's the rewrite we'd make on a typical AI-drafted appeal opening. The numbers are illustrative.

• Before: "In a world where hunger knows no borders, one child's tears can change everything. Will you be the hero she needs?"
• After: "Last Tuesday, 212 families picked up groceries at our Eastside pantry. Forty dollars restocks one shelf. Here's what that shelf looked like at 6 p.m."

The first version is what a model produces when you ask for emotion. The second only works if you have the real number and the real shelf, which is the point.

Myth 5: an avatar of the founder can stand in for the founder

An avatar of your executive director, trained on a few minutes of footage, is a reasonable way to deliver routine updates, event reminders and program explainers in several languages. Get her written consent, label it, and agree in writing what happens to the avatar when she leaves. Our AI avatar tools roundup compares how the main tools handle consent and access.

What an avatar can't do is ask a major donor for a gift. The objection we hear from development staff is that donors can tell, and for anyone giving at a level where they'd expect a phone call, that objection is right. Use AI for volume and reach, and keep the asks that depend on a relationship in the director's own unedited voice.

For the production side, our guide to making AI videos covers the workflow from script to export. Flowjam builds nonprofit video templates with the disclosure end card, AI labels and donate link locked in, so every version a volunteer makes, in any language, ships with them.

Frequently asked questions