Charities Are Using AI Images. Most Haven’t Told Anyone.

AI tools are everywhere, but charities are still largely unaware of the risks and the obligations around artificial-imagery, writes Emma Bracegirdle, Founder of The Saltways.

Like many people in the sector, I was shocked to read The Guardian’s investigation in October 2025. Major NGOs were using AI-generated images of people living in poverty to fundraise, without labelling them or telling their supporters. So I started asking charity communications teams about their own use of AI-generated imagery, and what I found made the Guardian story feel less like an outlier and more like the tip of the iceberg.

Smaller organisations especially, were using AI images regularly, often in fundraising content, and without labelling them. The speed at which these tools had been adopted without much thought or discussion surprised me. It reminded me of the conversations I had around GDPR several years ago, except that back then, charities were cautious. They took legal advice. They updated their privacy policies. They worried about getting it wrong. With AI imagery, that caution seems to have disappeared.

Charities are full of good people, and I believe what is happening here is a lack of understanding of the impact rather than any deliberate intention to mislead. We surveyed 116 people and spoke in depth to eleven charity communications professionals. Here is what we found.

Obligations around AI imagery are not well-known

What we found was not a sector full of people making deliberate bad decisions. It was a sector that was under pressure, confused, and in many cases unaware of the risks. Ninety-four percent were unclear about their disclosure obligations. More than half of those using AI imagery were not labelling it. Seventy-six percent agreed that authentic content builds donor trust and yet a significant number were using AI-generated images anyway.

The reasons people gave were understandable, especially budget pressures and tight deadlines. The most common reason given was the need to protect beneficiary privacy, particularly for organisations working with children or survivors of abuse. These are real constraints, however, replacing a real person with an AI-generated image does not protect them. It substitutes them with a composite built from decades of potentially biased existing imagery that the sector has spent years trying to move away from. 

Enhancement, not generation, is the biggest challenge

The finding that genuinely shocked me was not about generation at all. It was about enhancement. Organisations were using AI tools to expand the edges of photographs, adjust lighting, remove backgrounds and not considering any of it as “using AI” in a meaningful sense. Enhancement was being treated as something entirely different and far more acceptable, when in reality it raises some of the same serious questions. The person in that original photograph consented to their image being used in a specific way, they did not consent to it being altered by an AI tool, and they almost certainly did not consent to their image being uploaded into software that may then use it to train future AI models. That is a conversation the sector has not yet had, and it needs to.

For those charities who think they can ignore AI generated imagery, by simply saying ‘we don’t use it’ were unaware of the wider risks. This is happening quietly, inside tools that staff already use every day, often without anyone in a leadership position knowing about it. If your organisation has not had a conversation about AI imagery, the chances are that conversation is already happening on the ground, just without any agreed boundaries.

This is not an argument against AI

None of this means never using AI. Abstract illustrations, campaign graphics that are clearly not photographs, visualisations of data, these are all reasonable uses that carry very little risk. The question is always whether the image could mislead someone into thinking they are seeing something real when they are not.

For NGOs, trust matters

There is a bigger issue here, and it matters more than any individual piece of content. Trust in charities is not high. Donors are more sceptical than they have ever been, and public scrutiny of how charities communicate is intense. We cannot afford to chip away at that trust further. Think about your least tech-savvy supporter, the one who still donates by cheque, who has supported your cause for twenty years. If they saw your content and later found out it was AI-generated and unlabelled, how would they feel? That is the standard we need to hold ourselves to.

In a world where people are already scrolling and wondering whether what they are seeing is real, charities have a genuine opportunity to be the exception. We can be the sector where supporters know that if an image is AI-generated, it will say so clearly, in plain language that anyone can understand. Not buried in small print. Not in technical metadata. Written for the person who has no idea what a large language model is and does not need to.

Ask yourself: Would our supporters be comfortable with this?

The disclosure does not need to be complicated. “This image was created using AI” is a starting point. Explaining briefly why, because we wanted to protect someone’s privacy, is better still. Sharing the prompt you used is best practice. Which brings me to the question I think every communications team should ask before publishing: would I be comfortable if our supporters could see exactly how this image was made and why? If the answer is no, that is probably your answer about whether to use it at all.

Authenticity is our superpower

The sector’s real superpower has always been authenticity. Real people, real stories, genuine connection with the communities we serve. AI imagery, used without thought or transparency, puts all of that at risk. It only takes one organisation to end up on a front page for the whole sector to feel the damage. 

We cannot be the sector that prides itself on ethics, on careful consent processes, on treating the people we serve with fairness and dignity and then quietly use AI-generated imagery without a second thought. It simply does not fit. The values we talk about in our fundraising appeals and our annual reports have to show up in our content decisions too, including the small ones, including the ones nobody is watching.

We have a chance to get ahead of this, and to do it in a way that is consistent with everything the sector is supposed to stand for. The charities that will be trusted in five years’ time are the ones being honest with their supporters now about what they create, how they create it, and why.

Emma Bracegirdle is the Founder of The Saltways, an ethical video production company working exclusively with the third sector