Why fundraising imagery went generative
Charity communication has a permanent tension in it. The images that raise the most money are the ones showing identifiable people in distress, and those are precisely the images that raise consent and dignity questions.
Generation appears to resolve that. A picture of a child who does not exist cannot be exploited, cannot grow up and object, and cannot be recognised by their neighbours. Several organisations have used exactly that reasoning in public.
The trouble is that the image is still making a factual claim. An appeal illustrated with a generated scene is telling a donor that this is what the situation looks like, and if no photographer ever stood there, the claim is unsupported.
When it surfaces, and it does surface, the damage lands on the sector rather than on one organisation. Donors who learn that one appeal was illustrated with invented imagery apply that scepticism to the next appeal, including the honest ones.
Checking an appeal you have been sent
Start with the organisation and finish with the picture. That order matters because the most common fraud does not involve generated imagery at all: it involves a real photograph, a real disaster, and a payment link belonging to somebody unconnected to either.
-
Search the charity register
Most countries maintain a public register of registered charities with a number, filings and trustees. An appeal from an organisation that is not on it is the finding.
-
Reach the appeal through the register, not the message
Go to the organisation own site by searching for it, rather than following the link you were sent. Cloned appeal pages are the most common form of this fraud.
-
Reverse image search the main photograph
A hit on a news story from a different country or a different decade is common and immediately informative.
-
Run a pixel check
This reaches the generated case that a search cannot: an image with no earlier copies anywhere because it was never photographed.
-
Give through the organisation own channel
Never through a link in an unsolicited message, and never by transfer to a personal account.
The second step defeats most of the fraud in this category on its own. An appeal that is genuine will exist on the charity own website; one that exists only in the message you received is answering the question for you.
What a check tells you about an appeal
Read the result as information about the communication rather than about the cause. A generated image on a genuine charity appeal means the organisation made a questionable editorial choice; it does not mean the need is invented or the money will be stolen.
That distinction is worth holding onto, because collapsing it produces the wrong response. Withholding a donation to a legitimate organisation because their agency used a stock generator punishes the wrong thing.
| Register check | Image check | What it means |
|---|---|---|
| Registered | Photograph | An ordinary appeal |
| Registered | Generated | Honest cause, questionable imagery |
| Not found | Photograph | Likely a clone or a stolen appeal |
| Not found | Generated | Treat as fabricated |
| Registered, link differs | Either | Go to the site directly instead |
The last row catches the cloned appeal, which is the most professional form of this fraud and the one that survives every check on the imagery, because the imagery is the charity own.
If you work in fundraising
The reputational maths is unfavourable and it is worth stating plainly. Generated imagery buys a modest improvement in an appeal now, against a disclosure risk that compounds every time the practice becomes more detectable.
Where an illustration is genuinely needed, the alternatives are established and cheaper than the eventual apology. Commissioned illustration that reads as illustration, photography with informed consent and a named photographer, or an appeal that carries no image at all.
Where generated imagery is used, label it in the image caption rather than in a policy document nobody reads. Donors are considerably more forgiving of a disclosed illustration than of an undisclosed one they find out about later.
It is also worth checking your own back catalogue before somebody else does. Agencies have been quietly using generative fill for several years, and the first an organisation often hears about it is when a supporter runs an image through a checker.
Emergency appeals and the speed problem
Disaster fundraising compresses every one of these checks into a window measured in hours, which is exactly when fraudulent appeals appear. The pattern is well documented: a major event, a wave of new domains within a day, and social posts carrying real photographs from the event itself.
Speed is the attack. A donor who waits until tomorrow will find the established organisations, their registration numbers and their own donation pages. One who gives in the first hour is giving to whoever moved fastest, and fraudulent operations move faster than anybody.
The protection is a habit rather than a check. Decide in advance which organisations you would give to in an emergency, and give to them directly when one happens. That removes the entire category of risk without requiring any analysis of any image.
Where an appeal reaches you cold, the single most useful question is not about the picture. It is whether the organisation existed a week ago, and a domain registration date or a register entry answers that in seconds.
Regulators in several countries now publish guidance on the use of generated imagery in fundraising, and the direction is consistently toward disclosure rather than prohibition. Organisations that label early will find that easier than those waiting to be asked.