Psychology of Fundraising: Five Findings From the Newest Research

Does the form of a thank-you change the next gift?

In the best recent test, the difference was not statistically significant. Maja Adena, Steffen Huck, and Levent Neyse ran two preregistered field experiments with a German opera house, published in Experimental Economics in 2025. The letters were mailed in May 2023. Preregistered means the researchers registered their analysis plan before receiving the data, which guards against searching for a result after the fact. In the first experiment, 8,800 fundraising letters either ended with "Thank you in advance" and "Thank you for your support" or did not. Response was 1.6% without the phrases and 1.8% with them, a difference that was not statistically significant, according to the authors' discussion paper (Table 2).

In the second experiment, 1,132 past donors received a thank-you postcard, either printed or individually handwritten, alongside the next fundraising letter. Response was 19.3% after a handwritten card and 18.7% after a printed one, and the average gift was about 87 euros against 114 euros, with neither difference significant. The authors write that their "zero results are quite precisely estimated", meaning the sample was large enough that a meaningful effect would likely have shown up. The study tested the form of thanks, not its content, so it leaves open whether what a thank-you says changes the next gift. Our thank-you timing guide covers speed, which this study did not test.

When does overhead cost a nonprofit donations?

Overhead is the share of spending that goes to administration and fundraising rather than programs. ChiaKo Hung, Jessica Berrett, and Ji Ma tested it in a survey experiment published in Nonprofit and Voluntary Sector Quarterly (published online in May 2024, in the April 2025 issue, volume 54), randomly showing participants an organization with an overhead ratio of 5%, 20%, 35%, or 50%. Donations to human-service nonprofits fell substantially once overhead reached 35%, while donations to health care nonprofits did not fall until 50%. The mechanism was trust: participants shown higher overhead trusted the organization less, and lower trust led to lower donations.

The finding revises the belief that every point of overhead loses donors. In this experiment, donations fell substantially only once overhead reached the threshold, and the loss came through trust. Because the effect runs through trust, the practical response is to explain costs before a donor has to guess at them: say what the finance office, the database, and the development staff make possible. A survey experiment measures stated donation decisions, so the exact thresholds are a guide rather than a rule. The donor trust guide covers how small organizations earn trust.

Does telling donors you used AI lower engagement?

In a 2026 study, yes. Yujie Zhao, Pete Pengcheng Zhou, and Zengxiang Chen report in the Journal of Business Ethics (2026) that in a Facebook field experiment with a sample of 118,927, charity ads labeled as made with AI received significantly lower click-through rates, meaning a smaller share of the people shown each ad clicked on it, than the same ads without the label. Controlled follow-up studies with 761 participants found why: AI use made the organization's motives look extrinsic, meaning motivated by cost savings rather than care, and made the ad feel less authentic. A study of 599 participants found that framing the AI use around the organization's mission reduced the penalty, and a study of 995 found that nonprofits were penalized more than corporations for the same AI use.

The abstract does not report the size of the click-through gap, so the study shows the direction of the effect rather than its size. For an organization that uses AI tools and chooses to disclose that use, the finding suggests explaining the use through the mission, for example that it lets staff spend more time with the people served. AI-drafted copy can also contain errors, so a person should check every draft before it goes out. The AI and your donor file guide covers what AI should and should not do in a development office.

What should a fundraiser change because of this research?

Each study points to one change that costs little to try:

1. 1.Choose the messenger deliberately. Identification with the fundraiser had the strongest association with giving (r = .36), so sign appeals from someone donors can recognize as like themselves.
2. 2.If an appeal says what peers gave, pick the amount by goal. Saying that peers gave a higher amount suits established donors, where larger gifts matter, and a lower figure suits campaigns where the goal is more donors.
3. 3.Stop treating handwriting as the purpose of a thank-you. In a field experiment, the share of past donors who gave did not differ significantly between handwritten and printed cards, so test what the message says about the gift instead.
4. 4.Explain overhead before donors infer it. The loss came through trust once overhead crossed a threshold, so say plainly what administrative spending makes possible.
5. 5.If you disclose AI use, tie it to the mission. The penalty for AI-labeled ads shrank when the use was framed around the organization's purpose.

Research on other organizations' donors is a starting point, and your own file is where a change proves itself. Donor Insights reads your contacts and gifts the way a fund reads a portfolio, so a change in messenger or suggested amount shows up as a change in response rate and average gift that you can compare against the same appeal last year. The methodology starts from your own giving records, and the same platform tracks which of these findings hold on your donors.