Who’s asking? Using the Right Voice Makes AI Fundraising More Effective

“I need help”—a uniquely human phrase laden with emotion, power, and humility. It’s a sentence uttered by a friend, stated by a victim, and written by a fundraiser. Tied to the deeply human desire to offer support and protect vulnerability, it’s the kind of thing that—in theory—sounds clunky, even inauthentic, coming from AI. But as AI permeates every sector, philanthropy notwithstanding, it’s worth asking what happens when a request for help comes not from a human, but from agents equipped with proven fundraising narratives and techniques.

New research from Chicago Booth’s Thomas Talhelm investigates how we respond to human versus AI solicitors, and how the rhetoric and storytelling shift depending on who's speaking. The goal of the study was to identify whether what we already know about persuasive storytelling in fundraising, specifically the choice between first-person and third-person narration, holds true when AI becomes the one telling the story. 

The setup is straightforward: A fundraising appeal can either be written in first person (“I lost my home in the flood), which tends to act as an effective emotional tool, or in third person (“She lost her home in the flood”), which has a different psychological effect grounded in credibility. Today, the appeal can come from either a human or from AI. Talhelm and his coauthors ran people through all four combinations: first- and third-person appeals from a human or AI. Then they asked people how willing they would be to donate to the cause.

The unexpected result? Third-person narration works better for humans, while first-person narration works better for AI. The outcome directly contradicted Talhelm’s prediction. "I think most people's intuition is that if you feel like it's AI talking about itself in the first person, that just seems strange," he explained.

The novelty of this study stems from a new kind of question. The comparison of an AI narrator to a human narrator only recently became something that could be meaningfully tested, because this is the first time in history where it has made sense to consider, let alone ask, the question.

The Theoretical Backbone

The research is grounded in Narrative Transportation Theory, which argues that the more immersed someone becomes in a story, the more that story shapes their attitudes and decisions. But the research pushes the theory in an entirely new direction. Narrative transportation has always been studied as a question of how a story is told; this research asks whether it also depends on who is telling it, testing whether a story pulls people in differently depending on whether the narrator is human or AI. "Humans are moved by stories in ways that textbooks or facts cannot match," Talhelm explained. Now, it's worth asking who's doing the moving. 

Talhelm's explanation for the results centers on empathic concern, the sympathy that comes from being confronted with another person's suffering, and the mechanism that connects transportation to the decision to actually give. It's one of the most basic, automatic human reactions there is. The researchers also tested whether perceived credibility explained the pattern, and it didn't; the effect runs entirely through emotion, not through people rationally judging AI as more trustworthy.

The findings, Talhelm suggested, might be capturing two separate psychological effects at once. Research has established that a specific, identifiable person is more persuasive and moving than messages from an abstract group. For example, "Jane needs your help" is more resonant than "thousands of children need help". The second effect, coined algorithm appreciation, captures how people place more trust in a computer's telling of events than in a person's. "If it's put in the words of a person, people seem to be a little more suspicious, almost like a human is biased about asking for money," Talhelm explained. The perceived objectivity behind AI’s telling of the same story doesn't prompt the same suspicion that inhibits emotion. 

First-person narration from AI seems to get both benefits in a single package: the intimacy of an individual story, without the suspicion a human narrator would trigger when telling it the same way.

"Maybe it's marrying those two benefits into one," Talhelm said.

Building the Experiment

The research took shape across three studies, each one building on the last to isolate the mechanism behind the result and then to test how far the effect actually extends. 

Study 1 recruited 345 participants through a survey platform and randomly assigned each person to one of four versions of a fundraising appeal: a first-person or third-person narrative, presented as coming from either a human or an AI fundraiser. The scenario described a fictional young girl seeking help for a serious illness. Donation intentions were measured using a Charitable Dictator Game: participants received 100 tokens, free to keep or donate any portion to the help-seeker, and those tokens were later converted into real money.

Study 2 recruited 355 new participants and repeated the same four conditions, but swapped in a different cause: a college student who lost her parents in a car accident and was struggling to stay in school while repaying medical debt. This time, the researchers added direct measures of empathic concern and perceived credibility, so they could test which of the two was actually driving the effect rather than just observing that it existed.

Study 3 scaled up further, recruiting 663 participants and shifting to a disaster-relief scenario. This time, the researchers layered in a third variable, identifiability, by describing the person affected either as a specific named individual or as an anonymous group.

Across all three studies, it was made clear to participants when they were reading an AI-generated appeal.

Key Takeaways

Strip away the studies and the statistics, and five clear patterns remain. Expand each one to explore the details of the findings.

A Surprising Truth 

None of this fits seamlessly into the story people tell about themselves and AI. Ask someone in the abstract how they feel about a machine standing in for a person's voice, and skepticism is the reflexive answer. But put the same person in front of an actual appeal and let them decide with something on the line, and the skepticism doesn't show up nearly as often as expected. That gap, between the story people tell and the choice people actually make, is what Talhelm keeps circling back to.

He pointed to a concept from social psychology called pluralistic ignorance to explain it. The classic example is the final years of the Soviet Union, when most people privately doubted communism but assumed everyone around them still believed in it, right up until the illusion collapsed all at once. 

Talhelm wondered aloud whether something similar might be happening with AI. Individually, people may be more willing to trust it than they let on, all while assuming their own openness is the exception rather than the norm. If that holds true, the public conversation about AI skepticism may not be capturing how people actually behave once they're the ones making the decision. He admitted he wasn't exempt from that gap himself.

"Before I ran this study, if you'd asked me how optimistic I'd be about using AI for this kind of fundraising, I would have said be really cautious," he said. "The results surprised me at how open people were to it."

Fundraising has always been a human activity, dictated by human stories, told in human voices. AI didn't invent persuasion, but it just walked into a role people assumed only a person could fill, and now it's rewriting the rules for how that role works. Talhelm and his coauthors weren't trying to settle whether that's a good thing. They were trying to understand it, and this research is the first real look at what's actually driving the difference now that the narrator has changed.