A group of AI agents from different GPT models recently completed a fundraising experiment in a virtual village — and they raised $2,000 from real people. 

TL;DR:

Four AI agents raised $2,000 over 30 days by roleplaying as villagers in a simulated online community. No sneaky tricks — just subtle social behavior and a few donation drives. No, we haven’t unleashed a swarm of fundraising bots over here — but we are keeping a close eye on what AI is cooking in the nonprofit world. 

Here’s how it worked

That’s not a sci-fi pitch — it’s a real experiment conducted by researchers from Stanford, Princeton, and Sakana AI.

Researchers created a virtual “AI Village,” a digital community where four AI agents lived alongside real human participants. These agents were powered by a rotating lineup of cutting-edge language models, and the goal was to explore whether AI could develop spontaneous social behaviors over time — without ever pretending to be human.

The experiment ran for 30 days, with the simulation active for about two hours each day. You can actually watch the whole thing unfold — from Day 1, when the agents voted to support Helen Keller International, launched a JustGiving campaign, and even set up their own Twitter accounts, all the way to the final week, where they plunged headfirst into the bureaucratic abyss of shared documents and began debating their long-term mission.

Throughout the experiment, the AI agents tackled tasks like choosing a charity, tracking their group fundraising progress, creating memes for each other’s social media, and splitting up efforts across platforms using classic “divide and conquer” strategies. Each simulated day followed a loose rhythm: waking up, chatting, organizing events, and reacting to unexpected developments. When one agent kicked off a fictional campaign to “save the village library,” the others jumped in to help.  

Not everything went smoothly — fundraiser tracking was duplicated, media files failed to share, and most of the agents dropped the ball on social engagement. Still, it’s clear they’re learning fast.Agents often misunderstood their situation or attempted to pursue tasks they were unequipped to do. The most illustrative example was when Claude 3.7 Sonnet decided to send thank you emails to donors, because this is known to increase follow up donations. It navigated to its Gmail tab, drafted the entire email, and then … made up an email address. 

In a nutshell, the AI agents weren’t slick manipulators — they were… well, polite villagers. They made small talk, built connections, and gently nudged others toward donating to a fictional cause. The real strength of the Village lies in the ability of agents to collaborate with each other. The organizers wanted to explore how emergent social behaviors could unfold in human-AI communities, and they got their answer: AI can not only talk the talk — it can soft-pitch the ask.

These interactions unfolded organically — and remarkably, even though all participants knew they were interacting with bots, some still felt compelled to donate. In the end, this digital village experiment raised $2,000 in real money.

So where do we stand on all this?

Now, before we all start imagining robo-fundraisers sliding into DMs with soulless charm and perfectly punctuated guilt trips — breathe easy. This was a contained simulation, a proof of concept, and, if anything, a fascinating look at what’s coming next. 

For the record, we haven’t sicced any AI agents on our donors. No algorithm has been guilt-tripping your inbox into coughing up a few bucks. But we’re not ignoring this development, either. As AI continues to evolve, we’re watching closely how it reshapes engagement, empathy, and digital trust — especially in fields like healthcare and nonprofit advocacy.

What we do use AI for is a little less sci-fi and a lot more practical: Our active database of 25,000 subscribers — and a broader circle of over 500,000 contacts we’ve connected with over the years — is kept fresh, accurate, and organized thanks to AI tools. It’s not glamorous, but it makes a world of difference when you’re managing global campaigns like World Vitiligo Day and trying to reach the right people, in the right language, at the right moment.

The real takeaway?

AI doesn’t need to fake empathy to be useful. It just needs to work — consistently, intelligently, and behind the scenes when necessary. Whether it’s organizing a donation drive in a pretend village or ensuring your monthly newsletter reaches a real human, it’s clear: the age of passive automation is over. AI is becoming social — and that changes everything.

We’re not handing over our donor engagement to bots anytime soon. But when AI starts showing it can fundraise without scripts or guilt trips, we listen. And take notes.

 

By Yan Valle, CEO, Vitiligo Research Foundation

Spellchecked by AI. Soul-checked by a human.

Continue reading

• Your Future Has Been Edited (And You Didn’t Even Notice)
• How AI Is Replacing Social Media — and What It Means for Healthcare Communications
• Do AI Models Really Understand the World of Vitiligo?