Coaching Peer-to-Peer Fundraisers
Every peer-to-peer campaign has a handful of participants who raise thousands and a long tail who register, never personalize their page, and never send a single ask. The difference between the two groups is rarely motivation. It is that the top performers know what to do next, and the rest are staring at a blank page hoping someone tells them. That gap used to be closable only with staff time nobody had. It is now closable with a coaching layer that AI can help you build.
Peer-to-peer fundraising asks something unusual of the people who sign up for it. A person who registers for your ride, walk, birthday campaign, or DIY challenge has volunteered to do a job that most professionals find uncomfortable, which is asking people they know for money. They have no training, no script, and no manager. In many cases they have never fundraised before in their life. Then they get a confirmation email with a link to a personal page and are essentially left to it.
What happens next is predictable and it shows up in every campaign report. A small group takes off immediately, usually people who have done this before or who have an obvious occasion attached to their ask. A larger group logs in once, does not personalize the page, and never returns. Somewhere in the middle sits a group that would raise money if someone told them what to write and when to send it. That middle group is the entire opportunity, and it is much larger than most organizations realize.
The economics here have always been the problem. Coaching individual fundraisers works, and everyone who runs a P2P program knows it works, because the participants who get a personal phone call from staff outperform the ones who do not. It simply does not scale. A development team of three cannot call four hundred participants, so they send a general email to everyone, which the people who most need help are the least likely to act on. The result is a coaching program that reaches the fundraisers who need it least.
This article is about closing that gap. It covers what your strongest fundraisers actually do differently, which of those behaviors can be prompted at scale, how to write nudges that work better than leaderboards, and where an AI coaching layer will do damage if you let it run unsupervised. If you are still setting up the campaign itself, our overview of AI in peer-to-peer fundraising covers the campaign-level mechanics. This piece is about the participant layer that sits underneath.
Activation, Not Recruitment, Is the Number That Moves
Most P2P programs are managed as a recruitment problem. The goal is registrations, the marketing budget goes to acquisition, and the campaign is judged on how many people signed up. This makes sense right up until you look at what fraction of those registrants ever raise a dollar. Activation, meaning the share of registered participants who secure at least one donation, is the number that determines whether a campaign meets its goal, and it is far more responsive to intervention than recruitment is.
Industry commentary on 2026 planning has been fairly direct about this. Analysis from Stephen Thomas on P2P trends notes that campaign volume has been rising while activation rates have stayed flat, and argues that fundraising teams will increasingly be measured on activation rather than reach. The same analysis makes the point that launching without a fundraiser coaching plan leaves activation rates on the floor, which matches what most program managers observe firsthand.
The reframe is worth taking seriously because it changes where the effort goes. Adding a hundred registrants to a campaign with weak activation adds mostly inactive accounts. Moving activation up by a few points across an existing roster produces real revenue from people who already raised their hand. It is also cheaper, since you are not buying attention, you are supporting people who already gave you theirs.
There is a second reason to prioritize activation. A participant who raises nothing has had a mildly disappointing experience with your organization and is less likely to come back next year. A participant who raises three hundred dollars from six friends has had a small success, has told six people about your work, and has brought you six donor records you did not have. The compounding value of that is what makes P2P worth running at all, and it only happens for participants who actually get started.
Activation rate
The share who raise anything
The single most diagnostic number in the program. Track it weekly during the campaign, not at the end, because a participant who has done nothing for three weeks is far harder to reactivate than one who has done nothing for four days.
Average raised per active fundraiser
Coaching effectiveness
Once someone is active, this measures whether your support is helping them go further. It responds to different interventions than activation does, which is why the two should never be collapsed into a single average.
Fundraiser retention
Year over year return
Returning participants raise more with less support, which makes retention the compounding asset in any P2P program. It is also the measure most damaged by a first-year experience where nobody helped.
What Your Best Fundraiser Does That the Median One Does Not
Before automating any coaching, it is worth being specific about what you are coaching toward. The behaviors that separate strong participants from weak ones are unglamorous and consistent across campaign types, which is exactly what makes them coachable.
They personalize the page, replacing the default template text with a reason that belongs to them. A summary of P2P research compiled by Kindsight reports that fundraisers who personalize their pages raise substantially more than those who do not, and that participants who send fifteen or more emails raise dramatically more than those who send fewer. Both findings point the same direction. The work that produces results is direct, personal, and repeated.
They ask individuals rather than broadcasting. A single social media post reaches many people and asks none of them, which is why it converts so poorly compared to a message that names the recipient. Strong fundraisers work a list. They think about who in their life would want to support this, and they write to those people one at a time, which feels laborious and is the entire mechanism.
They ask more than once, and they close. Most donations to a peer fundraiser come after a follow-up rather than the initial message, because the first ask arrives when the recipient is busy and gets mentally filed rather than acted on. Strong participants also announce progress and finish with a specific closing ask tied to a deadline. Weak participants send one message, get three donations from their most reliable friends, and stop.
They give to their own campaign first, which sounds trivial and is one of the more reliable predictors of eventual total. A page showing a self-donation signals commitment and removes the awkward emptiness of a zero balance. It is also the smallest possible first action, which is why prompting it early works so well.
The coachable behavior list
Everything a coaching layer should be driving toward
- Add a photo and replace the default page text with a personal reason
- Make the first donation to their own page
- Build a list of specific people to ask, not an audience to broadcast to
- Send individual messages rather than a single mass post
- Follow up with anyone who did not respond to the first ask
- Post a progress update at the midpoint
- Send a deadline-specific closing ask in the final week
- Thank every donor personally, which is what makes them give again next year
Where an AI Coaching Layer Actually Fits
With the behavior list in hand, the question becomes which of those steps stall for reasons a machine can address. The answer is most of them, because the barrier is almost always the blank page rather than unwillingness. Someone who has agreed to fundraise for you has already cleared the hard motivational hurdle. What stops them next is not knowing what to write.
The largest platforms have moved in this direction. GoFundMe launched an AI-powered smart fundraising coach in March 2026, described in its launch announcement as a personal command center offering recommended next steps, prewritten social posts and emails, and suggestions for who to contact next. Whether or not you use that platform, the design is instructive. It treats the fundraiser as someone who needs a next action rather than a report on their progress.
Page setup is the highest-leverage application. A participant who is asked three short questions about why they are doing this can be given a draft page in their own register, which they then edit rather than compose. Editing is a vastly easier task than writing, and the resulting page is meaningfully personal because the raw material came from them. The critical design choice is that the draft must be presented as a starting point they are expected to change, not as finished copy they can publish untouched, because a roster of pages that all sound like the same model is worse than a roster of thin but genuine ones.
Outreach drafting is the second. Most participants stall at the message itself, particularly the first one and the follow-up. A coaching layer that generates a short draft they can adapt for a specific person, with the recipient's relationship as an input, removes the exact friction that stops people. The same applies to thank-you messages, where the failure is usually not reluctance but the accumulation of twenty small writing tasks at the end of a campaign when the participant's energy is gone. Similar thinking applies to organizational messaging, which we cover in our guide to AI-assisted email subject lines.
Page setup assistance
Turning a blank template into their story
Ask two or three questions about their connection to the cause, generate a draft in their voice, and require an edit before publishing. This converts the highest-abandonment step in the whole funnel into a five-minute task.
Outreach and follow-up drafts
Message-level help, not mass sends
Short drafts adapted to a specific recipient and relationship, plus follow-up variants for people who did not respond. The follow-up is where most incremental revenue lives and where most participants give up.
Stall detection
Knowing who needs a human
Flagging participants whose behavior has diverged from a healthy pattern so staff can spend their limited calling hours on the people where a call will change the outcome, rather than on whoever is already at the top of the leaderboard.
Contact list prompting
Answering who should I ask
Structured prompts that help a participant think through categories of people in their life rather than staring at their phone. This works better than importing a contact list, which most people find intrusive and abandon.
Why Personal Nudges Beat Leaderboards
Most P2P programs lean on leaderboards and thermometers as their engagement mechanism, and there is a reason those tools have persisted. They work well for the participants who are already competitive and already performing. What they do to everyone else is worth thinking about carefully. A person who has raised fifty dollars and sees themselves in position 214 of 300 has been given information that is discouraging rather than actionable, and the most common response to discouragement is disengagement.
The alternative is a nudge addressed to the individual, referencing their specific situation and naming one next step. Practitioner analysis of 2026 P2P programs suggests that personal nudges to fundraisers who have not posted recently lift activation more than leaderboards do, which is consistent with what behavioral research generally finds about goal proximity and social comparison. Being told you are behind other people is demotivating. Being told that adding a photo takes two minutes and roughly doubles the chance of a first donation is a task you can complete.
This is where AI contributes something beyond drafting, because the nudge has to be specific to be worth sending, and specificity across four hundred participants is a generation problem. A message that references the fact that someone set up their page but has not yet sent a message, mentions the amount they raised on day one, and suggests one concrete action is far more effective than a general reminder. Producing that message individually for four hundred people is trivial for a model and impossible for a three-person team.
Frequency deserves as much attention as content. Peer fundraisers are volunteers, not staff, and an over-communicating coaching system is experienced as harassment. A defensible rhythm is a welcome sequence in the first few days, then contact triggered by behavior rather than schedule: a nudge when someone stalls at a specific step, a congratulation at a real milestone, a reminder as the deadline approaches. Silence in between is a feature. Competitive elements still have a place in the mix, and our piece on gamified fundraising with AI looks at how to use them without demoralizing the middle of the roster.
A nudge that names a number is not automatically a good nudge
There is a difference between specificity that helps and specificity that shames. Telling someone they are forty dollars from their goal with a week left is useful. Telling them they have raised less than eighty percent of participants is a fact that gives them nothing to do with it.
A working test for any generated nudge is whether it contains an action the recipient could complete in under ten minutes. If it only contains information about their standing, it belongs in a dashboard the participant chooses to open, not in their inbox.
Segment the Roster Before You Automate Anything
The single most common mistake in P2P communication is treating registrants as one audience. A returning participant who raised four thousand dollars last year and a first-timer who signed up because a coworker asked them to need almost nothing in common from you, and a message written for both will help neither. Segmentation is what makes an automated coaching layer feel personal rather than mechanical.
Four groups cover most rosters. Returning high performers need very little coaching and a lot of recognition, plus early access and any tool that saves them time. Returning modest performers are the most improvable group in the program, because they have already demonstrated willingness and usually stalled at a specific step you can identify from last year's data. First-timers who have taken an action need step-by-step guidance and encouragement at each stage. First-timers who have done nothing at all need the smallest possible first ask, which is usually adding a photo or making their own donation, not a request to build a contact list.
Segments should be behavioral and updated during the campaign rather than assigned at registration. A first-timer who personalizes their page on day two and raises three hundred dollars in week one has moved into a different group and should stop receiving beginner content, which is condescending once you are ahead of it. This kind of continuous re-segmentation is straightforward to automate and is where a lot of the perceived intelligence of a good coaching system actually comes from.
Data from prior campaigns makes all of this considerably sharper. If you can see where last year's participants stalled, you know which nudges to prepare and roughly when to send them. Many organizations have this data sitting in their P2P platform and have never analyzed it because doing so meant exporting several thousand rows and building a funnel by hand. That analysis is now a short task, and it should precede any coaching design rather than following it. The same logic applies to event fundraising more broadly, where registration and participation data usually goes unexamined.
Returning modest performers
The most improvable group
They came back, which means the experience was acceptable, and they stalled somewhere identifiable. Look at where their activity stopped last year and design the intervention for that exact step. This group typically offers the largest available gain per hour of staff attention.
Inactive first-timers
The largest group in most campaigns
Registered and then nothing. The intervention has to be tiny, because any ask that sounds like work confirms their suspicion that this will be hard. One photo, one sentence, one self-donation. Everything else can wait until they have taken a single action.
What You Should Not Hand to the Machine
The boundary here is sharper than in most nonprofit AI applications, and it follows a simple rule. AI should help your participant write to their own network. It should never write to their network on their behalf without them seeing it, and it should never write to the participant in a way that pretends to be a person who cares about them individually.
Sending on a participant's behalf is the line that matters most. Their name is on the message. Their relationship with the recipient is what makes the ask work, and it is the thing that gets damaged if a message goes out that they did not read. A participant who discovers your system sent something in their voice to their mother-in-law has a legitimate grievance, and no amount of fundraising uplift is worth it. Draft and hand over. Never draft and send.
The second boundary is emotional content generated about the cause itself. A model asked to write a compelling story about why this cause matters will produce something plausible and unverified, and participants will send it as fact. If your coaching layer supplies substantive claims about your programs, those claims should come from a fixed, approved library that your communications team wrote, with the model doing arrangement rather than invention. This is a straightforward control and it prevents a category of problem that is very hard to unwind once four hundred people have repeated it.
The third is the human touch at the top and the bottom of the roster. Your highest performers should hear from a person, because recognition from a machine is not recognition. Participants who are struggling in a way that suggests something else is going on, such as someone fundraising in memory of a family member who has gone quiet, should be handled by a staff member with judgment. The purpose of automating the middle is precisely to free the hours needed for those conversations, and an organization that automates everything has spent the savings on nothing.
Proving the Coaching Layer Did Anything
It is easy to add a coaching layer, see a decent campaign, and conclude it worked. Campaign totals move for many reasons, including the weather on event day, so attributing an increase to the coaching requires slightly more care than most organizations take. The good news is that P2P is unusually well suited to clean measurement, because you have many participants and can hold some of them out.
The most defensible approach is to randomly withhold a specific nudge from a portion of an otherwise identical segment and compare activation between the two. This is straightforward, it takes no additional tooling beyond a random assignment, and it will tell you within one campaign which of your nudges are doing work. Most organizations that try this discover that two or three interventions produce nearly all the effect and the rest are noise, which is useful because it lets you cut the volume of messaging without losing anything.
Watch the funnel rather than the total. If page personalization rates rise but activation does not, the page nudge is working and something later in the sequence is broken. If activation rises but average raised per active fundraiser falls, you have successfully brought in a lot of small first-time efforts, which is genuinely good for retention even though it flatters one metric and depresses another. Reading these together prevents the wrong conclusion.
Finally, measure the year-two effect, which is where P2P programs actually compound. A coaching layer that lifts activation this year should show up next year as a higher return rate, because more participants had a successful experience. That is a slower signal than campaign revenue and a more important one. Donors acquired through peer fundraisers also deserve their own follow-up path rather than being dropped into the general file, and pairing that with matching gift identification often surfaces revenue that a P2P campaign leaves behind entirely.
A simple test design
Enough rigor to learn something, little enough to actually run
- Pick one nudge and one segment, ideally the largest one
- Randomly hold the nudge back from a fifth of that segment
- Compare the specific behavior the nudge targets, not total revenue
- Keep the holdout group for the whole campaign so late effects show up
- Write down the result before the next campaign, when nobody will remember
Conclusion
Peer-to-peer fundraising has always contained an unresolved tension. It works because the ask comes from a friend rather than an institution, which means the institution cannot make the ask, which means the outcome depends on hundreds of untrained volunteers doing something uncomfortable without supervision. Coaching resolves that tension, and until recently coaching was available only to the fraction of participants a small team could personally reach.
What AI changes is the reach of that coaching rather than its nature. The advice a good P2P manager gives has not changed: personalize the page, give first, ask individuals, follow up, close with a deadline, thank everyone. What is new is the ability to deliver that advice as a specific next action to each participant, at the moment they stall, with a draft attached so the action takes minutes rather than an evening. That is a genuine capability shift for a program that has always been constrained by staff hours.
Getting it right requires holding two lines. The first is that the participant sends everything themselves, because their relationships are the asset and a message they did not read is a message that can damage one. The second is that automating the middle of the roster is worthwhile only if the hours it frees go into human contact with the people at either end. An organization that automates coaching and then reduces its P2P staffing has optimized a program into something less than it was.
A reasonable place to start is a single segment and a single stall point. Look at last year's data, find the step where most participants stopped, and build one well-written nudge with a draft attached for that exact moment. Hold it back from a fifth of the group so you learn whether it worked. That is a week of work, it will tell you more than a platform evaluation, and it is the version of this that organizations actually finish.
Want More From Your Next P2P Campaign?
We help nonprofits analyze where participants stall and build coaching that reaches the whole roster without losing the personal touch that makes peer fundraising work.