I recently came across an insightful study from Whole Whale about how AI is learning about nonprofits and deciding who to recommend to their users.
You should certainly read the whole study for a fascinating look into bias in AI as well as to get a peek into the complex technology behind these large language models (LLMs).
ChatGPT from OpenAI is the most commonly referenced, but there are many LLMs out there that your constituents, donors, and potential donors may be using to help decide which organizations to support financially.
It becomes increasingly important now to create content that teaches these AI LLMs who you are as an organization, what you do, how trustworthy you are, and why your organization is the right choice to recommend.
Remember, LLMs aren’t actually answering questions or giving advice; they’re regurgitating patterns of words and pieces of data that are statistically likely to make a good answer to the question or prompt.
To state the obvious, they sound increasingly human, but they aren’t human. Nor are they a more-or-less predictable algorithm like a search engine. A Google search would probably not lead you to believe that the word strawberry has only two Rs, but… ChatGPT has been known to insist upon exactly that.
Complicating matters even further, LLMs are limited to the snapshot of language and data they learned from. So if you’re using a version of ChatGPT or Gemini or Grok that was trained on a snapshot from 2022… then you’re getting potentially outdated information. This date is called the “knowledge cutoff date”.
Phew.
What's a small nonprofit to do?
For starters, if you’re a small nonprofit, it’s unlikely that you’re going to put out enough content to arrest an LLM’s notice quickly.
Because LLMs use statistics to determine content relevance and accuracy, the sheer volume of references to larger nonprofits is likely to overwhelm general inquiries.
For instance, the study referenced above notes that when searched for general causes like education, cancer research, or women’s rights, only the largest organizations made the cut of being recommended. And that makes sense.
On the other hand, if you’re a local, niche, or religiously-specific organization, you may have LLM users asking for recommendations for more specialized results.
Here’s where you have your chance.
Let’s see what the free version of ChatGPT recommends when I ask specifically about the best small nonprofits in my local area.
(Note: this study was done on a desktop computer, so if you’re on a mobile device, try rotating it to horizontal so you don’t go blind trying to read the following screenshots.)
(Note: this study was done on a desktop computer, so if you’re on a mobile device or laptop with a smaller screen, try tapping/clicking each screenshot to open it in a lightbox, which should be a bit easier to read.)
Great start!
The LLM notices here that I haven’t specified a cause that I’m interested in and recognizes that the list provided is not exhaustive.
Now we’re getting somewhere. Take a look at the sources listed in the small gray buttons next to the recommended nonprofits. I highlight a few examples to note:
- Lancaster County Food Hub – ExtraGive (local giving day event) and WITF (local news source)
- CAP Lancaster, Boys & Girls Club, Bridge of Hope – official websites
- HOPE International – Wikipedia
- CrossNet Ministries – 2 x United Way (local resource directory)
- DVS – Touchstone Foundation (local resource directory) and United Way (local resource directory)
You can see from just this one result that linking from other highly ranked websites and making sure your organization’s listings within local resource directories could be hugely important to how an LLM understands and ranks you. Your organization can borrow from the larger organization’s status by being listed and linked to from them.
And Wikipedia as the source for information about HOPE International… Who’d have expected that?
ChatGPT only linked to the main official site of the organization 40% of the time, the other 60% relying on data from third-party sources!
This is a huge vote in favor of backlinking, which should already be part of your organization’s SEO strategy for your website.
I also want to highlight that with prompts for more specific or niche organizations (e.g. “Christian nonprofit in lancaster pa” and “pro-life nonprofit in lancaster pa”), ChatGPT several times included sources from charity evaluators like Candid GuideStar, Charity Navigator, and Cause IQ, reflecting a similar result from Whole Whale’s much more comprehensive study.
The study reveals that AI models often recommend charities by citing third-party validators such as Charity Navigator, GuideStar, and GiveWell. OpenAI and Grok, for instance, are heavily influenced by effective altruism principles, frequently citing GiveWell and specific Charity Navigator ratings.
I doubt anyone other than a researcher like myself will add this last question (below), but I wanted to give some context for how quickly the changes and updates you make to your online presence could potentially impact how well ChatGPT and other LLMs understand you.
Most companies to don’t announce their updates far in advance, so you might not know when, for instance, ChatGPT is suddenly updated to have a more recent knowledge cutoff date. But, as it mentioned, it does use live web searches to verify data, so at least there’s that.
Final note before I move on to my recommendations:
This was a VERY limited review to gather only the most preliminary data. I only used ChatGPT, not Google’s Gemini, Facebook’s LLaMa, or X’s Grok… or any of the other LLMs that are out there learning and gaining influence over your supporters. My aim was not to provide sweeping truths but to take a snapshot of data and review it briefly.
Welcome Aboard, LLMs
In their study, Whole Whale refers to “what LLMs know about your organization” as your organization’s “AI Brand Footprint”. I’ll take it in a slightly different direction because I don’t think we need to be stomping on AI hard enough to leave a mark (kidding)…
I’ll call it “AI Training & Onboarding”.
What SEO is to your organization’s official website, AI Training & Onboarding is to your organization’s presence on alllllll of the web.
The more you do to get the right information about your organization disseminated* across the web, the more LLMs can know, understand as important, and present to users.
More sources of data = more statistical weight
AND, of course, getting the right information presented by the right sources (directories, charity evaluators, news outlets, etc.) is even better.
More data from already statistically-weighty sources = exponentially weightier weight
Three Ideas for AI Training & Onboarding for 2026
- Include AI T&O in your annual marketing plan by taking every opportunity to tell your organization’s story through reputable online sources. Think—giving days, resource directories, websites of well-known sponsors, etc.
- Keep local resource directory listings up-to-date with your mission and service offerings.
- Send out press releases to get covered by reputable news outlets and link to any resulting articles from your site (backlinking!).
* Yeah, I said disseminated because it’s a great word and even though we’re supposed to write at a 6th grade level, sometimes I just want to use multi-syllabic words for the beauty of them. Recapitulation. Haberdashery. Discombobulate. Extraterritoriality.
