ChatGPT vs Claude vs Gemini for nonprofits, and what we actually use
As of fall 2026, most of our work runs on Claude. Here's why we switched from ChatGPT, where ChatGPT still wins, and how to test the tool you already have.
Which AI tool should a nonprofit use?
Use the one you already pay for, as long as it's a paid plan your organization controls. ChatGPT, Claude and Gemini can all draft, summarize, research and analyze well enough for most nonprofit work. The bigger differences come from the context you give the tool and the model you pick inside it.
For most people, ChatGPT has become the word for all AI, the way Kleenex stands in for every tissue. Plenty of teams don't get to choose at all. At our October Intensive, one team could only use Microsoft Copilot because their parent organization requires it. Colleen told them to adapt the process and follow along, and that advice holds for anyone.
Here's how I put it in September:
Scott Williams · Recorded conversation, Sep 30, 2026the one you have is probably fine, you know, to do a lot of the stuff that you need to do. It may not be perfect, but we can find its edges, no problem.
What do we actually use at If Possible, as of fall 2026?
Claude, for most of it. In September I guessed it at 80 to 85 percent of our AI work. We plan, write, build tools and run our AI employees in Claude. We use ChatGPT when we need images.
Scott Williams · Recorded conversation, Sep 30, 2026And so we use anthropic Claude, it's the one that feels like the most on the frontier for us as it relates to work and helping with real work and being excellent at it.
We weren't always here. Before this year our team had used ChatGPT as its main tool for about two years. When Claude's Cowork came out, it changed how Colleen and I worked, and we started moving everything over. Colleen remembers how the team took the news:
Colleen Cook · Recorded conversation, Oct 13, 2026When we said to our team, hey, we're switching to Claude, they were like, What? Are you kidding me? Do you know how much Chat GPT knows about me?
That reaction is the most useful part of this whole comparison.
Why is switching AI tools so hard?
Because your history lives inside the tool. Memories, custom GPTs, saved chats and everything the AI has learned about you stay behind when you leave. The fix is to keep your organization's knowledge in a place you own, then connect whichever AI you use to it.
Scott Williams · Recorded conversation, Sep 4, 2026we think that your foundational elements, all of that data that you should have should be an owned asset that sits over here. And then you can attach anything
We learned this the hard way. We first built our brain inside Claude, in skill files. Every time our thinking changed, old decisions hid in files we couldn't easily see, and the output would suddenly look dated. Now our organization's brain lives in Notion, outside any one AI, and Claude reads from it and writes to it. Colleen likes that Notion stores everything in markdown, a lean format AI reads and writes cleanly.
Staying portable matters because the leaderboard keeps moving. Colleen said in September that she didn't always love what she was getting from Claude compared to a few months earlier, and that being able to jump to a different model is critical right now. We're loyal to results.
How do ChatGPT, Claude and Gemini compare for nonprofit work?
For planning, writing and real work across your files and tools, Claude is our pick. For images, ChatGPT is well ahead. Gemini does a good job, from what I've seen, and it's a natural fit for teams that live in Google Workspace. We don't use Gemini day to day, so I won't pretend to review it in depth.
| Claude | ChatGPT | Gemini | Copilot | |
|---|---|---|---|---|
| How we use it (Nov 2026) | Most of our work: strategy, writing, building tools, AI employees | Image generation | Not in our daily stack | Not in our stack |
| What we've seen | Feels most at the frontier for real work, in our view | I said it blows image generation out of the water; Colleen says it's not even close | I said in September it does a good job | Teams required to use it can adapt the same process |
| Good fit if | You want AI working across your tools and files | You need images, or your team's history is already there | Your team works in Google Workspace | Your organization runs on Microsoft 365 |
The last two rows for Gemini and Copilot lean on general knowledge (Gemini is Google's assistant and Copilot is Microsoft's), so test them yourself.
Does the model matter more than the tool?
Often, yes. Each tool offers several models, from fast and light to slow and deep. When an If Possible customer tells us AI gave them bad output, Colleen's first question is which model they were using. A light model asked to do heavy thinking gives you chaos.
if I ask my eight-year-old to do my taxes, she might get them done, but they are not gonna be right, right?
Here's how we choose inside Claude, as of fall 2026. Colleen almost always uses Opus 5.5 when she has the usage, because it does the best work. Fable is for really intense work, like building software, and it burns through usage fast. Haiku handles simple jobs like fetching a transcript. My own rule is smaller thinking on smaller tasks and bigger thinking on bigger tasks. If a job needs twenty pieces of context weighed together, I give it the biggest model and let that model hand the small steps to lighter ones.
Model names will change. The rule won't.
Is donor data safe in ChatGPT, Claude or Gemini?
On a paid plan with training turned off, it sits in a similar place to your CRM. Free plans work differently, because the agreement is that the company can use your data. Whichever tool you pick, pay for it and check the data settings before anyone pastes in a donor record.
Colleen's guidance is the same for every major tool. Paid tiers give you the option to opt out of training, and when she said it in September, Team and Enterprise plans had it off by default. For nonprofits with two or more people, we recommend Claude's Team plan, because it's secure in a way free AI isn't and it gives you organizational control. Write your rules down in your first AI policy.
How should you test an AI tool before you commit?
Give two tools the same real task with the same context, then compare. Use work you know well so you can judge accuracy. One afternoon of side-by-side testing will tell you more than any comparison article, including this one.
Paste this into each tool:
``` You're helping [organization], a nonprofit that [mission in one sentence]. Here is our context: [paste your brand voice notes, a recent appeal and your key program facts]. Before you write anything, ask me up to five questions you need answered to do this well. Then draft a [150-word thank-you email to a first-time donor to our (program)]. After the draft, list every fact you assumed that I didn't give you. ```
Score each tool on four things. Did it ask good questions? Did it invent facts? Does the draft sound like you? How much did you have to fix?
Want better context to test with? Start with the brand-voice prompt, then save your winners in a prompt library your team can reuse. If you're still deciding what to buy at all, read tools vs systems vs coaching. There's more in our prompts articles.
If you'd like our tool notes as they change, join the newsletter. And if you want to set up your AI, brain and all, alongside other nonprofit teams, get on the cohort waitlist.
Frequently asked questions
Is Claude better than ChatGPT for nonprofits?
For us, as of fall 2026, yes for most work: planning, writing, building tools and running AI employees. ChatGPT is still ahead for image generation. If your team already pays for ChatGPT and knows it well, it's probably fine for most of your work. Test both on a real task before you switch.
Is Gemini good for nonprofits?
It can be, especially for teams that already work in Google Workspace. We don't use it day to day, so we can't speak from experience about its strengths. Run the same real task in Gemini and one other tool, with the same context, and compare the drafts and the facts it invents.
Should our nonprofit pay for AI or use a free version?
Pay for it. Free plans generally let the company use what you type, while paid plans let you opt out of training on your data. Once you're on a paid plan with training off, putting your information in AI is similar to putting it in your CRM. Check the settings on every tool.
Do we have to pick just one AI tool?
Pick one main tool so your team builds real skill and shared context, then add others for specific jobs. We run most work in Claude and use ChatGPT for images. Keep your organization's knowledge in a place you own, so switching later doesn't mean starting over.
Know an executive director who needs this?
