Dale Nirvani Pfeifer
← All insights

November 21, 2026 · 5 min read

What do ChatGPT, Claude and Perplexity tell donors about your nonprofit? How to check

More donors ask AI assistants which charities to support and whether one can be trusted. Here is how to see what they hear about you, the questions to ask, and what to look for.

Donors used to search for you. Now many of them ask. "Is this charity legitimate?" "Who runs it?" "How do I donate?" "Which nonprofits help families in my city?" The answer comes back as a few confident paragraphs from ChatGPT, Claude, Perplexity or Google's AI, and the donor often never visits your website at all.

So it's worth knowing what those answers say. But checking is harder than it looks.

Why your own search isn't a reliable guide

There is no single answer to "what does ChatGPT say about us?" Every answer depends on:

  • Who's asking. Assistants use your past chats, saved memories and, on some accounts, search history. Yours has probably heard a lot about your organization from you. A donor's hasn't.
  • What's connected. If your assistant is connected to your email, documents or calendar, it may answer from those. A donor's can't.
  • The exact words. "Is [name] legit?" and "Is [name] a trustworthy charity?" can get different answers. So can your organization's full name and its short name.
  • What it finds that day. Assistants search the web as they answer, and the pages they pull change from one try to the next.

So when you ask about your own organization, you're seeing your answer, not your donors'. It may look better than theirs, or worse. It's a starting point, not a measure.

The best you can do by hand is get close to what a first-time donor sees, and ask enough times to see the pattern. Here's how, in about an hour.

First: ask the way a donor would

This step matters most, and it's the one people skip.

To see what a donor sees:

  • Sign out, or use a fresh account. ChatGPT and Perplexity can be used signed out. Others, including Claude, need an account: use a free one that has never asked about your organization, with memory and connected apps (email, documents) turned off. A private or incognito window helps, but only if you're signed out in it.
  • Don't add context. Ask the bare question, exactly as a stranger would.
  • Ask each question more than once. Answers change from one try to the next. If an assistant gets something right once and wrong once, note both.

The questions to ask

About you, by name. These show whether assistants know you and get the facts right:

  1. What is [your organization] in [your city] and what do they do?
  2. Is [your organization] a legitimate and trustworthy charity?
  3. How do I donate to [your organization]?
  4. Are donations to [your organization] tax-deductible? What is their EIN?
  5. Who runs [your organization]?
  6. What has [your organization] accomplished recently?
  7. How can I volunteer with [your organization]?

Without your name. These show whether you're found by donors who don't know you yet:

  • What are the best nonprofits working on [your cause] in [your city]?
  • I want to donate to help with [your cause] in [your city]. Where should I give?
  • Where can I [what your program offers, in a donor's words] in [your city]?

What to look for

Read each answer as a donor would, and note:

  • Is it about you at all? Assistants often answer about a different organization with a similar name. See When AI mixes up your nonprofit with another one.
  • Are the facts right? Your website address, donate link, city, leader, EIN and what you actually do.
  • Does it hedge? "I couldn't confirm their tax-exempt status" or "check before giving" can cost you gifts even when everything is in order. See Why AI says it can't confirm gifts to your nonprofit are tax-deductible.
  • Does it warn donors? A low rating, an old controversy, a deficit. These are usually true, so the fix is to tell your side clearly, not to hide them.
  • Is the leader current? See Why AI names the wrong leader after a leadership change.
  • Which pages did it use? Most assistants show their sources. Those pages are where your fixes go: your own site first, then Candid, Charity Navigator, directories and news.
  • For the questions without your name: were you named? If not, who was, and where did the assistant find them?

Then what?

Fix what's wrong at the source, starting with your own website: one clear About page, a giving section with your EIN, your current leader, and the same plain description everywhere. Then work outward to your profiles. Changes can show up within days in assistants that search the web live, and may take a few months elsewhere. See Which profiles AI reads about your nonprofit for where to go next.

If your answers seem to ignore your website entirely, check that AI tools can read it. Some security settings turn them away. See Should AI tools be allowed to read your nonprofit's website?.

Or let us do it

The free Nonprofit AI Visibility Check asks these questions and more of four assistants, the way a first-time donor would: no history, no connected accounts, the same wording every time. It asks again whenever an assistant couldn't find something, so one unlucky answer doesn't become a finding. It checks every answer against your website and IRS record, and gives you a fix list in order, with the exact steps for each one. It takes a minute to start and about ten minutes to run.