I had been quoting the same number for a month. Every morning GiveReady asks ChatGPT, Claude and Perplexity ten charity questions and writes down which websites they cite. One line in that tracker kept coming back zero: the number of times an AI found our page and then chose someone else. I loved that zero. It meant the whole problem of AI visibility for charities came down to being found, and being found is something you can fix.
This morning I had a brief ready for the biggest page we have put on GiveReady, and that zero was in it, doing a lot of the work. Before writing the page I asked Claude to check every number in the brief against the raw files.
It came back with this. Perplexity and ChatGPT hand back one list of websites, and that one list is both what they searched and what they cited. So "found us but chose someone else" could never be anything but zero for those two. The zero came from how I'd built the measurement. Only Claude reports what it retrieved separately from what it cited, and in 24 readings it has retrieved a GiveReady page zero times. Also, Claude and ChatGPT had only been measured 24 times each, against 324 for Perplexity. "Three engines, daily" was a generous way to describe it.
That stung a bit. It was in the daily digest and in this morning's brief.
Why GiveReady is changing again
This is the second time I have changed what GiveReady is in public. The first one was in May, when I retired the idea that AI agents would find the directory and fill in the missing data themselves. They read it and never wrote anything back.
What changed my mind this time came from the other project. A few days ago I wrote about the Google ad grant for the fund: ten thousand dollars a month in free ads, zero applicants, while half the real applicants came from ChatGPT. Since then, four applicants to the Finn Wardman World Explorer Fund have told us directly that they found it by asking an AI assistant. They asked a chatbot how to pay for a trip, and it sent them to us. About one in six ChatGPT visitors to the fund's site went on to apply, roughly 17%.
If that is happening to a tiny grant fund, I'd bet it is happening to plenty of small charities, and most of them have no idea whether ChatGPT knows they exist.
So GiveReady is pointing at them. For the next few months it stays a directory built for AI to read, 41,000 charity profiles, and the job becomes getting charities to claim and correct their profile so AI finds the right version of them. The x402 payments the directory was originally built for are parked until agents actually start giving money away, which they aren't yet. The go or no-go date is 23 November. If nothing external has moved by then, GiveReady goes into maintenance and the hours come back to client work, and we use it to help the World Explorer Fund to get found.
What AI visibility actually means for a charity
People call this answer engine optimisation, AEO, which is SEO for the AI answer at the top of the page rather than the ten blue links underneath. When someone asks an AI "what charities help young people in Cape Town", the AI runs a web search, reads what comes back and writes an answer from it. If your pages aren't in what comes back, you can't be named. Each engine searches its own index, too. Our Cape Town guide was cited by Perplexity 8 times out of 8 and by ChatGPT 0 out of 8 on the same day.
Today I also counted which websites the AI answers cited. Across 372 AI answers to questions about youth charities, the site cited most often after GiveReady was the Charity Commission register, in 91 of them. Then LinkedIn, GreatNonprofits, GlobalGiving, GoFundMe and gov.uk. None of them is a charity's own website. For a small charity, start with the register. Make sure the entry is current and reads well to a stranger, because that is the text an AI is most likely to quote.
What did work was the guides. Perplexity cited GiveReady in none of our ten daily questions for the first three weeks of the tracker. We published short, checked guides, each one with the charity's registration number, readable without JavaScript. The first citation came about two weeks later. Since the start of August, Perplexity has cited GiveReady in 309 of 460 answers. Claude, which searches using Brave's index, has cited us in none, and I haven't fixed that yet but I just got started on it, and likely it's due to lag time in their indexing.
What shipped
The page went live on GiveReady today: How to Get Your Charity Found by AI. Five steps, the tables, a section on exactly how we measured and what our numbers can't tell you. It has a free AI audit form embedded after the self-check. The audit asks ChatGPT, Claude and Perplexity 19 questions about a charity and sends back a one-page report within two working days. Requests from that page are tagged, so I will know if the page brings in anyone at all.
The home page nav went from seven links that jumped around the page to four that go somewhere.
If you run a small charity and want help setting this up, it's what I do at TestVentures, at half price for youth nonprofits. The audit on GiveReady is free either way.
Changing course when people are watching
For anyone reading this who is halfway through rebuilding a career: changing direction in public is less dramatic than it sounds. You say what you thought, what the numbers said, and what you're doing next. The only hard part today was admitting the zero I liked so much was one I had built in myself.
The page will take weeks to be picked up by the engines. The go or no-go date is 23 November, 59 days from today.
Day 181 of 365.