The giveready daily learnings loop had been static now for about ten days, the self-learning loop I discussed when I automated the daily readout. Citation share: zero out of ten. Fifth day running. Ten charity questions put to an AI assistant, and on not one of them did the directory I have spent months building get cited, or even pulled into the list of sources it read before answering.
What did get cited: a children's mental health site, a couple of philanthropy directories, an aggregator listicle. Established names. Sites that have been around a while, that other people already link to.
One of the more attractive ideas of starting a charity B2A giveready.org platform was marketing to AI agents instead of marketing to people. It is a directory of about forty thousand nonprofits, built so that when you ask an AI assistant where to give, it can find the small charities that never show up in a normal answer. The bet underneath it was a tidy one. Do AI everything. Make the data so clean and so machine-readable that Claude, ChatGPT and Perplexity cite it when someone asks where to donate. I called it B2A, business to agents, and I retired the first part of that theory in public a month ago.
Part of what I liked about the bet, if I am honest, was that it let me skip the part of marketing I have always done and was tired of. No cold emails. No asking strangers for a link. No slow relationship work. Just better data than anyone else, and let the machines do the distributing. I even built a chunk of it around a large foundation grant whose whole premise is AI making charitable giving easier to find.
The number that will not move
Three months in, the one number that matters is still zero. And it is not the data holding us back. The pages are clean, structured, server-rendered, carrying all the right machine-readable detail. Everything an AI crawler is supposed to want is there.
The problem is authority. Perplexity does not answer from memory. It builds a list of sources from a web search index, mostly Bing's, and ranks what it finds by how trusted each site is. GiveReady is new. Almost nobody links to it. So it never enters the list in the first place. We are not losing a close call on the final answer. We are not in the search rankings where the answers are given. In other words, I'm building a platform for 2027 and beyond that is underpinned by domain authority from 2013.
I went looking for who was in the rankings instead. The sites beating us are not better built. They are older, linked-to, cited by other people. A directory that has existed for years. A mental health charity with a press history. An aggregator listicle that ranks for everything. Their advantage is not their data. It is that humans have spent years pointing at them.
What the machines actually reward
This has taken some time to sink in. The machines are trained on a web that humans built by vouching for each other. A link is one person saying this is worth reading. A mention is one person saying I used this and it helped. The AI did not get rid of any of that. It reads it, weighs it, and sits one step downstream of it.
So the dream of doing AI everything has a hole in the middle of it. You cannot automate your way to authority, because authority is other people pointing at you, and you still have to give them a reason to point. The machine I built for turns out to want the exact human signal I was trying to route around.
The daily change
So the daily work on GiveReady changes from today. I have been tinkering. Another robots.txt fix, another bit of structured data, another internal link, the small mechanical jobs that feel productive at eight in the morning and move the citation number not one inch. I learned that lesson once already on the marketing funnel and somehow had to learn it again here.
The new rule is small but it will keep me from doing the easy thing, building for machines, and force me back to doing the thing that only humans can do, or do better than the machine. Thirty minutes a day, every day, on human outreach. Email a charity that is already listed and ask it to link its GiveReady profile. Ask the surf-therapy orgs I actually know to mention us. Get the directory into the roundups that already rank, one editor at a time. One of the nonprofits in the directory is the fund we run in my son Finn's name, so the very first link I am asking for is one I can give myself. The rest I have to earn the old way.
It is slow. It is the work I built a machine to avoid. But the data is clear: the B2A marketing only keeps running if a human keeps building the authority underneath it. I made a product for the agents, and the agents handed me back the phone.
Thirty minutes, starting this morning, with the surf charities.
Monthly Revenues $11,000 | Clients 2 | Prospects (AI outbound agent now live) | Team: Me + part time Jan (CTO)
Day 96 of 365.