AI discoverability for nonprofit websites

The new front door

For years, the homepage was the front door. Increasingly, the first impression happens somewhere you don’t control: inside an AI assistant answering a supporter’s question. If the assistant can read and trust your site, it brings people to you. If not, it sends them elsewhere.

That shift changes what “findable” means. A donor researching where to give, a parent looking for services, a reporter checking a fact: each may now get a summarized answer instead of a list of links. Your site becomes the source material for that answer: the assistant reads your pages, and the person asking sees its summary of them.

What AI tools need to “see” you

Each item is checkable. Could a stranger — or a machine — answer “what does this organization do, where, and for whom?” from the first screen of your About page? If that answer lives in a brochure PDF or an image, assistants can’t use it.

What makes a site machine-readable

Machine-readable sounds more technical than it is. In practice it means the page is organized so that a system skimming it can pull your facts out without guessing, and four things do most of that work.

Answer-first pages

Assistants quote short, self-contained passages. A page that opens with the direct answer (who the program serves, what it costs, how to apply) gives them something they can use. A page that spends its opening paragraphs on scene-setting usually gets passed over.

A real heading hierarchy

Headings are the table of contents machines use: one H1 that names the topic, H2s that describe each section honestly, and no bold paragraphs pretending to be headings. Decorative headings leave assistants unable to tell which passage answers which question.

Structured data

Schema.org markup (Organization, Article, FAQPage) restates in machine language what your page says in human language. It won’t rescue a vague page, but on a clear one it removes ambiguity about your legal name, your mission, and which question each section is answering.

Consistent facts

Assistants cross-check what they read. If your site, your annual report, and the directories that list you disagree about your name, service area, or programs, the machine hedges, or it leaves you out of the answer altogether. Pick one version of every core fact and use it everywhere.

Two kinds of assistants, two ways of being found

Not every AI tool finds you the same way, and the difference decides what to fix first.

Retrieval-based assistants (Perplexity, Google’s AI Overviews, and chat tools with browsing turned on) search the live web at the moment of the question, read a handful of pages, and compose an answer with citations. They reward what search engines reward: crawlable pages, plain titles, and answer-first writing. Improve a page today and they can reflect it on their next read.

Training-data assistants (ChatGPT and Claude in their default modes) answer from what they learned before a training cutoff. Their picture of your organization formed over years of published information, and it changes slowly. You influence it the long way: accurate, consistent facts on your site and everywhere else your name appears.

Most tools now blend the two approaches. In practice, page-level clarity pays off quickly with retrieval tools, while consistency builds up slowly in training data, so plan to work on both rather than choosing between them.

Five practical steps

Start with an audit of how assistants currently describe you. Then fix your organization and FAQ schema, rewrite key pages to lead with the answer, tighten your About page into an unambiguous statement of identity, and create a maintained source of truth for your core facts.

1. Run the audit yourself

Ask several assistants to describe your organization and to recommend nonprofits doing your kind of work. What’s wrong, missing, or outdated becomes both your priority list and your baseline for measuring progress.

2. Fix your organization and FAQ schema

Add or repair Organization markup with your legal name, location, and mission, and add FAQPage markup wherever you genuinely answer questions. Then run it through a validator and fix whatever it flags, because markup with errors in it can do more harm than having none at all.

3. Rewrite key pages to lead with the answer

Start with the pages supporters actually ask about: what you do, who you serve, how to get help, how to give. Put the answer in the first sentence, then get into the nuance and caveats after it.

4. Tighten your About page

This is the page machines lean on hardest to work out who you are. State your full name, where you work, what you do, and for whom. Say it plainly, near the top, in text rather than images.

5. Keep a source of truth

Maintain one internal document with your core facts: name, mission, locations, program names, leadership. Update your site and external profiles from it.

None of these steps involve gaming a system. They amount to stating what you do clearly and keeping your information organized and accurate, which is work your human readers benefit from just as much.

A starter checklist

If you do nothing else this quarter:

Little of this work is specific to AI, either. A site that states its facts plainly is also easier for the donors and families who read it themselves, however they happen to arrive.