When a language model is asked about an organisation it has not memorised, it goes looking for the page that says plainly what the organisation is.
That is almost never the homepage.
Homepages are written to make someone feel something. About pages are written to state facts. Only one of those survives summarisation.
What most nonprofit about pages are missing
The year it started. The place it operates. The number of people. What it actually does, in a sentence that would survive being quoted on its own.
These sound trivially obvious. They are absent more often than not, replaced by a paragraph about believing in a world where.
The rewrite
Put the plain facts in the first two paragraphs. Move the mission language below them.
Nothing is lost, because a person reading with intent scrolls. Something is gained, because a machine reading without intent does not.
Then add the same facts as structured data. It is a small block of markup and it removes the guesswork entirely.
How to check it
Ask a few different models what your organisation does. Where the answer is wrong, find the page that most likely produced it.
Usually the page is not wrong. It is vague enough that the gap got filled in by something else.
