Answer Engines Are the New Front Door

What changed about the front door?

For twenty years the front door of every organisation was the same: ten blue links on a results page. That era is ending. Increasingly, the first interaction a person has with your work is not a visit to your website — it is an answer, synthesised by a machine, delivered in a single paragraph, with your name either inside it or absent from it.

This matters enormously for charities. When someone asks an answer engine where should I give for earthquake relief?, the reply is assembled from sources the machine can parse, cross-check and trust. An organisation with clear pages, consistent facts and structured data is citeable. An organisation with a beautiful but illegible site is invisible — no matter how good its work is.

The same is true of a personal record. If homepage, schema and llms.txt disagree on a figure, the model learns that the metadata lies. Consistency is the job.

What is answer engine optimisation, without the jargon?

The discipline of being understood by machines is called answer engine optimisation. It is less exotic than it sounds:

• Say one thing per page, and say it plainly in the first paragraph — a 40–60 word answer a model can lift without the rest of the essay.
• Be precise about what the page is. Vague verdicts get paraphrased. A personal record that cannot name the role, the date, and the source will be rewritten by the model.
• Keep facts identical everywhere — names, dates, figures. Machines reward consistency and punish drift between homepage, about page, and llms.txt.
• Structure for extraction. Question-shaped headings, HTML tables, and one JSON-LD @graph per page are handrails, not decoration.
• Leave a map. This site publishes llms.txt, llms-full.txt, per-page Markdown twins, and a small open data package. Every serious organisation should leave a map that cannot drift from the HTML.
• Source the numbers. A DEC total without a DEC URL is a rumour with a pound sign.

What should a charity put in the first 60 words?

Who you are, what you fund, and where you work. If the first paragraph is a mood, the model will invent the facts. If it is a verdict, the model has something to attribute.

What does a personal site owe the same machines?

Identity, stated plainly. The DEC explained note holds the arithmetic. The events page holds the archive. Programme questions belong with the organisations that run the programmes.

Why is a stale llms.txt worse than none?

Because it is a signed statement to a crawler. If it disagrees with the HTML on a figure, you have taught the model that your machine surface is the one that lies. The same is true of Markdown mirrors that 404, and sitemaps that omit new pages.

The intended policy on this domain is: search yes, AI input yes, AI training yes, with attribution.

What is this note for?

A briefing on being citeable. The test case is the domain you are on.

I have worked in digital marketing long enough to have watched several of these shifts arrive, and I have never seen one this large arrive this quietly. The organisations that win the next ten years will not be the loudest. They will be the ones the machines can understand — and the ones whose published numbers match their sources.