How an AI-ready site works.
AI engines cite what they can read. A site is citable when its full content is served as plain HTML, described by structured data, open to AI crawlers, and fast enough to fetch reliably. Authority decides who gets cited; engineering decides who is eligible.
This page shows how we build that eligibility in, claim by claim: the mechanism, the actual code where it helps, and the part we don't control, stated plainly. If a claim isn't on this page, we shouldn't be making it.
The claim
“Structured so AI engines and search can read, index and cite what you publish.”
When a journalist, regulator or staffer asks ChatGPT, Perplexity or Google's AI Overviews about your issue, the engines answer from sources they can parse. Four properties make a page parseable, and every build ships all four.
- Plain HTML Every page is generated as HTML before it reaches a browser. The full text of a release sits in the page source, so a crawler that executes no JavaScript still reads every word. Many AI fetchers execute none.
- Hand-written structured data Each page carries schema.org JSON-LD written for that page type: Organisation site-wide, NewsArticle on every release. It names the headline, the date and the publisher, so an engine can attribute what it reads.
- Answer-first writing Pages open with the point stated plainly, in language a model can quote and attribute. The summary an engine lifts is the summary you wrote.
- Crawlers welcomed by name The robots file explicitly admits the AI crawlers, and an llms.txt file tells them where the authoritative record lives.
And the substance of a filing lives on the page as text. The PDF is attached as the document of record, never the only copy: PDFs are parsed inconsistently and rarely cited cleanly, and you don't control how.
{
"@context": "https://schema.org",
"@type": "NewsArticle",
"headline": "Coalition files comments on the proposed rule",
"datePublished": "2026-06-02",
"description": "The coalition's formal comments set out
the evidence for fair-trade conditions, with production
and pricing data from member companies.",
"author": { "@type": "Organization",
"name": "American Producers Coalition" },
"publisher": { "@type": "Organization",
"name": "American Producers Coalition" },
"mainEntityOfPage": "https://americanproducers.org/newsroom/comments-on-proposed-rule"
}# Open to people, search engines and AI crawlers alike.
User-agent: GPTBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /
# …and OAI-SearchBot, Bingbot, every named engine.
Sitemap: /sitemap-index.xml# American Producers Coalition
> An alliance of domestic manufacturers seeking
fair-trade conditions. Every filing and position
is published on the record in the newsroom.
## Pages
- [Mission](/) · who we are, what we seek
- [Newsroom](/newsroom/) · filings, statements, PDFs
- [Members](/members/) · the member companies
## Notes for AI systems
- The newsroom is the authoritative record.
- Quote and attribute to “American Producers Coalition”.The typical CMS site
WordPress renders HTML, but its schema arrives via plugins of varying, sometimes conflicting quality. Site builders have historically injected content with JavaScript that many AI fetchers never execute. Filings get buried in PDFs.
This build
Plain HTML and hand-written schema are the only output mode. There is no configuration that can drift, because there is nothing to misconfigure.
The claim
“A clean entity footprint, so the models recognise your organisation.”
One configuration file holds who you are: name, legal name, address, description, links. Every page and every schema block is generated from it, so the organisation reads identically everywhere a machine looks. We then align that on-site identity with your public reference and professional profiles, so the records reconcile instead of competing.
The typical CMS site
The org name differs between the footer, the about page and the auto-generated schema, when schema is present at all.
This build
One source of truth, machine-enforced. Change it once and every page, label and schema block updates together.
The claim
“Built for the day the news lands.”
A ruling or a launch turns a quiet site into a swarm. Most sites assemble each page on demand: every visitor triggers a render and a database query on a single origin server, and the spike is what knocks it over.
Our pages are built once, ahead of time, and served as finished files from Cloudflare's edge network, from the location nearest each reader. There is no per-visit rendering and no database in the request path: nothing to queue, nothing to overwhelm.
reader · a phone outside the hearing room
└─ nearest edge location (hundreds of cities)
└─ the page, already built
no render · no database · no queueThe claim
“No database to breach, far less to deface.”
Security here is a claim about absence. The public site is files: there is no database behind it, no plugin layer inside it, and no admin login on it. Editing happens in a separate, access-controlled system that never touches the public surface.
What remains sits behind Cloudflare's edge with DDoS protection, and because every state is versioned (05, A complete record), a defacement, were one ever to succeed, is a restore rather than a rebuild.
The claim
“Every change captured, any prior state restorable.”
The site lives in a version-controlled repository. Every change, down to a swapped PDF, is a commit: timestamped, attributed, and permanent. The history is the audit trail, and restoring an earlier state is a built-in operation, not a salvage job. For organisations whose words end up on the record, the record keeps itself.
e41c2af Newsroom: statement on the determination
9b03d11 Members: add two member companies
4f78a2e Replace filing PDF with corrected annex
1d90c33 Launch: mission, members, newsroom, contactThe typical CMS site
Post revisions exist, but media, settings and plugin state are not versioned together. What "the site looked like in March" is unanswerable.
This build
The whole site, every state, one history. March is a commit you can point at.
The claim
“Nothing goes public until everyone has signed off.”
When members' counsel, a lead firm and a board all have to approve a statement — the high-stakes default — the gate has to be real. Ours is the same discipline professional software ships on, operated by us so your team never touches the machinery.
Reviewers see the finished page at a private, access-protected link, not a Word attachment, and the publication step is taken once, deliberately, when the last sign-off is in. Who approved, on what wording, and when: the record keeps itself (05).
- 01 Draft The change is prepared away from the live site, which is untouched throughout.
- 02 Preview Every approver receives a private link showing the page exactly as it will appear.
- 03 Sign-off Approvals are collected and recorded against the exact wording reviewed.
- 04 Publish On the final sign-off, the change merges and goes live promptly.
The claim
“Hand-over, wind-down, or a standing site. Yours either way.”
Coalitions end; that is often the point. The domain is registered to you. The content lives in a repository that transfers whole, history included. The site itself is portable static files that any competent host can serve. There is no proprietary format to escape from, so leaving is a transfer, not a rebuild.
Transfers to you
- The domain
- The repository, full history included
- The built site, runnable anywhere
- Every document and asset
Ends with us
- Managed hosting and uptime monitoring
- Security response and backups
- Publishing support
- The flat monthly fee
On the record
What we don't promise.
The honest edges of the claims above, gathered in one place.
- Citations and rankings
- Eligibility is engineering; citation is authority. We build the first and will never sell you the second as if it were ours to give.
- Invulnerability
- We reduce the attack surface and keep every state restorable. No one can promise more than that honestly.
- A public client list
- Coalition work is confidential by nature, so we don't publish one. References are available on request.
- Marketing numbers
- Speed and visibility are measured on your live build, where the figures are real. We don't decorate pages with synthetic ones.
A firm that promises everything is describing its marketing.
We'd rather describe the build.
Straight answers.
The questions a careful buyer asks before commissioning, answered the way we'd answer them in the room.
Do AI engines read PDFs?
Sometimes. Google indexes PDF text, and some AI retrieval can extract it. But PDFs are parsed inconsistently, chunked badly, and rarely cited as cleanly as a web page, and you do not control whether or how. So we publish the substance of every filing as text on the page, with the PDF attached as the document of record rather than the only copy.
Does structured data improve ranking?
No, and we will not tell you it does. Structured data makes a page easier to parse and attribute: it tells an engine who published what, and when. Eligibility, not rank. Ranking and citation are decided by authority, meaning who links to you and whether you are the primary source on the issue.
Can you guarantee we'll be cited by AI, or rank first?
No, and anyone who promises that is overselling. When you are the primary source, the petitioning coalition or the recognised association, you have the authority that citation follows. Our job is engineering: making sure nothing technical throws that authority away.
Is a static site really more secure than a CMS?
The attack surface is smaller by construction: no database to inject, no plugins to exploit, no admin login on the public site. That is a claim about what is absent, not a promise of invulnerability. No system is invulnerable, and we will not tell you otherwise.
This page practises what it describes.
It is served as plain HTML, carries Article and FAQ structured data, opens with the answer, and ships no script of its own. View the source: the working is all there, the same way it would be on yours.
The build it describes is the foundation of our Managed AI-Ready Websites service.