A claim on this site should be traceable to the document behind it — by a reader who wants to check us, and by anything reading the page mechanically. That is a design constraint, and this page describes what follows from it: what each of our two quote stores records, what the schema markup does and does not carry, and which version of a bill an analysis was written against.
What this page will not do is tell you how any AI system or search engine treats a source. We have no way to check that from here, so we make no claim about it.
Primary Source Proximity
The difference between a citable claim and an uncitable one is whether the reader can get from it to the document behind it. "Senator X supports this bill" is an assertion. A verbatim floor statement, attributed to the member by name and party and dated, on a page that links to the bill it concerns, is a claim you can go and check.
What that looks like here, precisely. Every member quote is reproduced verbatim from the Congressional Record or discarded — each sentence has to appear in the fetched text or the quote is never written. The floor-quote collection records the GovInfo granule each statement came from on 430 of its 468 entries; the quotes on a bill's own page come from a second extractor that does not record a granule, and none of those 417 carries one. Every bill page links back to that bill on Congress.gov. The analysis is written against the newest applicable text version at the time it was fetched — enrolled, then engrossed amendment, engrossed, reported, and only then introduced — so a bill reported out of committee is read from the reported text, not the version as introduced. A bill that advances after it has been analysed is not automatically re-fetched, so the stored text can be older than the newest version Congress.gov now lists.
Structured Data: Telling Crawlers What the Content Is
Structured data is machine-readable markup embedded in a page, stating in a fixed vocabulary what the page is and what it is about. Every article and bill page on this site carries schema.org markup: Article on both, Legislation on bill pages (carrying the bill identifier, date, and sponsor), and BreadcrumbList for position in the site.
All of it resolves to a single identity. Rather than repeating an author block on every page, each article and bill page references one canonical Person node — the editor who reviewed it — and one NewsMediaOrganization node for the publisher, carrying explicit links to our publishing principles, fact-checking policy, corrections policy, and ownership. Anything reading any page on this site can resolve who is accountable for it and what standards it was produced under, rather than finding the same author block copied inline on page after page.
The two node types carry different things, and it is worth being exact about which. The bill page for S. 1884 carries a Legislation node with the bill's name, its identifier (S.1884), its date, its type, and its sponsor (John Cornyn). An article about that bill carries a LegislativeAction node in its about field with the name and the identifier — this one also carries the date, though most do not — and never a sponsor; that fact lives on the bill page, not on the article. Reading the article's markup tells you what the article is about; the sponsor is one link away, on the bill page.
Factual Density
A generic description ("the bill helps veterans") gives a reader nothing to verify. A specific one — S. 3971 extends SBIR and STTR grant programs through FY2031 and requires screening against eight federal watchlists for foreign adversary ties, signed April 13, 2026 — names five things that can each be checked against the text.
The drafting rules require it: lead with the dollar amount, percentage, threshold or deadline whenever the source states one, use the real figure rather than a round approximation, and where no number exists in the source, describe the mechanism instead of estimating. Every one of those figures is then checked against the fetched text. One the checker cannot find is flagged as a possible insertion from memory rather than from source. That flag warns rather than blocks, so clearing it — confirming the figure against the text, re-sourcing it, or recording the evidence in a tracked adjudication file — is a step a person takes, not one the pipeline takes for them.
No Editorial Spin: Why Neutrality Makes a Claim Checkable
A description written in one coalition's vocabulary is only usable by readers who already share it. Describing the same provision in the statute's own terms is usable by everyone, and is checkable against the statute.
That is what the "no editorial spin" policy is for. When the question is "what does FISA Section 702 authorize?", the useful answer is the mechanical one. Our analysis of H.R. 8322 describes the mechanism and leaves the judgment to the reader: "Section 702 of the Foreign Intelligence Surveillance Act lets the NSA collect the communications of foreign targets without a warrant — and when those targets contact Americans, those conversations are swept up too."
The Floor Activity Page: A Quote Database
Our Floor Activity page is built as a quote database rather than a feed. Each of its 468 entries carries the speaker's name, party and state, the chamber, the date the statement was recorded, and a link to the bill it concerns where there is one — 432 of them have a bill. What it does not carry is a session number — no quote record in either of our stores has one.
The point of building it this way is not traffic. It is that a specific, dated, attributed statement can be checked, and a general characterisation of what somebody said cannot.
Drafted from primary source documents by LegislationPatch's automated research pipeline, then reviewed, verified, and edited by James Shearn, Editor. Every figure and citation is checked against the official text before publication. See our editorial standards and AI disclosure.