Machine-Readability of Official U.S. State Statutory Sources — October 2026

Version 2026.2 · Published 2026-10-14 · Coverage 2026-10-14/2026-10-14

The second reading of a first-party, reproducible measurement: can an identified, non-JavaScript HTTP client retrieve the text of one specific citable statutory section from each U.S. jurisdiction's own official online source on a single request? Of 47 jurisdictions with an official source URL, 26 returned the cited section and 21 did not (25 of 47 in the August 2026 reading). Failures by class: section text absent 5, pdf only 4, blocked by server 3, js shell 3, tls certificate error 2, dns resolution failed 2, not found 2. 9 jurisdictions changed class since August with the same section URL: California (not_fetched_robots_disallow → blocked_by_server); Connecticut (connection_failed → tls_certificate_error); Illinois (not_found → tls_certificate_error); Iowa (pdf_only → dns_resolution_failed); Michigan (connection_failed → blocked_by_server); New Hampshire (connection_failed → machine_readable); Tennessee (connection_failed → not_found); Vermont (connection_failed → machine_readable); Wisconsin (machine_readable → dns_resolution_failed). This version splits failed connections by their network error class (an incomplete TLS certificate chain, a failed DNS lookup) so that a failure is attributed to what actually failed, and it moves both controls to govinfo.gov after the previous control host began answering missing pages with HTTP 200.

Downloads

• Per-jurisdiction outcomes and HTTP evidence (CSV) — text/csv, 11,874 bytes, SHA-256 d9dcbd8df0c41c6b7f5134e28af15be35aad5345531271c5e6d45c2ef0308ca1
• Full audit record including method, controls, counts and limitations (JSON) — application/json, 40,934 bytes, SHA-256 439973cd423052ef20913f94a2095a988511fcc28a303e48f39bec1d92fa3a29

Plain-text citation · BibTeX · Machine-readable release metadata

Methodology

1. One personal-injury limitation-period section was chosen per jurisdiction (50 states and the District of Columbia) as the retrieval target, and the URL of that section on the jurisdiction's own official statutory website was recorded in advance.
2. For each jurisdiction, robots.txt was fetched FIRST from the section URL's host and parsed for the wildcard user-agent group. Where robots.txt disallowed the path, the section was NOT requested and the jurisdiction was recorded as not_fetched_robots_disallow.
3. Where robots.txt permitted it, the section URL was requested exactly once with a GET, following redirects, using an identifying User-Agent that carries a contact URL, with a 25-second timeout.
4. A second request was made to the scheme-and-host origin of that same URL, derived mechanically from it and never authored separately, so that a failure at the section URL could be distinguished from a host that does not serve an identified client at all.
5. Responses were classified in this order: connection_failed (no response), blocked_by_server (401/403/429), not_found (404/410), server_error (5xx), pdf_only (a PDF body, never parsed), js_shell (HTTP 200 carrying fewer than 400 characters of text after markup removal), machine_readable (the cited section identifier appears in the extracted text), and section_text_absent (HTTP 200 with substantial text in which the cited identifier does not appear).
6. HTML entity decoding places dash-like entities before the generic numeric-entity sweep, so that a section rendered with a non-breaking hyphen is not silently mangled into a different identifier before it is searched for.
7. Two controls were run before the audit and both had to pass or nothing would have been written: a positive control on a known-good federal source outside the audited set (28 U.S.C. § 1332 on govinfo.gov, the Government Publishing Office's official U.S. Code), which had to classify machine_readable, and a negative control on an impossible path on that same host, which had to classify not_found.
8. At least five seconds separated consecutive requests, each URL was requested once, no form was submitted, and no block, rate limit or robots exclusion was circumvented.
9. All counts in this release are computed over official_primary sources only. Four jurisdictions for which no official source URL was available to the auditor are retained in the data with source_class marking them and are excluded from every total.
10. Changes in version 2026.2, applied only to this reading: (1) a failed connection is split by the network error the client received: tls_certificate_error where the server's certificate could not be verified (most often an incomplete certificate chain that ordinary browsers repair on their own), dns_resolution_failed where the host name did not resolve, and connection_failed for anything else; all three remain not machine-readable. (2) An HTTP 200 response whose final URL is an error path, or whose page title says the page was not found, is classified not_found (a "soft 404"); on this date that rule changed no jurisdiction's class and applied only to the negative control. (3) The controls moved from uscode.house.gov, which now answers the old control URL and any impossible path with HTTP 200 "Page Not Found", to govinfo.gov.
11. A robots.txt that could not be read because the host refused the request is recorded as no_robots_txt, and the section was then requested once, as RFC 9309 permits when robots.txt is unavailable; where the section request was refused too, the row is blocked_by_server.
12. Comparison with version 2026.1 counts only official sources whose section URL is unchanged (0 changed). The 2026.1 files are immutable and were not re-classified under the new rules.

Coverage and limitations

• The section URL for each jurisdiction was selected by the auditor. A different URL may succeed where the audited one failed. The origin stage is included precisely so that a section-level failure can be separated from an unreachable host, but it cannot rule out the existence of a better URL.
• This measures single-request retrieval by a non-JavaScript client. It is NOT a measure of whether the law is publicly available, whether a person using an ordinary web browser can find and read it, or the quality of any state's website. A person with a browser can reach all 51 jurisdictions.
• PDF-only sources are counted as not machine-readable for this purpose and are deliberately not parsed. The statutory text in them is public and readable by a person; it is single-request automated retrieval that fails. Parsing a compressed PDF stream and then pattern-matching the result risks a false positive, which would be worse than an honest gap.
• robots.txt outcomes are recorded as observations and were obeyed. A disallow is a publisher's legitimate choice and is not scored as a defect.
• A jurisdiction classified section_text_absent may be publishing the section at a different URL, rendering it on the client, or presenting an index page. This instrument returns HTTP evidence and does not distinguish among those causes.
• These are single readings taken on a single date. Websites change, and a host that failed to connect or resolve may be healthy an hour later. On this date the two dns_resolution_failed rows (Iowa, Wisconsin) were confirmed from three independent resolvers (the auditor's network, a second server, and Google Public DNS, which reported a failure at each host's own authoritative name servers), but they are still a single night's reading. Any reuse should re-run the measurement rather than treat this file as a current status.
• Four jurisdictions (Arkansas, Georgia, Mississippi, New Jersey) are recorded against a third-party source because no official source URL was available to the auditor, and are excluded from all totals.
• A tls_certificate_error means a standards-conforming client that does not fetch missing intermediate certificates on its own could not verify the server. Many browsers do fetch them, so a person may reach the page normally.

Data dictionary

Sources and provenance

• United States Code, 28 U.S.C. § 1332 (Government Publishing Office, govinfo.gov) — positive control for this release (official_primary, retrieved 2026-10-14)
• A.R.S. § 12-542, Arizona official statutory source — an audited official source that returned the cited section (official_primary, retrieved 2026-10-14)
• LawyerLand statute-source availability audit run, 2026-10-14 — instrument, controls and per-jurisdiction HTTP evidence (first_party_measurement, retrieved 2026-10-14)
• The previous reading: Machine-Readability of Official U.S. State Statutory Sources — August 2026 (version 2026.1) (first_party_measurement, retrieved 2026-10-14)

License and citation

This release is available under Creative Commons Attribution 4.0 International. Attribution identifies the source; LawyerLand does not require a followed hyperlink as a condition of access.

LawyerLand (2026-10-14). Machine-Readability of Official U.S. State Statutory Sources — October 2026 (Version 2026.2) [Data set]. https://www.lawyerland.com/research/state-statute-source-availability-2026-10/

Version history

This is version 2026.2. Prior version directories remain immutable. Material corrections receive a new version and release-feed entry.