The 9pm Test: Spanish-language intake signals on 308 US personal injury law firm websites, September 2026

Of 107 readable forms, 10 asked a qualifying question and 9 carried a Spanish promise. Median score 2 of 7; highest 5 of 7, by 2 firms; none reached 6.</p> <p><strong>The seven signals.</strong> (1) a page actually written in Spanish; (2) a phone number that dials when tapped, which means a tel: link; (3) a Spanish lead form with a sentence saying what happens after you press send; (4) the site says it takes contact after hours AND names a Spanish speaker, both on the same site; (5) the "typically replies" badge on the firm's public Facebook Page; (6) paid-acquisition markers in the page source; (7) the form asks at least one question about the accident.

Each signal is pass, fail or unknown with an evidence line, the URL and the date. Unknown is never counted as a fail.</p> <p><strong>Limits, stated plainly.</strong> Signal 5 returned unknown for all 308 firms, because Facebook does not put the reply badge in a Page's static HTML for most Pages, so the practical maximum on this run was 6 of 7 rather than 7 of 7. This is the public half of the question: it measures what the pages promise, not whether anyone answers.

Nothing was called, no form was submitted, no message was sent, no accident was invented, nothing behind a login was touched. 27 of 308 home pages could not be fetched with an ordinary browser request (bot wall, HTTP 403, 500, 404, 307 or DNS failure) and are reported as unreachable, never as failures, and left out of every share. Signal 6 is unknown where only a tag manager is present. Signals 3 and 7 are unknown on the 64 readable sites where no form could be read from the page source.

The source lists were compiled from public directories and firm pages and are not a census of either metro.</p> <p><strong>Ethics.</strong> Aggregate only. No firm is named for failing, here or anywhere else this research is published. The method names firms only when they score 6 or 7, and no firm did on this run, so these files contain no firm names at all.</p> <p><strong>Method and study pages.</strong> Method and definitions: <a href="wealthmarketing.ai/9pm-test">https://wealthmarketing.ai/9pm-test</a>.

Los Angeles study: <a href="wealthmarketing.ai/research/9pm-test-los-angeles">https://wealthmarketing.ai/research/9pm-test-los-angeles</a>. Houston study: <a href="wealthmarketing.ai/research/9pm-test-houston">https://wealthmarketing.ai/research/9pm-test-houston</a>.</p> <p><strong>Files.</strong> 9pm-test-los-angeles-2026-09-15.csv and 9pm-test-houston-2026-09-15.csv hold every count behind the two study pages; 9pm-test-both-metros-2026-09-15.csv concatenates them.

Columns: metro, crawl_date, section, item, value, denominator, note. README.md carries the full method, the codebook and the limits.</p> <p><strong>Citation norms.</strong> Anyone may run and publish a 9pm Test. Two conditions exist to protect the businesses being measured rather than the author: publish the date every signal was checked alongside the score, and name only what passes.</p> <p><strong>Also deposited at Zenodo</strong> with DOI <a href="doi.org/10.5281/zenodo.22815427">10.5281/zenodo.22815427</a>.</p>