What changed on this site, newest first. Every entry names the page or file that changed and the reason; no entry is a projection of future work.
2026-11-02
Trust and methodology round.
What changed: Added /methodology/, /changelog/, /corrections/, /contact/ and /research/ai-visibility-variance-2026-11/; rewrote /about/; added a dated verification line and an author byline to content pages; the homepage now states what is retained and for how long (180-day contact records, 120-day event counters).
Why: The sampling method and the first 45-answer pilot were described in sales copy but not published as data. This round publishes the method, the CSV and a correction path, and names the person who does the work.
2026-10-31
Published /learn/chatgpt-citations/.
What changed: A 1,373-word guide to what makes a page citable in ChatGPT answers, with a six-step pass, a covers/does-not-cover pair and a six-question FAQ backed by FAQPage, Article and HowTo JSON-LD.
Why: The question was already in the buying conversation; the page answers it with a runnable check and a stated limit rather than a citation promise.
2026-10-31
Added the GEO page pipeline.
What changed: content/geo/chatgpt-citations.json holds the page data; scripts/build-geo-page.py renders it through templates/page.html and registers the URL once in sitemap.xml and llms.txt. scripts/geo_blocks.py and assets/geo.css carry the block renderer and styles.
Why: So a new guide can be added as data, with the shared header, footer and JSON-LD, without copying the shell into a second file.
2026-10-31
Added internal links to the citable-pages guide.
What changed: The shared footer gained a “Citable pages in ChatGPT” link, and the homepage links to the guide; the guide links back to the free check, the diagnostic and the method note.
Why: A page that nothing links to is a page that is hard to find, for readers and for crawlers.
2026-10-30
Published /single-vs-sequence/.
What changed: A comparison page: what one check can and cannot show, the stability rubric (unanimous absent, unanimous present, swing, flip), the day 0 → day 60 calendar, and a four-question FAQ with Article and FAQPage JSON-LD.
Why: Buyers were asking why one screenshot is not a measurement; the page answers the objection before the pricing conversation.
2026-10-30
Expanded /method/.
What changed: Added the stability labels and their limits, the “a single check is a screenshot” section, the Perplexity Sonar / logged-out ChatGPT disclosure, the 0–100 score formula from api/check.js, a synthetic-sample statement, two extra “will not claim” items, and a three-question FAQ.
Why: The method is part of the offer. Publishing it before the sale lets a skeptical buyer audit the work instead of taking a dashboard on trust.
2026-10-30
Added local static verification.
What changed: scripts/check-static.mjs and npm run verify check titles, descriptions, one H1, canonical URLs, JSON-LD parsing, internal links and sitemap coverage with no network access.
Why: Several of those failures are invisible in a browser and only show up after indexing. The check now runs before a deploy.
2026-10-30
Made the delivery promise explicit.
What changed: The homepage, /diagnostic/, /pricing/ and /sample/ now state 30 dated answers and five business days; the sample page carries the same service-sheet link as the pricing page.
Why: The deliverable and the date are the two things a buyer needs before paying; leaving them implicit made the offer look vaguer than it is.
Last updated: 2026-11-02
Read who runs the site, or see the AI visibility diagnostic this work supports.