Baseline established — the day-zero snapshot

GeoSalience launched on 2026-05-31. This experiment records the zero point — the state of the site on day one — so that every later experiment has an honest "before" to measure against.

Key facts at baseline (2026-05-31):

• Live corpus: 2 articles, both cornerstone — What is GEO? and The llms.txt spec: adoption and setup — plus 6 glossary terms. Ten further articles exist as drafts, gated on primary research, and are hidden from listings.
• Crawler policy: robots.txt welcomes the major AI crawlers (GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and others). Cloudflare runs in DNS-only mode, so those crawlers reach the origin directly rather than being filtered at the edge.
• GEO surface live: llms.txt, llms-full.txt, .md aliases on every article, Article/Dataset/Breadcrumb JSON-LD, canonical URLs, sitemap, and RSS/Atom/JSON feeds — all verified returning 200 over HTTPS at launch.

What is measured at baseline, and what is not

This is the honest part. A baseline is only useful if it says exactly what it knows.

• AI-crawler hits — not yet measured. The nginx-log parser that turns crawler footprints into a daily dataset is scheduled but was not streaming on 2026-05-31. The day-zero record for crawler hits is therefore "pending first run," not "zero confirmed."
• Citation rate (North Star) — not yet measured. The 50-prompt, 4-LLM harness had not run a baseline at launch. The expected day-zero value is near 0%, but we record it as "pending," not as a measured 0%.
• Traffic — off by design. Plausible is env-gated off at launch.

We do not fabricate a number we did not collect. As each pipeline produces its first run, this page's changelog is updated with the real figure.

Why a near-zero baseline is valuable, not embarrassing

The entire point of a build-in-public GEO lab is the curve, not the starting height. A site that can show "citation rate went from a measured 0% on 2026-05-31 to X% on a later date, and here is the experiment that moved it" has primary research no competitor can copy. That story is only possible if the zero point is recorded honestly today.

How to reproduce this snapshot

Every claim above is checkable: fetch https://geosalience.com/robots.txt, https://geosalience.com/llms.txt, and any article's .md alias; count the articles visible on the home page; read the launch entry in the project changelog dated 2026-05-31. Nothing here depends on private data.

Limitations

This is a snapshot of corpus and policy, not of measured traffic. The crawler and citation pipelines were not yet streaming at launch, so their day-zero values are "pending first run." The corpus is small and single-domain. Those limits are the reason exp-001 exists: to make the starting line explicit before anything is claimed to have moved.