One day of GEO work on Pancake, June 11, 2026: what we observed

This is a historical record of one focused day of Generative Engine Optimization work on Pancake on June 11, 2026. At that time, Pancake was presented as a platform for building, running, and hosting trading strategies. The product and its website have since changed. This article describes what changed that day, not Pancake's current positioning and not a measured lift in AI answers or citations.

Scope and starting point on June 11, 2026

The Pancake site then had an llms.txt file, segmented sitemaps, robots.txt rules, and pages intended to answer product questions. The work focused on two question families: how to host a trading strategy and how to evaluate a trading strategy.

We treated the work as an implementation exercise. The counts below are the team's change record: ten question pages, three entity-definition pages, three comparison pages, one comparison list, and machine-readable surfaces for selected artifacts.

What changed

• Question pages opened with a direct answer, then added context, scope, and limitations.
• Definition pages explained the product's use of the term strategy hosting.
• Comparison pages included a sincere section on when another tool may fit better.
• A list compared six hosting options and described the alternatives rather than treating them as strawmen.
• Selected artifacts exposed Markdown representations and a preferred citation URI where the implementation supported them.
• An IndexNow notification was sent after deployment.

These mechanisms have different jobs. The Robots Exclusion Protocol describes crawler access requests, while a sitemap helps search engines discover URLs but does not guarantee crawling or indexing. An IndexNow response confirms receipt of a notification, not that a URL was crawled, indexed, cited, or recommended.

What this experiment cannot show

This was a single-site, one-day implementation record. It had no randomized rollout, control group, stable pre-change and post-change answer sample, or independent causal analysis. Model responses, search results, crawler schedules, competitor publishing, and user behavior can all change at the same time.

We therefore do not attribute a later answer, citation, recommendation, visit, or conversion to the work described here. A page change can be documented. Its downstream effect must be measured separately under recorded conditions.

What to measure next

• Answer samples: exact prompt, provider, model, date, location, and browsing state.
• Citations: the URL or domain shown in each recorded answer.
• Search demand: queries, impressions, clicks, and index state.
• Human referrals and conversions: observed sessions, goals, and revenue under a stated attribution rule.
• Crawler evidence: identified agents, requested paths, source coverage, and HTTP outcomes.

Keep those records separate. This case study is useful as a reproducible historical change log and a list of hypotheses to test. It is not evidence of downstream answer selection or recommendation.