The agent successfully understood CircleCI's purpose, positioning, pricing model, and competitive differentiation by combining a single machine-readable fetch (llms.txt) with web search results and prior knowledge. The site's main pages returned only generic HTML scaffolding; the agent had to rely on CircleCI's LLMs-optimized text artifact and external third-party sources to construct a coherent answer. CircleCI publishes its key messaging in a fetchable, structured form (llms.txt), which significantly improved agent-readiness, but core details like exact credit rates and competitor comparisons required supplementary research.

• ›Step [15] (circleci.com/llms.txt) was the only direct, content-bearing fetch from CircleCI's domain that returned machine-readable, structured information. It contained the core definition, platform capabilities, and target personas—all of which appear in the final response's explanation of what CircleCI does and who it's for.
• ›Steps [1]–[7] fetched the main marketing pages (/pricing, /why-circleci, /features, /continuous-integration, /docs) but returned only HTML boilerplate with truncated payloads; the agent could not extract text content from these fetches, forcing reliance on web search and prior knowledge instead.
• ›Steps [9]–[13], [16], [18]–[19] were web searches that supplied competitive comparisons (vs. GitHub Actions, Jenkins, GitLab CI), pricing clarifications (credit rates, plan tiers), and feature differentiation (test splitting, Docker layer caching). These external sources filled gaps the site's own pages did not expose in a machine-readable form.
• ›The agent explicitly cited [15] (llms.txt) in its sources, confirming it was recognized as a first-party artifact. It also cited external third-party pricing and comparison guides, indicating the site does not publish comprehensive competitive positioning or detailed pricing models internally in a format the agent could fetch.
• ›The site's agent-readiness is mixed: it publishes an llms.txt summary (high readiness for identity), but its main pages are JavaScript-heavy marketing sites with truncated HTML responses, forcing agents to guess URLs or resort to web search. Pricing details and credit rates are present but scattered across docs and require manual navigation.