PostHog promises LLM observability in five minutes. Its docs page mostly delivers, but starts in the wrong place.

PostHog earns a 7.6 (B) on this audit. The LinkedIn ad cluster is unusually disciplined: nearly every variant repeats the same two-part promise of a five-minute setup and 100K free events per month, with playful one-liner body copy carrying the brand voice.

The destination is a documentation page rather than a marketing landing page. That tonal switch is fine for a developer-focused offer - clickers do expect an install path - but the page opens with the H1 'Getting started with AI Observability,' which describes the category without echoing the specific promise that earned the click. The five-minute setup hook and the 100K free events reassurance both exist on the page; they just live below the fold.

The destination is the AI Observability getting-started guide. The 'Capture LLM conversations' section sets the scope (track conversations, model performance, traces, cost-per-user, multi-conversation sessions) and links to install paths for 40+ providers including OpenAI, Anthropic, the Claude Agent SDK, Vercel AI SDK, LangChain, LangGraph, LiteLLM, AWS Bedrock, Azure OpenAI, and Ollama.

Three follow-on sections cover 'Track AI generations' (token counts, latency, cost calculation, trace IDs), 'Evaluate model usage' (the AI Observability dashboard), and 'Integrate customer data' (product analytics, error tracking, session replay). A 'Use for free' block at the bottom confirms 100K LLM events per month free with 30-day retention, no credit card, and usage-based pricing at $0.00006/event above the cap.