What people are prompting in PostHog AI

An analysis of 3.67 million messages sent by 230,000 users to PostHog AI between October 2025 and October 2026 reveals how different job roles use analytics assistance, when they prompt, and how their language changes with familiarity. Tuesday at 11 AM local time was the busiest period overall, while founders prompted more often at night and on weekends, asked heavily about retention and churn, and swore more than other roles; marketers focused on ad spend, traffic, and UTMs, engineers on debugging and feature flags, data professionals on databases and SQL, sales teams on customer research and session replay, and product managers on adoption and surveys. Common topics included dashboards, session replay, funnels, web traffic, experiments, feature flags, and error tracking, while users also sought career, reporting, investor, and pricing advice. The study found that users became more likely to say “please” and swear as their prompt counts rose, that questions made up one-quarter of messages, and that many users accepted AI suggestions with short replies such as “yes” or “go ahead.” The methodology used PostHog event data, SQL, Python, and reproducible notebooks, excluded employee messages, repetitive “hi” tests, and identifiable information, and acknowledged limitations including message-level rather than conversation-level counting, optional job-role data, keyword inaccuracies, and gaps in location data.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.