How CreateWith collects and interprets creator data
Twitch coverage, collection history, metric freshness and the limits of creator research in CreateWith.
By CreateWith · Published
What does CreateWith cover?
CreateWith is an agent-first creator-discovery service built on Twitch data. It collects point-in-time observations and derives performance history from those observations. A subset of creators also has published research about interests, projects and past brand work, with sources.
Collection began in June 2026. A tracked creator record does not mean continuous observation, a complete research dossier or full coverage of every Twitch category. Off-platform research is not equivalent to a YouTube, Instagram or TikTok analytics dataset.
Read the dated coverage report for measured totals and definitions.
Where does the history come from?
Twitch supplies current stream and follower information. CreateWith builds its own time series by collecting observations over time; it cannot backfill viewer or follower history from before collection. Missing observations remain gaps.
Raw stream observations are retained for about 30 days. Hourly summaries preserve collected history beyond that retention window. The estimated raw observation count on the statistics page is therefore different from the lifetime total in a dated coverage report. Do not add them together: their underlying observations overlap.
VOD and clip view counts are native cumulative counts and are different from live concurrent viewers. None of these is a count of unique audience members.
How fresh are the metrics?
Live data describes the most recent capture available. Historical metrics are computed on a separate schedule. Use the observation or metric publication timestamp, rather than assuming a new response means newly collected data.
- 3d window
- hourly
- 7d window
- every 6 hours
- 14d window
- daily
Longer windows need enough collected history and an available published rollup. A selectable or requested window is not a promise of complete data for every creator.
What can the rankings tell me?
Trending rankings compare viewer growth with category and follower-size cohorts, reward acceleration and discount one-stream spikes. They help identify candidates for further investigation.
Growth, follower counts and viewer-to-follower ratios do not establish authentic engagement, sponsorship prices, audience demographics, location or campaign effectiveness. A stream language is not evidence of the creator’s country or their audience’s location.
How should I use creator research?
Read the cited sources and their dates. Separate a documented brand relationship from an inference about suitability. Missing research means unknown, not that a creator lacks an interest or has never worked with a brand.
Use research to form questions, then review recent content and contact the creator before making a campaign decision. Our shortlisting guide shows that workflow.
How do agents access the data?
Agents can discover the service through llms.txt, then use the REST API or MCP connection. Product documentation and the dataset statistics endpoint are public. Creator-data access retains the service’s account, trial and usage limits.
Connection instructions and current plans explain access. Report data questions or corrections through Contact.