How to build a Twitch creator shortlist with your AI agent

For this illustrative example, imagine a PC accessory campaign looking for English-language creators who stream racing games and average 10–100 concurrent viewers. These are example criteria, not a recommendation or a claim about actual campaign results.

Connect CreateWith to your AI app, then ask:

Find Twitch creators streaming racing games in English, averaging 10–100 viewers over an available recent window. Show the window, observation freshness, stream count and any missing data. Do not infer audience location from stream language.

Have the agent use the same metric window for each candidate. Compare average live viewers, streams observed and growth, rather than mixing a peak from one stream with another creator’s multi-day average.

Ask the agent to distinguish a small sample from sustained activity and explain why each candidate meets the requested filters. If the results are sparse, relax one condition explicitly instead of silently changing the brief.

For promising candidates, ask for available published research and source links. A dossier may include interests, projects or past brand relationships; many tracked creators do not have a dossier.

For these candidates, separate sourced facts from possible campaign angles. Cite evidence for any previous brand work, show its date where available, and label missing research as unknown.

Ask for a shortlist with the following fields. This is a review template, not a set of fabricated creator results.

• Creator and public channel link.
• Observed performance, comparison window and freshness.
• Evidence matching the brief, with source links.
• Potential campaign angle, clearly marked as a proposal.
• Unknowns to verify: rates, availability, audience demographics and interest in the product.

Review recent streams and confirm terms with the creator. A growth ranking is not a price estimate, a brand-safety guarantee or proof of authentic engagement.

One question can trigger several data calls, including searches, result pages and research unlocks. The pricing page explains the shared API/MCP allowance. Read the methodology before presenting findings as campaign evidence.