Each measurement in plain language, with the number, the comparison it should be read against, and what it can and can’t tell you.
Live-stream audience audits
Know if your influencer actually has influence.
Find out who is actually in the room before you pay for it. Viewer counts, chat activity and subscriber numbers can all be bought. We measure a streamer’s audience from the inside, every message and every account, and tell you which numbers hold up.
Audit readout
Room A · Kick · one stream, Nov 2026
- Viewers displayed by the platform
- 639
- Accounts that typed in chat
- 110
- …with no proof of ever watching an hourKick awards collectible emotes for 60 min of watch time
- 106
- Messages that were emote-onlyTypical for real rooms: 30–45%
- 0.6%
- Accounts first seen anywhere that day
- 7 of top 30
- Messages carrying the same machine formatting mark
- 2
Why this exists
Brand deals, clip campaigns and platform payouts are priced on audience numbers. Those numbers are easy to inflate: viewbots hold a count steady, scripted accounts fill chat with plausible lines, and paid subscriptions are traded between streamers.
Most checks look at one account at a time, and modern automation passes them. Ours looks at the whole room over time and at the things automation can’t cheaply fake.
What we measure
Each test was checked against an audience independently known to be real. Tests that looked convincing but didn’t separate real from fake were dropped, and we say which ones in every report.
-
Proof of watching
The platform rewards an hour of actual viewing with collectible emotes. A real audience accumulates them; accounts driven by software almost never do. We count them across every account that speaks.
-
Who is new, and when they arrived
Account registration batches, arrival times, and whether an account exists anywhere outside this one room. Rented audiences arrive together and leave together.
-
Chat that only looks like conversation
Lines drawn from a shared pool, captions describing the screen, a shared posting rhythm, formatting left behind by the software that wrote them. Invisible per account, clear across the room.
-
Numbers that contradict each other
Polls that take more votes than the viewers the platform says are watching. Viewer counts that stay flat while chat swings. These are the platform’s own figures disagreeing.
-
Subscribers who are an audience
The current subscriber roster: how many of them chat in the room, how long they’ve stayed, and how many are other streamers trading subscriptions.
-
Real chatters per hour, not views
Service bots, the streamer and known viewbot sellers removed. What remains is the people who show up and talk, measured against airtime.
-
Whether the audience acts
Influence is an audience that does something when the streamer asks. When they raid another channel, how many arrive and how many are still there days later. How many chatters come back stream after stream. How many chatters become paying subscribers. Whether chat answers a call to action, measured against what the streamer actually said on air.
The report
Every audit is written so that someone who disagrees with it can check it.
How each figure was produced, from which source, over which window. No black-box score.
When each data point was captured and by what. A fingerprint of the underlying data is kept, so the report can be shown later to be unaltered.
What we couldn’t see: periods we weren’t recording, viewers who never type, rooms we don’t monitor. Stated up front, every time.
What we don’t do
- No verdicts on people. We measure audiences and report the figures; we don’t label streamers.
- No follower counts. Followers are cheap to buy and prove nothing, so they never appear as evidence.
- No single magic score. One number hides the method. You get every measurement behind it.
Request an audit
Tell us the channel, the platform and what you’re deciding: a sponsorship, a clip campaign, a partnership. We’ll tell you what we can measure and how soon.
Currently covering Kick. Audits are delivered in English and Japanese.