Room A
Public sample. The room and its accounts are anonymised; dates are reduced to the month.
This report sets out measurements of the chat audience in Room A over 14 days to 2026-11. It reports what was measured, how, from what data, and what the measurements cannot show. It does not label the streamer or any account. Conclusions are the reader's.
Reference room: A Kick room whose audience is independently known to be real (paying subscribers known to the operator). It is the comparison for every figure.
Findings
1. People who type in the room
| Measure | This room | Reference |
|---|---|---|
| Chatters per 15 min of airtime | 62.3 | 4.8 |
| Human messages per live hour | 489 | 123 |
| Distinct accounts that chatted | 601 | 514 |
| Live hours in window | ≈26 | ≈130 |
How to read it. The audience that can be counted: distinct accounts writing in chat, per 15 minutes of airtime and per live hour, with the channel's service bots, the streamer's own messages and known viewbot-seller accounts removed. Compare rooms on these, not on displayed viewers.
What it can't show. Viewers who watch without typing are not counted here, in either room.
2. Proof of watching
| Measure | This room | Reference |
|---|---|---|
| Heavy emote users in the room | 41 | 257 |
| …that never used a collectible | 38 | 32 |
| …as a share (%) | 92.7 | 12.5 |
How to read it. Kick awards collectible emotes only after 60 minutes of watching in a day, followed by a claim prompt that blocks the page until answered. Among accounts in the room that have posted 100+ emotes anywhere we monitor, this is how many have never once used a collectible. Heavy emote users in a watching audience accumulate them almost without exception.
What it can't show. We see emotes used in monitored rooms, not an account's inventory. A zero is evidence about a population, not proof about any one account: we have measured a long-standing paying subscriber with none.
3. What the chat is made of
| Measure | This room | Reference |
|---|---|---|
| Emote-only messages (%) | 0.4 | 16.2 |
| Collectibles among emote uses (%) | 2.6 | 36.8 |
How to read it. Share of human messages that are emotes only, and the share of emote uses that are collectibles. Fast real rooms on Kick lean heavily on emotes; text written by software tends not to use the emote picker at all.
What it can't show. Texture differs by language and room culture (Japanese rooms type short fragments instead of emotes). Read it against a room of the same language.
4. Messages written by the channel's own services
| Measure | This room | Reference |
|---|---|---|
| Messages from service accounts (%) | 29.2 | 53.3 |
How to read it. Share of all chat messages posted by service accounts (channel bots, announcement bots, game bots). These are removed from every other figure, but they inflate any raw message count quoted for the room.
What it can't show. Only service accounts we have identified are counted. A new service bot can sit in the human total until added.
5. Accounts new to everything we monitor
| Measure | This room | Reference |
|---|---|---|
| Accounts active in the room | 616 | 583 |
| …first seen anywhere in the window | 18 | 8 |
| …as a share (%) | 2.9 | 1.4 |
How to read it. Accounts active in the room during the window whose first message anywhere in ~1,300 monitored Kick rooms falls inside the window. Accounts merely new to this room are normal (raids, clips) and are not counted.
What it can't show. Bounded by our own history (from 2026-09-01 JST) and our room list, not by Kick registration dates. This room's chat has been recorded since 2026-10 JST.
6. The platform's displayed count against the measurable audience
| Measure | This room | Reference |
|---|---|---|
| The platform displayed (median viewers) | ≈680 | ≈37 |
| Displayed count, day-to-day change (%) | 0.4 | 10.8 |
| Chatters, day-to-day change (%) | 9.6 | 19.4 |
How to read it. Median viewer count the platform displayed while live, and how much it moved from one day to the next compared with how much the number of chatters moved. A displayed count carried by people tends to move with them.
What it can't show. The displayed count is the platform's figure and cannot be audited from outside. It is shown only to compare it with what can be measured, never as a size.
7. Accounts that post on a timer
| Measure | This room | Reference |
|---|---|---|
| Accounts posting on a steady interval | 6 | 0 |
| …also single-room, no replies, arrived together | 0 | 0 |
How to read it. Accounts whose messages arrive at a regular interval (median gap 90 s–20 min, low variation, no bursts), and of those, how many also post only in this room, never reply to anyone and arrived on the same day as three or more others like them. Reported whether or not it finds anything.
What it can't show. A pattern test. It does not catch automation that varies its timing, and a steady human can match the first condition.
8. Poll votes against the displayed count
| Measure | This room | Reference |
|---|---|---|
| Polls observed | 0 | 1 |
| Highest votes as % of displayed viewers | – | 22 |
How to read it. Polls observed in the room, and the highest number of votes any poll took as a share of the viewers the platform displayed at the time. Over 100% means more votes than the platform said were watching.
What it can't show. Only polls we observed while they ran. No polls means no reading, not a clean one.
9. Subscribers
No current subscriber roster was available for this room at the time of the audit.
How to read it. The current direct-subscriber roster (gifted subscriptions excluded): its size, how many of them chatted in the room during the window, median tenure, and the share who are themselves streamers.
What it can't show. Roster from the latest scrape only; subscribers who lapsed are not counted. Silent subscribers are normal: people pay without typing.
10. Influence: chatters who pay
No current subscriber roster was available for this room at the time of the audit.
How to read it. Influence is an audience that acts when the streamer asks. The first measure is conversion: current direct subscribers per 100 active chatters (accounts with more than 10 messages in the window). A room whose regulars back the streamer with money converts at a higher rate.
What it can't show. Covers only what we can see on the platform. Raid follow-through and response to calls to action are not in this report yet, and sales through a sponsor's own links or codes need the sponsor's data.
Method
Every figure comes from the monitoring system's analysis API, which reads a continuous log of chat, moderation, poll and subscription events recorded live from ~1,300 Kick rooms. Each finding names its source below; numbers.json lists every figure with its endpoint, JSON path and pull time.
| Source | Pulled |
|---|---|
| /analysis/channel/Room A | 2026-11 |
| /analysis/streamer/Room A?days=14 | 2026-11 |
| /analysis/collectibles/Room A?days=14 | 2026-11 |
| /analysis/composition/Room A?hours=336 | 2026-11 |
| /analysis/arrivals/Room A?days=14&min_msgs=3 | 2026-11 |
| /analysis/clockwork/Room A?days=14 | 2026-11 |
| /analysis/polls/Room A?days=14 | 2026-11 |
Tests we do not use
Each of these was tried against the reference room and against confirmed automation, and failed to separate them. None contributes to this report.
- Profile picture present or absent: manufactured accounts carry custom avatars more often than real paying subscribers
- Follower counts: cheap to buy, and spam accounts follow in order to post; they never appear as evidence
- Empty bio, no socials, no banner: most real subscribers leave these empty too
- Number of rooms an account chats in: real paying subscribers range from 5 to 50+ rooms
- “Never sleeps” / active every hour: follows the broadcaster's schedule, not the account's nature
- Posts only while the channel is live: humans do this too
- Variety of laugh words, rate of “W”: real reference accounts overlap suspect ones completely
- Shape of a poll's vote curve: tracks poll length and device mix, not authenticity
- Mostly-emote output: most such accounts hold collectibles; it is how people chat in fast rooms
- On-topic, contextual comments: software can read the stream as well as a person can
Provenance
The raw response from every endpoint was saved unchanged before this report was rendered, and the report was rendered from those files only. The extract hash is the SHA-256 of the SHA256SUMS manifest listing every raw file's hash. Recomputing it from the retained files shows whether anything has changed since issue.
Limits
- Chat history starts 2026-09-01 JST. Nothing before it is visible; “first seen” means first seen by us.
- Only monitored rooms (~1,300) are seen. An account's activity elsewhere on Kick, or on other platforms, is not.
- Viewers who never type are invisible to every chat measure, in this room and the reference room alike.
- The platform's displayed viewer count is quoted only as the platform's figure. Follower counts are not used.
- Data comes from Kick's public chat feed and public pages, which are undocumented and may change.
- All tests describe populations. None is proof about an individual account.