Discovery

How the Twitch directory sorts channels, and where discovery actually happens

"Just stream and they will come" assumes someone can find you. On Twitch, findability is a sort order — and the sort order is concurrent viewers, descending.

Open any Twitch category and the live channels are listed with the biggest at the top. That single fact governs discovery on the platform more than anything else. Twitch layers recommendations, tags and the front page on top of it, but the category page is where most browsing happens, and the category page is a viewer-count leaderboard.

The category page

Rows of channel cards, sorted by current viewers, highest first. Most browsers look at the first row or two and click. A minority scroll a page. Almost no one reaches page three. So a channel's viewer count decides not just how it looks but whether it exists to a browsing viewer — a channel at position 300 in a category of 400 is not "far down the list"; it is not in the list anyone reads.

What tags do

Tags filter, they do not rank. A viewer who filters a category by "English" or "First Playthrough" sees a shorter list — still sorted by viewers. Tags narrow the competition; they do not lift a small channel past a larger one. They matter most in big categories, where a good filter turns a 3,000-channel page into a 200-channel one.

Category choice is the one discovery lever fully under your control. Twelve viewers is page four in a large game and the top row in a small one. The rank estimator shows the placement for any viewer count in categories of different sizes.

The recommendation layer

The front page, "Live channels we think you'll like", and the sidebar are personalised — driven by what a viewer has watched, followed and lingered on. These surfaces do send meaningful traffic, but they are trained on existing behaviour: a viewer is recommended channels similar to ones they already watch, which means channels that already have viewers. Recommendations amplify discovery; they do not start it.

Where that leaves a new channel

At the bottom of the sort, invisible to browsers, and absent from recommendations because nobody has watched it yet. The loop has no entry point. Every route in works by changing the sort position: streaming in a small category, being raided into a viewer count, a clip going viral off-platform — or holding a viewer count for a session so the channel sits where browsers can see it. That last one is the mechanism a live-viewer session provides, and it works because the directory does not ask where the viewers came from. It sorts.

Kick

Identical model, smaller categories. The same twelve viewers sit higher on Kick than on Twitch because the tail is shorter. The Kick versus Twitch discoverability guide compares the two directories directly.

Related reading

Fill the room for your next stream

Real concurrent viewers, live in minutes, held for the whole session. First 25 free.

See packages →

Twitch's own description of how tags and categories work is on the Twitch guide to tags, and it says the same thing in fewer words: tags describe a stream so viewers can filter, they do not reorder the list. That leaves viewer count as the number that decides position, which is the only reason a held-viewer session is worth anything — a session from £1.89, with the first 25 viewers free, moves a channel up the sort for the hours you are live. It does not touch the thumbnail, the title or what happens after the click, and those still decide whether a browser stays.

Frequently asked questions

Does Twitch ever randomise the directory order?

No. It sorts by viewers, descending. Some experiments have shown a small number of low-viewer channels in "up and coming" rows, but the main list is a leaderboard.

Do tags help small channels get found?

They narrow the list a viewer sees, which helps in large categories. They do not change the sort order.

How does the front page choose channels?

Personalised recommendation based on watch history. It favours channels similar to ones the viewer already watches.

Is Kick's directory the same?

Same model — sorted by viewers — with smaller categories, so a given viewer count sits higher.