Structure a back catalogue into seasons

YouTube will let you turn a playlist into a show: seasons, episode numbers, artwork. Shows may be eligible to appear in search, under Recommended shows, and in Continue watching. That is the structural change. It is also an easy way to hide four years of unrelated uploads behind a poster.

A show, in YouTube's own wording, is "a collection of high-quality episodes with a clear connection across episodes." The connection has to match how a viewer arrives. People who land from search on a standalone how-to do not want Episode 14 of 37. People who finish a story episode do. Dumping the catalogue into Season 1, ordered by publish date, optimises for your archive, not for their session.

This is a grouping method, then a conversion pass, then one metric so you can tell whether the grouping was right.

Group by arrival, not by the year you filmed it

Before Studio, write down how viewers actually reach the videos you want to bind. YouTube Analytics → Content → a representative video → Traffic source. You are looking for the dominant door, not the romantic one.

Dominant door	What the viewer is trying to do	Grouping that fits	Serial or non-serial
Search / suggested as a standalone topic	Solve one problem	Cluster by question, not by series name	Non-serial (any order, newest first)
Suggested / Up next from a sibling video	Continue	Cluster by the chain they already watch	Serial (oldest first, numbered)
Channel page / browse	Sample the channel	Cluster by format they can recognise in a thumbnail	Non-serial, unless the format is a story
External / notifications	Watch the new thing	Do not force old videos into this show	Leave the new flagship out, or give it its own season later

YouTube's show tools match that split. Non-serial is the default when you convert a playlist: any order, listed newest to oldest. Serial is for sequential watching and lists oldest to newest. Pick this on purpose. The default will turn a recipe archive into Episode 1 through 90 in filming order, which is the wrong story for a search viewer who wanted one cake.

Three filters decide whether two videos belong in the same show. All three have to pass.

Same promise family. A 12-minute workshop and a 45-second Short that came from it are not episodes of one show. The Short is a trailer or a clip, not Episode 0. Runtime can vary inside a family; the job cannot. "Watch me rebuild X" and "X explained" are two shows.
Same arrival door. If video A's traffic is 60 percent search and video B's is 60 percent suggested from A, they can be serial. If both are 60 percent search for different queries, they can be a non-serial topic show, or they should stay out. Mixing those two in a serial list trains Continue watching to serve the wrong next step.
Same look, on purpose. YouTube's best-practice list is blunt: consistent story or topic, same characters or hosts where you can, same intro, same thumbnail style, same format. Videos with no visuals are called out as unlikely to be treated as a show. If half the cluster is talking-head and half is screen-only slideshows, you do not have a look yet. Either commit, or split.

A practical clustering pass:

List recent long-form videos. Ignore Shorts.
Tag each with one promise family and one dominant traffic source.
Pile per family. Piles smaller than four stay playlists, not shows.
In a pile of eight or more, check "What they watch next." If viewers already go A→C→B, that is serial order, even if you filmed A→B→C. If they bounce to another family, it is a topic shelf, not a serial.

This is the opposite of putting everything in Season 1, and the opposite of a channel hub that uses playlists as year folders. Folders remain useful as archives. Shows are for piles a stranger would binge or sample as a unit. A repeatable series you are still filming belongs in its own show from episode one. Old format, new format: two shows, even if the host is the same.

Convert the pile without flattening the order

Once a pile passes the three filters:

Build a playlist that contains only those videos. Remove trailers, Shorts, and sentimental near-matches.
In Studio → Content → Playlists, hover the playlist, Options, Add show features. Converting a playlist selects non-serial automatically. If this pile is a story or a numbered build, change it to serial.
Check episode numbers. Conversion puts everything in Season 1. Manual playlist order wins; otherwise oldest-to-newest publish date. For serial shows, confirm against "watch next," not shoot memory.
Creating from scratch: Create → New show, then add content. You must set a season number. Use Season 1 for the current run. Park a reboot (new host or format) in Season 2.
Add the two artwork assets YouTube asks for, both 16:9, 1080p minimum, 4K encouraged, 10 MB max, JPG or PNG:
- Banner: poster with the show title or logo.
- Backdrop: the same art without the title or logo, used behind UI on browse surfaces.
Put the show on the channel Home tab as a single-playlist shelf (or a multi-playlist shelf if you now have several shows). A show that only exists in the Shows tab is a filing cabinet.

Do not stamp one thumbnail template across a non-serial topic show if it hides the search promise. Consistency is host, palette, type, and frame; the title can still name the question. An image model or thumbnail generator is enough for the 16:9 poster pair. Do not let artwork replace the grouping step.

Season cuts that hold up:

A season is a run, not a year. "Season 1 = 2024" is a folder. "Season 1 = the original workshop series, 11 episodes" is a season.
Serial seasons need an end. A run of roughly a dozen episodes is a unit most viewers will treat as a season. Non-serial topic shows can be longer because order does not matter; they still need a visible end or the shelf looks like a dump.
Leave orphans out. A one-off that happens to share a tag is not Episode 12. Orphans keep doing their search job better as standalone videos.

If you also run short-form series, that is a different object. YouTube Shows are built around the long-form playlist. Short repeating formats belong on their own cadence; the eight series formats and a scheduled series are that problem.

The playlist metric to watch after

Single-episode impressions will move for a dozen reasons. The show question is: did we make a bingeable (or sampleable) unit? Watch views per playlist start on the playlist you converted, plus whether Continue watching actually sends people to the next episode.

That is a playlist-context number, not a channel-wide audience card. YouTube names it views per playlist start (viewsPerPlaylistStart in the Analytics API): videos watched each time a playlist is started, counting only views in the playlist. In Studio: Content → Playlists → Analytics on that playlist. A jump there after conversion is the grouping working. A jump in generic impressions is not.

In practice, for two weeks after you convert:

Open a flagship episode in the show. Traffic sources: look at Playlist and Suggested. You want Playlist / Continue-watching share up relative to the same videos' last 28 days as a plain playlist, if you had one.
On the playlist's own Analytics, write down views per playlist start the day you convert, then again at day 14. That pair is the verdict.
For serial shows, click through Continue watching on a living-room session if you can. If episode 3 autoplays episode 9, your order is wrong no matter what the poster looks like.

How to read the pair:

Views per playlist start up, watch time per view flat. The grouping is working. People are taking a second episode. Do not recut episode 1 in a panic.
Views per playlist start flat, playlist impressions up. The shelf is being seen and not used. Order, serial/non-serial, or the "next" episode's packaging is wrong.
Views per playlist start down. You tied together videos that do not belong. The most common cause is serialising a search pile. Switch that show to non-serial or break it into two topic shows.

Give it two weeks, not two days. A conversion changes merchandising; wait through a weekend of living-room traffic. Filling future seasons is a cadence problem. A slot from an AI video generator only helps if the new episode would pass the same three filters.

FAQ

Should every video on the channel go into a show?

No. One-offs that win search should stay one-offs. Shows are for clusters with a clear connection and a consistent look. Forcing orphans in dilutes the connection YouTube asked for and trains Continue watching to serve a random next click.

Serial or non-serial if I am not sure?

Non-serial, which is also the default on playlist conversion. Switch to serial only when you can point at a watch-next chain or an actual story order. You can edit episode and season numbers later; starting serial on a topic archive is the mistake that is expensive to unwind because it lists oldest first.

How many episodes make a season?

Enough that Continue watching has somewhere to go, not so many that Season 1 is the whole channel. For serial, a run of about a dozen is a unit most viewers will treat as a season. For non-serial topic shows, cap it when the promise family starts to split (beginner vs advanced is two shows).

Do I need new artwork if the thumbnails already match?

Yes, because YouTube asks for two 16:9 assets the episode thumbnails do not cover: a banner with the show title or logo, and a backdrop without it. Matching episode thumbnails still matter for the consistent-look test. They do not substitute for the poster pair, especially on a TV-sized browse surface.