Anime Episode Counts: A 3,539-Entry Catalogue Study

Attack on Titan appears in this study six times. Its TV entries run 25, 12, 12, 10, 16 and 12 episodes, and each counts as one row, carrying the same weight as all 366 episodes of Bleach. Keep that in mind when you read the headline figure: in Fantera’s catalogue snapshot of 14 September 2026, the median finished TV entry has 13 episodes. That describes how anime is released, one or two quarter-year blocks at a time. For how long a show you pick will actually take, including stories that span several entries, the practical guide to seasons and cours has the answer.

The shape of the distribution is strong evidence that the one-cour block is the standard unit among the popular TV anime this catalogue holds, and weak evidence about how long any story lasts. Both halves of that sentence matter.

The distribution

Percentages are rounded independently, so they need not sum to exactly 100. The median is the middle value after sorting all 3,539 eligible counts, and every entry has equal weight: a franchise hit and an obscure one-season show count the same.

How the sample was built

The snapshot holds 4,998 entries from Fantera’s AniList sync, which collects the most popular TV series and films on AniList. Short-form series, OVAs and online-only releases sit outside the catalogue from the start. From the snapshot we kept entries whose format is TV, whose status is FINISHED and whose episode count is a known positive number. That leaves 3,539 eligible entries and sets aside 1,459, most of them films, with the rest TV entries still airing, announced, cancelled or missing a count.

We kept finished broadcasts only because an airing show’s count is still moving. That filter says nothing about whether the story finished.

The bands are descriptive. The 7–13 band is roughly one cour, 14–26 roughly two, 27–52 a year or so of weekly broadcast, and 53 or more the long-runners. Nobody should read the boundary at 13 as the point where a show changes character.

What the shape shows

Two findings survive every caveat. First, the one-cour block is the default unit: 65.0 percent of eligible entries fall in the 7–13 band, and the two cour-sized bands together hold the large majority. Pick a finished TV entry from this catalogue at random and expect a quarter-year of weekly episodes, about five hours of viewing.

Second, very long entries are rare as entries. Only 134 eligible rows, 3.8 percent, have 53 episodes or more. That understates how much viewing time they represent, because a single one of those rows can be Naruto Shippuden at 500 episodes, more than the entire 1–6 band put together.

The 1–6 band is nearly empty, at 36 entries, for a structural reason. The sync never collects short-form series, which AniList files under a format of their own, so the band holds only the rare TV entry with a handful of episodes, such as Kaguya-sama: Love is War’s four-episode The First Kiss That Never Ends.

One story, six rows

Attack on Titan is the clearest illustration of what the study cannot do. Its six TV entries in the snapshot are Attack on Titan with 25 episodes, Season 2 with 12, Season 3 with 12, Season 3 Part 2 with 10, Final Season with 16 and Final Season Part 2 with 12. Four of those rows land in the 7–13 band and two in 14–26. The story runs 87 episodes before its closing specials, and the chart records it as four short entries and two medium ones.

The reverse case exists too. Natsume’s Book of Friends has seven seasons of 11 to 13 episodes each, all seven in the biggest band. The show is largely episodic, a boy who can see spirits helping one of them at a time, so each season is a reasonable place to stop, and here the entry really is a fair measure of your commitment.

The data has no way to tell those two shapes apart. Both look like a cluster of one-cour rows, and only knowing the shows reveals that one is a single escalating war story and the other is a shelf of gentle, separate evenings.

What it cannot tell you

The catalogue is a popularity-ordered sync, a sample of what AniList users track most rather than a census of everything broadcast in Japan. Popular shows get sequels, and sequels add rows, so franchises with many seasons may be over-represented compared with a random draw of titles.

We did not collapse franchises into stories or check whether each row’s story ends. The snapshot records no runtimes, and films are excluded rather than scored as one-episode shows. Nothing here estimates how long a real viewer takes once breaks and abandoned seasons are counted.

Used for what it is, the result still earns its place. It tells Fantera’s Finder what a short entry looks like, and it tells you that a filter for short entries returns mostly single cours, some of which are chapters of something much longer. The recommendation method explains what the length filter reads.

Reproduce the figures

Download the snapshot CSV. Each row holds an AniList ID, format, episode count and release status, which is enough to look up any entry without redistributing descriptions or artwork.

From the repository, run node --experimental-strip-types scripts/analyze-episode-counts.ts --check. It confirms the CSV matches its recorded SHA-256 checksum, recomputes the sample, median and bands from it, and fails if the result differs from the saved report. It makes no network request and changes no database.

These figures belong to the 14 September 2026 snapshot. When the catalogue changes, the right move is a new dated study published alongside this one, so the evidence behind these numbers stays checkable. For picks already checked for a real stopping point, use the weekend guide or the completed-story guide.