Similar Anime Recommendations

← Back to Game

To run the game, Anime Semantle precomputes how closely 800 anime resemble each other by pushing their synopses through a sentence embedding model. The point is measuring the distance between the answer and your guess — but the same data answers "I loved this, what should I watch next" just as well.

It isn't a score ranking and it isn't a tag filter. It only looks at how close two stories feel.

1. What each similarity band means

The band table the game uses, translated into recommendation terms.

Similarity	What it means for picking a show
0.85 and up	Same franchise or effectively a direct sequel. You've probably seen it
0.75 – 0.85	The useful band. A genuinely different show your taste is likely to cover
0.62 – 0.75	Same feel, but tone may diverge. Good when you want something new
Below 0.5	A different family. Not recommended unless you're broadening on purpose

These differ slightly from the in-game reading of the same bands, because the purpose differs. For narrowing down an answer, see the main strategy guide.

2. Cross-franchise pairs, measured

High-similarity pairs pulled from the data, with every same-franchise entry, sequel and movie excluded. "You liked Mushoku Tensei, try Mushoku Tensei season 2" isn't a recommendation. These are pairs of genuinely separate works.

Sim.	If you enjoyed	Try
0.865	Nichijou	Danshi Koukousei no Nichijou
0.863	Yuu☆Yuu☆Hakusho	Kimetsu no Yaiba: Yuukaku-hen
0.854	Seihantai na Kimi to Boku	Boku no Kokoro no Yabai Yatsu 2nd Season
0.836	Kaoru Hana wa Rin to Saku	Kaguya-sama wa Kokurasetai: Ultra Romantic
0.829	Sakamichi no Apollon	Shigatsu wa Kimi no Uso
0.827	Kono Sekai no Katasumi ni	Koe no Katachi
0.827	Seishun Buta Yarou wa Yumemiru Shoujo no Yume wo Minai	Kimi no Suizou wo Tabetai
0.827	Hikaru ga Shinda Natsu	Higurashi no Naku Koro ni Kai

Sakamichi no Apollon → Shigatsu wa Kimi no Uso (0.829) shows the method working. Both are coming-of-age dramas built around music, but by tag they're just Drama/Romance — lumped in with hundreds of others. Similarity isolates a far narrower axis: "students growing up through playing an instrument."

Nichijou → Danshi Koukousei no Nichijou (0.865) is not about the shared word in the titles. They cluster because both are plotless gag omnibuses. By tag, each is simply "Comedy."

3. Finding them yourself

The site has no lookup for browsing an arbitrary title's neighbors. There are two ways around that.

Option 1 — the past answers archive

The past answers archive expands the 12 most similar anime and characters, with scores, for every dated answer. Find a day when a show you love was the answer and you get its neighbor list directly. One more entry lands every day.

Option 2 — guess with it in-game

Entering a favourite title during play returns its distance from today's answer as a number. Enter several and compare: independently of what today's answer is, you start to sense where the titles you entered sit relative to each other.

Rule of thumb — aim at 0.75–0.85. Anything above 0.9 is usually the same franchise and already watched, and the 0.6s often match only at the genre level, which raises your odds of being disappointed.

4. What this can't do

Better to state the limits plainly.

It ignores ratings and popularity. Similarity only. A close but poorly reviewed title still surfaces
It can't read art, direction or music. The input is synopsis text, so "shows with a similar art style" is out of reach
It doesn't care when something aired. A 2000s title sits next to a current-season one
Briefly described shows are penalised. Chinese animation in particular scores systematically low because the source descriptions are thin. Details in When Similarity Betrays the Tags

Which model computes the similarity, and the weighting behind it (70% structure + 30% synopsis), is published with the supporting measurements on the about page.