Japanese Game Localization Pitfalls: A Checklist for Honorifics, Register, and Pronouns

Japanese players are often called some of the most demanding in the world — yet what sinks a Japanese localization is rarely vocabulary. It's honorifics pointed in the wrong direction, characters who don't sound like that character, and one misplaced「あなた」. This checklist unpacks the three densest minefields in Japanese game localization — honorifics, register, and pronouns — and ends with a pre-launch checklist you can copy as-is.

Honorifics Aren't "Politeness Points" — They're a Map of Relationships

Japanese keigo falls into three broad systems, each pointing in a different direction:

System	What it does	Where it shows up in games
Sonkeigo (尊敬語)	Elevates the other party's actions	NPC lines addressed to the player or a superior, official announcements
Kenjōgo (謙譲語)	Lowers the speaker's own actions	Butler and attendant characters, shopkeepers
Teineigo (丁寧語)	The polite です/ます register	System messages, UI, tutorials

The first minefield is direction: using sonkeigo about yourself or kenjōgo about the other party is the kind of rookie mistake Japanese players spot instantly. The subtler one is relationship change: if the dialogue stays stiffly formal after the protagonist goes from "stranger" to "comrade-in-arms," the intimacy arc breaks. Conversely, a devoted attendant shouldn't drop keigo even in a mid-battle shout — it's part of the characterization.

System text has one iron rule of its own: never mix the polite style (です・ます) with the plain style (だ・である). Pick one for UI, announcements, and tutorials, and stick to it — mixing them makes the whole interface feel like a patchwork from multiple hands.

Register and Role Language: How a Character Speaks Is the Character

Japanese has a rich tradition of「役割語」(role language): the elderly scholar's「〜じゃ」, the refined young lady's「〜ですわ」, the rough warrior's「〜だぜ」— one line is enough for players to picture the character. For localization this cuts both ways:

• Get it right, and the character carries their own persona before they even appear — text, voiceover, and art reinforce each other;
• Get it wrong or inconsistent — the same character is gruff in Chapter 1 and suddenly bookish in Chapter 5 — and players are yanked out of the story.

The practical move: before work begins, build a character language profile for every major character — first-person pronoun, sentence endings, keigo level, and how they address each other major character — and carry it through every translation batch. This is also exactly the kind of dimension a machine is good at watching: in our AI LQA scoring framework, "register and character voice" is a standalone scoring dimension that can be regression-checked line by line against the profile, instead of relying on a reviewer's memory.

Pronoun Minefields: Neither "I" nor "You" Has a Default

English I/you and Chinese 我/你 have no safe one-to-one mapping in Japanese.

First person: 私 (watashi), 僕 (boku), 俺 (ore), あたし, わし, 拙者… each carries its own age, gender, and personality tags. Render a hot-blooded teenager's「俺」as「私」and the character gets a personality transplant; let a dignified priestess call herself「俺」and you have an incident report.

Second person is the bigger trap:「あなた」is "you" in the textbook, but in real usage it often reads as distant, condescending — or carries the specific nuance of a wife addressing her husband. The idiomatic Japanese move is to avoid second-person pronouns altogether — use name + suffix, a title (隊長, 先生), or simply omit it. Every "you" in the source text deserves a pause: does this line actually need to say "you" in Japanese?

Name Suffixes and Variable Concatenation: {player}さん Is Not a Universal Answer

さん/くん/ちゃん/様/殿/先輩 — every suffix encodes a layer of relationship. Game text adds a special complication: variables.

• Hard-code {player}さん and you erase the differences that matter — the tsundere calling the player「{player}ちゃん」, the enemy faction using the bare name. The suffix should follow the speaker, not the template;
• Concatenation also collides with word order: Japanese pushes the verb to the end of the sentence, so an English template's word order jammed into Japanese often reads broken — the line needs rewriting as a whole, not word-by-word substitution;
• Keigo level bleeds into verbs too: the same "item obtained" line is「入手しました」in the system's voice and「手に入れたぜ」from a rough character. Reusing one string across multiple speakers is manufacturing a failure.

Variables, placeholders, and concatenation are "structural problems a machine can see" — run automated checks and clear them before review. Honorific direction and role-language consistency, on the other hand, are context problems, and they're among the most typical failure modes of generic AI translation — we broke these down in Where AI Breaks in Games: 7 Failure Modes. The common fix is feeding character and scene context to the engine instead of translating lines in isolation.

Pre-Launch Checklist (Copy As-Is)

• Every major character has a language profile: first-person pronoun, sentence endings, keigo level, forms of address for each counterpart
• Spot-check sonkeigo/kenjōgo direction: no reversed honorifics that elevate the self or lower the other party
• System text uses one style (polite or plain) consistently; UI and tutorial tone match
• Every source "you" reviewed individually: switch to a name, a title, or omission — use「あなた」sparingly
• Name suffixes around {player}-type variables vary by speaker, not hard-coded to さん
• Register shifts around relationship milestones (alliance, betrayal, confession) are deliberately designed
• A native Japanese reviewer does the final feel-check, sampling character dialogue rather than UI

Conclusion

The hard part of Japanese localization isn't "translating correctly" — it's sounding like that character in that relationship. Behind all three minefields — honorifics, register, pronouns — sits the same fix: build character language profiles first, then regression-check every translated line against them. Before your next batch of Japanese text goes out for translation, spend half a day locking down each major character's first-person pronoun and sentence endings — that bit of preparation saves rounds of rework later.