Players have changed what they are grading. They are no longer only asking whether a game is translated. They are asking whether the publisher respects their language, their culture, and their community. When a localisation feels rushed, generic, or machine-generated without real human craft, communities now have a name for it: "AI slop." And once that label sticks, it moves fast, into reviews, into sentiment, into sales.
For a publisher, that sounds like a risk. It is actually an opening. The teams that treat multilingual content as a system, rather than a word-count bought at the end of production, get to turn a reputational landmine into a growth lever. That is the real argument for content operations: not cheaper translation, but games that feel authored in every market they ship to.
"AI slop" has become shorthand in gaming communities for localisation that puts speed and cost ahead of quality. You can see it in fan-translation and retro-game threads, where players call out AI-assisted work for weak nuance, thin editing, and a lack of respect for the people who used to do the craft.
It is not confined to niche corners either. In the Kingdom Come: Deliverance II AMA controversy, concern about localisation quality and the suspicion that AI had replaced translators grew into a broader trust problem, even though the studio said publicly that no AI-generated content shipped in the final game. The lesson is uncomfortable but useful: players react to perceived intent as much as to output. If it looks like you cut corners for non-English audiences, the backlash rarely stays in one language.
The commercial case has not softened. CSA Research's Can't Read, Won't Buy study, covering 8,709 consumers across 29 countries, found that 76% prefer to buy products with information in their own language, and 40% will never buy from a website in another language. In games, that preference reaches well past the store page. Language shapes discovery, wishlist conversion, community sentiment, support load, retention, and live-service spend.
The numbers can be dramatic. One Steam-focused analysis credited Hooded Horse with wishlist lifts of roughly 35 to 40% from French and German localisation announcements, over 50% for Spanish, and over 90% for East Asian languages. Meanwhile the market itself is moving. As AI compresses per-word pricing, plain translation revenue growth is slowing, while cultural adaptation grows faster. That is the signal worth acting on. Translation is becoming a commodity. Cultural resonance is not.
It is rarely the AI on its own. It is almost always an operating model that treats localisation as a late, mechanical task. The failure points are consistent:
Tabletop Simulator is the cautionary tale. The team ran machine translation across 29 Steam languages and invited the community to correct it. The reaction was not only about accuracy. Players objected to being handed responsibility for quality that they felt the studio should have owned.
A strong multilingual operation does not ask how cheaply it can translate the strings. It asks how to make the experience feel authored for each market, at the speed and scale the business needs. That takes five connected capabilities:
None of this requires reinventing production. It requires moving a few decisions earlier and making them deliberate.
Localisation belongs in production during development, not after content lock. Following IGDA best practice, that means a central terminology base for characters, items, skills, factions, UI, and lore; a style guide covering tone, humour, profanity, formality, and speech patterns; a culturalisation brief per launch market; and a localisation kit with source files, string IDs, context screenshots, reference builds, and instructions for triggering hard-to-reach text.
Not every string deserves the same treatment. A tiered model keeps budgets defensible without defaulting to low quality:
AI earns its place as a production layer: first-pass drafts for repetitive content, consistency checks across languages and builds, detection of untranslated strings and UI overflow, and QA support for missing or malformed text. The final creative call stays with language leads who understand the game, the genre, and the community. Hybrid workflows cut cost and turnaround, but narrative, wordplay, and context-dependent dialogue still need people.
Community feedback is valuable, and it should never be the primary quality mechanism. Run closed playtests with native speakers before release, open clear channels for terminology and cultural issues, respond quickly and in public when something is wrong, and credit the people who did the work. What players reject is not being asked for feedback. It is being asked to fix systemic failures for free.
Publishers do not need to pretend AI does not exist. They need to be credible about how they use it. A defensible policy says where AI is used, where humans lead, how quality is measured and who is accountable, how player feedback is handled after launch, and what the studio will not do, such as shipping unreviewed machine output for narrative or brand-critical content. The Alters showed the cost of staying quiet: players found undisclosed generative AI, and the conversation stopped being about translation errors and became about trust.
The way to defend this work is to stop counting words and start measuring outcomes: wishlist growth, conversion, reviews, and revenue by language; LQA defect rates and post-launch localisation complaints; time from content freeze to language-ready build; cost per active player per market rather than cost per word; sentiment in non-English communities; and the share of updates shipped day-and-date across supported languages. The evidence that this matters is already visible. Cross-language review research on Steam has found games heavily review-bombed in one language while performing strongly in others, with localisation at the centre of the split.
The shift into global content operations is well timed, and the opportunity is not to sell "translation with AI." It is to sell accountable multilingual content operations: creative localisation, cultural intelligence, AI-enabled scale, in-context QA, and community-ready publishing, run as one system with someone answerable for the result. Translation has its place, and it is not always the answer. In games, the publishers who understand that are the ones whose communities feel seen, and whose numbers show it.
Regulation (EU) 2023/1230 applies from January 2027. The language obligation is explicit and instructions may now be digital, turning a large documentation estate into a versioning problem.
Under the EU Medical Device Regulation, translation is closer to a subscription than a launch task: the obligation regenerates with every product revision and every new market.
The generic advice says lead with language when you're small. In Africa, that advice can quietly cost you market share. How to decide, stage by stage.
