TikTok's algorithm update in late 2025 deprioritized bilingual content creators, cutting reach for Spanish-English code-switching videos by up to 60%. The platform optimized for language consistency — and in doing so, made the fastest-growing demographic in American media functionally invisible on the fastest-growing platform.
Camila Reyes had 1.2 million followers. She made cooking videos — not the kind where someone in a spotless kitchen assembles a Buddha bowl while explaining macros, but the kind where her abuela's hands appear in frame without warning, adding cumin to a pot of frijoles while Camila narrates in a mix of English and Spanish that is not a performance but a language. Her audience was the thirty million Americans who speak exactly the way she speaks — switching between languages mid-sentence because some things are said better in Spanish and some are said better in English and the choice is not conscious, it is cultural.
In September 2025, her engagement dropped. Not gradually, the way it does when an audience moves on. Suddenly. Her cooking video of pozole rojo — a video that would have gotten 400,000 views a month earlier — reached 38,000. Her reels where she spoke entirely in English continued performing normally. The ones where she code-switched, where she spoke the way she actually speaks, fell off a cliff.
She was not alone. Across the platform, bilingual creators — particularly those who mixed Spanish and English — reported the same pattern. Creators who had built audiences by reflecting the actual linguistic reality of millions of Americans found that the algorithm had decided their language was a problem. Not explicitly. Not through a policy announcement. Through a systems optimization that treated language consistency as a quality signal and language mixing as noise.
Camila's analytics told her what the algorithm wanted: pick a language. Create content in English or create content in Spanish. The algorithm would reward consistency and penalize mixing. But Camila's audience didn't speak in one language. They spoke in the space between two languages — and that space had just become invisible to the recommendation engine that determines who sees what on the platform where sixty-five percent of Gen Z Latinos discover content.
Camila didn't stop posting. She started making two versions of every video — one in English, one in Spanish. Her engagement recovered partially. Her creative voice did not. The thing that made her content work — the fluid, unselfconscious movement between languages that her audience recognized as their own — was the thing the algorithm had learned to suppress.
When a platform optimizes for language consistency and the fastest-growing demographic in America speaks in two languages simultaneously, the optimization is a policy. The platform just made thirty million people harder to reach.
TikTok's recommendation algorithm — the core technology that determines which videos appear on which users' For You Pages — processes multiple signals to assess content quality and match it with likely audiences. Among these signals are audio transcription, language detection, and content categorization. In late 2025, creators and analysts documented a significant change in how the algorithm handled multilingual content.
The change was not announced. TikTok, like all major platforms, does not publish detailed algorithm specifications. What became visible through creator analytics and third-party monitoring was a pattern: videos containing mixed-language audio — particularly Spanish-English code-switching — experienced significant reach reductions compared to single-language videos from the same creators. Estimates from creator networks and analytics platforms placed the reach reduction at 40–60% for bilingual content versus English-only content from the same accounts.
Language consistency as a quality signal makes sense for monolingual markets: if a user speaks English, serving them a video in English is more likely to produce engagement than serving one in a language they don't understand. The problem is that code-switching is not "inconsistent language." It is a distinct language practice spoken by tens of millions of Americans. The algorithm treats it as the same category as random language mixing — noise — when it is in fact a coherent linguistic behavior with its own grammar, vocabulary, and audience.
The algorithm doesn't suppress Spanish. It doesn't suppress English. It suppresses the space between them — which is exactly where Latino identity in America lives.
TikTok's business model depends on maximizing time-on-platform, which depends on the recommendation engine's ability to match content with audiences likely to engage. Language detection enables this matching: English content goes to English-speaking users, Spanish content goes to Spanish-speaking users. Bilingual content breaks the matching model. The algorithm cannot confidently assign a bilingual video to either language audience, so it assigns it to neither with the same confidence — which means lower distribution.
The platform incentive is not anti-Latino. It is anti-ambiguity. But when the largest bilingual population in the Western Hemisphere creates content that is ambiguous to a language-detection model, the technical constraint produces a cultural outcome. Sixty-five percent of Gen Z Latinos in the United States report discovering new content primarily through TikTok. When the platform's recommendation engine deprioritizes the linguistic register in which those users naturally communicate, it restructures the attention economy for an entire demographic.
Spanish-language media in the United States generated over $9 billion in advertising revenue in 2024. The Spanish-speaking internet audience in the US is the third-largest in the world. A platform behavior change that reduces the visibility of bilingual content is not a minor optimization — it is a restructuring of who has access to the attention of one of the most valuable and fastest-growing consumer demographics in the Americas.
Bilingual creators have adopted three primary responses. Some, like Camila, produce duplicate content — the same video in two language versions. This doubles production cost and eliminates the creative authenticity that built their audiences. Others have shifted to English-only content, effectively code-switching their creative practice to match the algorithm's preference. A third group has migrated to YouTube Shorts and Instagram Reels, where bilingual content has not experienced the same suppression pattern.
The migration is significant because it represents a platform behavior change driving demographic audience shifts between platforms. If TikTok's algorithm makes bilingual Latino creators less visible, and those creators migrate to competitors, the platform risks losing both the creators and their audiences — a feedback loop that other platforms are actively exploiting. YouTube launched a Spanish-language creator fund in late 2025. Instagram increased promotion of bilingual content in its Reels algorithm during the same period.
Code-switching is not a content strategy. It is a language. When a recommendation engine treats it as noise, the engine is making a cultural judgment dressed as a technical optimization.
TikTok's algorithm changes frequently, and all creators experience reach fluctuations. The counterevidence is the language-specific pattern: English-only content from the same creators maintained performance while bilingual content dropped. If the change were general, all content types from affected creators would show similar declines.
Content in a single language may reach a larger addressable audience than content mixing two languages, making lower distribution a rational optimization. This is true in aggregate but wrong for the specific audience: bilingual Americans are the audience for bilingual content, and that audience is 65 million people. The algorithm is optimizing for global reach at the expense of demographic fit.
TikTok has not confirmed any algorithm change specifically targeting bilingual content; all evidence is inferential from creator analytics and third-party monitoring. Whether the pattern persists, intensifies, or self-corrects as TikTok adjusts its language models is unknown. The competitive response from YouTube and Instagram is emerging but not yet quantified in market share terms. Whether this pattern extends to other bilingual populations globally (e.g., Hindi-English in India, Arabic-French in North Africa) has not been documented.
The platform didn't announce that bilingual content would be deprioritized. It didn't need to. The analytics told creators everything the policy memo didn't: pick a language or lose your audience.