Every campaign speaks a language, and it should be one the audience understands. Audience Language shows the top languages of a creator's engaged audience with the share for each. It answers a question the creator's own content can't: a creator posting in English can easily have a third of their audience speaking something else. Like every audience data point, it only ships for creators where the sample behind it is large enough to be trustworthy.
Start From the Engaged Audience
For each creator, we analyze a sample of their engaged audience. These are the people a brand actually reaches, which makes them a better base for demographics than raw follower counts. Audience estimates are extrapolated from public engagement signals, and every audience statistic is an aggregate.
The Audience We Analyze
Every audience data point we publish is built on the people who actively interact with a creator's content, because they're the ones who see and respond to campaigns. The full reasoning behind this approach is on the Audience Analytics page.
Language Detection on Public Signals
The engaged audience goes through the same language detection that runs across the rest of our data, documented in full on the Language page. It works from public signals only, and where the signal isn't clear enough, that part of the audience is excluded rather than guessed.
Clear Signal Counts, Weak Signal Doesn't
Not all of an audience carries enough public signal for a language estimate. That part is simply left out of the calculation. The ranking is always computed over the share of the audience where the signal was clear, never padded with guesses.
Aggregate With Quality Gates
Languages are counted and ranked, and the top of the ranking ships with a share for each. The result is only published when the sample is large enough to be trustworthy; below that, the field stays empty rather than showing a shaky number.
From Many Voices to a Ranking
A ranking computed from a handful of people is noise, not data. We require the sample behind the field to cover at least 1,000 engagers with a clear language signal before publishing, and up to 10,000 for creators with larger audiences. Below that floor, Audience Language stays empty for that creator. An empty field is a quality signal, not a gap.
Illustrative Example
Two creators who both post in English, with audiences that don't overlap at all in what they speak.
Two Creators, Side by Side
An English-first audience with a real Spanish-speaking segment. English campaign copy reaches over half, and a Spanish variant could unlock another fifth.
No majority language. English copy reaches just over a third, so a single-language campaign leaves most of this audience out.
A Spanish-speaking audience isn't necessarily in Spain: it can span the US, Mexico, and half of Latin America. That's why language and location ship as separate audience data points, and the strongest targeting combines both.
Output Format
The exact fields this data point ships with, shown with example values, and how to use them in search.
The Actual Fields
The top 3 languages of the analyzed sample. Shares don't sum to 100% because the audience speaks more languages than three.
Filter creators by the share of their audience speaking a language, for example "audience English above 50%" for an English-only campaign, or find creators with a strong Spanish-speaking segment for a bilingual push. Combine it with audience location to separate Spain from Latin America.
Validation 
Audience Language is validated at both levels: the underlying language detection, and the aggregated rankings built from it. Both go through the same two-track process used for every data point.
Getting the underlying language detection right is not enough. We also validate the final rankings, comparing published audience languages against independently built reference data for the same creators. This catches errors that only show up in aggregate, and it is why the field ships with a sample floor instead of shipping every creator with any data at all.