The split between creator and business accounts is one of the most-used filters in the platform. A text classifier labels each profile as creator or business. It was built from a hand-labeled sample of about 10,000 profiles and tuned per platform.
How Profiles Split by Platform
The classifier labels every profile in the database. What changes between platforms is the balance between creator and business accounts, which reflects how each platform is used.
Creator vs Business Distribution
Instagram has the largest business share because it is widely used by brands, agencies, and local shops. YouTube and TikTok skew heavily toward individual creators.
How the Classifier Works
A deep-learning text model uses what is known about a profile and outputs a single label with a confidence score. The exact inputs we feed it differ by platform because each platform exposes different fields.
The Model in Plain Terms
What the Model Reads
Each platform has its own classifier. The model uses profile text, content metadata, and public engagement signals. The exact mix of available signals varies by platform, with some providing richer metadata than others.
How the Model Learned
Most of the work in a classifier like this is in the labels, not the model. The training data was built and rebuilt twice to refine label accuracy.
Two Training Iterations
Worked Example
Two illustrative profiles run through the same classifier. One triggers strong signals across multiple platforms. The other has fewer distinguishing features, but still lands in the right class.
How the Classifier Reads Two Illustrative Profiles
NikolaTravels triggers more classifier features because of a richer bio, niche-specific vocabulary, and cross-platform presence on three networks. The model reads travel terminology, location references, and content-style hashtags as strong individual-creator signals. BorisLifeLog has fewer distinguishing features: a shorter bio with generic lifestyle language and presence on only two platforms. The classifier still gets the right answer, but with moderate confidence instead of high. This is the difference between a profile that "screams creator" and one that merely does not look like a business.
Validation 
Every accuracy number on this page comes out of a two-track validation process. We compare at least two independent AI sources on tens of thousands of records, and we manually validate at least 1,000 records per subset, with subsets defined to cover every meaningful use case. F1-score balances precision (when the model predicts business, is it actually a business?) and recall (of all real businesses, did we catch them?).
Measured F1-Score by Platform
Creators are by far the larger class on every platform, so the model has more training examples and more opportunities to be correct. Businesses are rarer and visually similar to creators (especially small personal brands and solo founders), which is why business F1 sits in the 60s on YouTube and TikTok. We mitigate this with a high-confidence threshold for business labels: when in doubt, the profile is kept in the creator class rather than misclassified.
v1.0 production model