Instagram Secretly Chooses Who Sees Your Posts

Instagram’s algorithm doesn’t just show your posts to people who like your style — it tries to predict what kind of visuals will do well *in general*. But a new study from researchers at the University of South Florida shows this approach is mixing up two different things: audience preference (what people tend to like across the whole platform) and creative effect (what your specific followers respond to when *you* change your style).

The team analyzed over 232,000 sponsored beauty posts from 1,527 creators on Instagram. They found that in four out of eleven visual style categories — including skin tone — the two effects pointed in opposite directions. For example, the pooled data suggested lighter skin tones perform better overall, but when individual creators shifted to darker tones, their own audiences actually engaged more. Following the AI’s blended advice led creators to miss out on 31% of the potential engagement they could have gained by listening to their *own* audience.

The researchers call their method DIRECT: a way to separate audience preference from creative effect using advanced math and AI. They say this could help creators, brands, and even platforms make smarter decisions about what visuals to use — not based on what works for everyone, but what works best for *your* followers. It’s like finally getting a personalized coach instead of a one-size-fits-all rulebook.

The study suggests Instagram’s current recommendation system might be giving creators the wrong creative advice — and changing that could help people get more likes, comments, and sales from their posts.