Frequency is about text
Frequency is usually calculated from corpora: books, subtitles, websites, news, or other text collections. It is excellent for studying exposure and usage.
A technical word can be frequent within one domain and rarely recognized outside it. That makes frequency powerful, but audience-dependent.
Prevalence is about people
Prevalence starts with vocabulary knowledge. In the Brysbaert norms, participants judged whether they knew English lemmas, producing an estimate of the share of people who recognized each word.
For general readability, prevalence can be easier to interpret: known by about 90% of readers is clearer than a corpus count.