The Short Version
There is a real association between high ultra-processed food intake and dementia risk in several large observational studies and meta-analyses (pooled RR ≈1.44). However, the claim overstates the evidence in key ways: the underlying studies are observational (not proving causation), the pooled estimate has extreme statistical heterogeneity (I²≈97%), newer studies find no association for total UPF intake, and "regular consumption" is vaguer than the "high vs. low" comparisons actually studied. The link is plausible but not as settled or causal as the claim implies.