I have seen this criticism a lot. The argument is that intelligence is a circular concept where we define it as what IQ tests measure and then use IQ tests to prove intelligence exists. If a psychologist just picked a bunch of tasks that seemed smart-ish and called it an intelligence test, how is that different from what actually happens? Is there real theoretical grounding behind cognitive assessment or is it basically just vibes with statistics on top?
The circularity criticism misrepresents how the field developed. G was not invented by someone choosing tasks that seemed smart. It emerged from factor analysis of large diverse cognitive batteries where a general factor kept appearing whether researchers expected it or not. The theory followed the data. Modern frameworks like CHC are grounded in decades of replication across cultures and populations, and the structure shows up regardless of which specific tasks are included because g saturates almost any sufficiently complex cognitive demand. 
If intelligence were just whatever a psychologist put on a test, different tests would give uncorrelated results. They do not. Scores from completely independent batteries correlate strongly with each other. That convergence is the whole argument against circularity.