Researchers agree this effect is real. They do not agree on how strong it is, and that argument is still going on. So you can trust the direction, but not any exact number.
In 2008 a surgeon and musician at Johns Hopkins put jazz pianists inside an fMRI scanner and asked them to improvise. He expected the planning and control regions to work harder. The opposite happened. The area associated with self-monitoring and self-censoring, the part that asks whether you are doing this correctly, went quieter. The area associated with self-expression became more active. The musicians were not thinking harder while improvising. They had something switched off. It is worth being honest about the size of this: the study involved six pianists, which is very small. Later work looking at other kinds of improvisation, including freestyle rap, has found broadly similar patterns, which is why the finding has held its reputation. But it remains a small original sample describing something genuinely difficult to measure.
The state people are in when they do their best unscripted work is not the state of trying harder. It is closer to the state of not monitoring yourself. That matters for how you judge people, because most assessment settings do the opposite. A formal interview switches the internal censor on and the pattern recognition off, so it tends to reward someone who can deliver a rehearsed answer rather than someone who can respond well to something they did not expect. If you want to see how a person actually works, you have to put them somewhere unpredictable and watch, rather than listen to them perform something prepared.
Source: Limb and Braun, 2008, PLoS ONE. Six musicians. Similar patterns found in later improvisation studies.
This Is Your Brain on Music
Daniel Levitin, 2006
On what music actually does inside the head, from a neuroscientist who was a record producer first.
Draw your own card. It does not take long, and it rewards taking your time.