The way I like to think about it is that convincingness is a 2-place function - a simulation is convincing to a particular mind/brain. If there's a reasonably well-defined interface between the mind and the simulation (e.g. the 5 senses and maybe a couple more)

The 5 senses are brain events. There aren't input channels to the brain. Take taste. How many different tastes of food can you perceive through your taste sense? More than 5. Why? Your brain takes data from nose, tongue and your memory and fits them together to something that you can perceive through your smell sense.

You have no direct access to the data that your nose or tongue sends to your brain through your conscious qualia perception.

If someone is open by receiving suggestions and you give him a hypnotic suggestion that a apple tastes like an orange you can awake him. If he eats the thing he will tell you that the apple is an orange. He might even get angry when someone tells him that the thing isn't an orange because it obviously tastes like an orange.

it's cheating to bypass that interface and make the brain more gullible than normal, for example by introducing chemicals into the vat for that purpose.

You don't need to introduce any chemicals. Millions of years of evolutions have trained brains to have an extremly high prior for thinking that they aren't "brains in a vat".

Doubting your own perception is an incredibly hard cognitive task.

There are experients where an experimentor uses a single electron to trigger a subject to do a particular task like raising his arm. If the experimentor afterwards ask the subject why he raised the arm the subject makes up a story and believes in that story. It takes effort for the leader of an experiment to convince a subject that he made up the story and there was no reason he raised his arm.