Thought experiments are powerful thinking tools. To clarify the perhaps most important characteristic of them, I like to think of them as existing at the intersection of thoughts and experiments.
Here is an obvious Venn diagram to show that relationship.

It has some clear implications.
Experiments are specifically about testing hypotheses or ideas. You get to control variables and assumptions – the environment and context of your experiment – then apply rigor and logical thinking to explore probable outcomes. They’re a structured means of inquiry, designed to explore the relationships between causes and effects or to expose hidden truths.
In the case of thought experiments, this structure occurs entirely in the mind or abstract representations rather than in a laboratory of the physical world.
Thinking does not mean that this is all just in your head. Thinking out loud, thinking in writing, and so forth are likewise possible ways to engage in the activity. While the thinking is real, the subject of it is mental, essentially imaginary.
As ever, the map is not the territory. This is a necessary characteristic of a thought experiment: You conduct it by engaging in mental work only, not by changing a physical reality that you might be describing.
The experiment’s limitations are those of your thinking. Imagination and reasoning have their own constraints: biases, gaps in knowledge and subjective interpretations. However, a thought experiment is not bound by the conditions you observe in the physical world. It is free to explore “impossible” or hypothetical scenarios, constrained only by the coherence of its logic and the breadth of your imagination.
Thought experiments are an important type of mental model. Engaging in a thought experiment ideally leads you to clearer understanding or insights about a given context – perhaps even the discovery of an intriguing paradox.
Thought experiments are foundational to human ingenuity and innovation.