It’s the same way you know the things outside your window are real. You look at the light coming to you from that object and make inferences as best you can. As long as new observations and inferences line up with old observations and inferences, then you can be reasonably confident that your growing model of the outside world is accurate. When something doesn’t add up then you revise your model and keep iterating with new observations.
There’s no difference whether the object appears to be within our solar system or far outside it. We see something and we interpret what we can from the available observations. Occasionally, if something is close enough and interesting enough, we send a robot to orbit the thing or maybe land on it and gather better observations, like how Rosetta/Philae visited a passing comet.