You mention CGI, so let me give you a good rule of thumb to work off of.

If a shot involves a computer-generated element that must be added to real footage, and the camera is moving, that footage has to be motion tracked. This means using a program to place tracker "dots" on features of the scene that are solid and unmoving to approximate the camera motion and recreate it on the computer.

For this reason, if all you can see is sky/clouds or water (elements that are constantly moving) in a shot, it is damn near impossible to motion track it, making CGI unlikely.

Now, it could be 100% CGI, but recreating hand-held camera motion like that is not a simple task, nor something for beginners or basic hoaxers.

So whatever it is, it's probably legitimate footage.