Not at all true. For that to work, there would have to be, yes, eye tracking and a lot of image processing to change where your gaze appeared to be, but also knowledge of the physical size of the screen, the location of that particular person's feed on the screen, and where the screen is in relation to the camera. Now, it's possible that a lot of that could be inferred if it's assumed the screen is not moved in relation to the camera, and a lot of work went into tracking window locations even as you move things around, but that's a LOT of effort to expend for such a minor feature (let alone the uncanny valley effect if you don't get the image processing perfect, including the transitional "eye moved from the document you were looking at, to looking at them" - the software would have to somehow note that rather than just looking over an inch, you actually looked up 7 inches, and render that realistically).
Easiest way to confirm this is not true, though (rather than just describe the challenge involved) - have a Zoom/Teams/whatever call 1-on-1. As you're talking to the person, and they're talking back to you, note where their eyes are. 95% of the time they're going to be below where your eyes would "meet", i.e., the camera is at the top of their monitor/laptop, and they're looking at you on screen. In fact, I've had some people -never- 'meet' my gaze, because they never become self-aware enough to "oh, I should look at my camera when I am talking".