Pictures that only look deep
Not every picture that looks solid is backed by a three-dimensional model. A wire frame is one way to get depth cues. It is not the only way, and for a lot of finished pictures it is not the way that was used.
Photorealistic effects in flat graphics are often achieved with no wire-frame model at all. The result can be impossible to tell from a render. Some graphic software includes filters that apply those effects to flat vector art or to raster art on transparent layers. A drop shadow, a bevel, a bit of blur that implies a lens, a gradient that implies a room. An artist can also skip the filter and paint the same cues by hand, copying what a render would have done, or what a photograph would have done, without ever building the object. The labour moves. The file does not grow a third axis.
This is easy to sneer at if you have just learned to model, and it is a mistake. A flat picture that only has to be seen from one direction does not owe anyone a mesh. The mesh would be a private scaffolding the viewer never gets to walk around. If nobody is going to walk around, the scaffolding can stay in the artist's head. Flat pictures is the honest home of that decision. A model earns its cost when a second viewpoint, a simulation, or a printer is going to ask the file a question the first picture did not answer.
Two and a half
Some games use what gets called 2.5D. The environment is three-dimensional in a restricted way. The projection may be isometric. The camera may be locked to a fixed angle. The restriction can be there so the engine has less to do, or it can be there because the game plays better, or looks more like itself, with a camera that refuses to tumble. Isometric pixel art is an old and durable version of the same idea: depth you can read, a viewpoint you cannot change, and a style that would be damaged by a free camera.
By contrast, a game that uses 3D graphics without those restrictions is said to use true 3D. The phrase is a bit proud. What it actually marks is freedom of projection. You can put the camera where the scene allows, and the depth test will sort the surfaces out. Rendering is what makes that promise true or false. A free camera over a set of flat cards is still a set of flat cards. It is also, sometimes, exactly the picture someone wanted, which brings the argument back to the first paragraph. Depth in the file and depth in the impression are different properties. A careful picture can have either one without the other.
The practical test is dull and reliable. Ask what a second camera would see, one that the artist did not place. If the answer is "nothing, there is no scene," you are looking at a flat picture that learned the manners of a solid. If the answer is another coherent view of the same objects, you are looking at a model, whether or not anyone bothered to call it one.