An Airbus A350-LM, a widebody/higher-capacity plane that can go 15,000 km, has 141,000 liters of fuel. That'd be 0.106km/l, or .249mi/g.
I suspect these figures are a bit off, as I doubt "range" == "all fuel burned". But it's at least somewhat representative.
At 200 passengers you'd get 2.3 l per 100km per passenger (100 miles per gallon per passenger), in line with a small sedan car.
Using my figure, if you want to get per-passenger, if we say it seats 200, we can multiply .217km/l by (1 plane/200 persons) and get 43.4 km/l / person, or, 102mpg/p, which I think most people will understand reasonably intuitively, compared to what they already know. This is indeed a rather impressive claim. It makes me wonder why they don't just call this thing the 100mpg plane, if true.
To the best of my knowledge, most of the world reports transport efficiency in l/100km. And the poster even nicely included the US-centric mpg value for people in not-most-of-the-world.
English speaking countries (or at least most of them) usually do it the other way around and compare the distance one can travel by fuel unit.
fuel quantity per distance ~~ inefficiency or cost
Probably tied with hot air balloons for least efficient transport I've ever taken.
(For the leg from LAX to MEL i had the aisle seat beside the shitter galley in the back. 14 hours of people fanning their farts in my face.)
The fuselage and wing of a cargo jet in turbulence basically "banana" in perpendicular planes.
I don’t know how air freight is factored into fuel efficiency calculations, but it has become an increasing part of passenger airlines’ business models in the past decade, too.
(Source: I spent way too much time keeping tabs on Delta as a top level frequent flier)
A skilled glider pilot can go (slowly) as far as they like with no fuel. The typical limit is when you get tired and need to sleep (gliding can't yet be done well by a computer, because it takes quite some skill to find the necessary thermals and wind currents).
Typically it is very hard to glide at night since a large part of gliding is the air movement caused by local and regional thermal differences.
Ardupilot (or one of its offshoots, I can't recall exactly) does have a lift detecting autopilot in beta. I don't know how it works with a specific route, but it seems to lock in thermals pretty well.
But with a computer - train ML on human pilot behaviours? - it could stay there for quite some time.
Unfortunately, most planes fly far from their maximum capacity (even though most passengers are in fact flying in overcrowded planes).