How Flying Works
What Autopilot Actually Does (And Doesn’t)
On 18 June 1914, a 21-year-old named Lawrence Sperry flew a Curtiss C-2 biplane fifty feet above the River Seine in front of a crowd of Parisians, then stood up in the cockpit with both hands raised over his head. His mechanic, Émile Cachin, climbed out and walked along the wing to balance on the tip. The aircraft kept flying straight and level with the pilot’s seat empty. Sperry and Cachin had just publicly demonstrated the world’s first practical autopilot, and they walked away with a 50,000-franc prize for it. That original device, built by Sperry two years earlier by wiring his father’s gyroscopic instruments to the rudder and elevators, is a direct ancestor of the system flying every airliner today — and more than a century later, the same basic misunderstanding people had watching that empty cockpit in 1914 still shapes how most passengers think about it now: that engaging it means nobody is actually flying the plane. Mentour Pilot’s video “How Does Autopilot Work? A Pilot Explains What It Can and Can’t Do” exists specifically to correct that. Autopilot is an execution tool, not a decision-maker. It does precisely what it’s told, with more consistency than a human hand-flying for hours could manage, and nothing beyond that.
.jpg?width=1600)
What it does
- Holds a commanded altitude, heading, speed and vertical rate with far more precision than continuous manual flying could over hours.
- Flies a programmed lateral and vertical route from the flight management system, including complex arrival and departure procedures.
- On a CAT III-equipped aircraft and runway, can fly and execute the entire approach and landing itself, tracking the ILS beams more precisely than a human could in zero visibility.
- Reduces crew workload on long sectors specifically so pilots can dedicate more attention to monitoring systems, weather and traffic — not so they can stop paying attention.
What it doesn’t
- Decide where to go, when to climb, or what to do about an unexpected situation. Every instruction it follows was set by a pilot, and it does exactly and only that.
- Handle every phase of every flight. Many airlines require autopilot to be off below a set altitude on takeoff and, without CAT III certification, below a set altitude on approach.
- React to something it wasn't told to react to. If air traffic control or weather changes the plan, a pilot has to notice and reprogram it — autopilot doesn't improvise.
- Substitute for a pilot's judgment in an abnormal or emergency situation. The first action in nearly every real in-flight emergency checklist is still a human decision, often followed by deliberately disconnecting the autopilot to fly manually.
From wing-walking stunt to CAT III autoland
Sperry's original gyroscope could only hold an aircraft level. What it couldn't do was navigate, land, or respond to anything it wasn't already correcting for. Every decade since added a layer: heading hold, then altitude hold, then the ability to intercept and track a radio beam for an approach, then full flight-management-system integration that can fly an entire route programmed before departure. The last major leap, CAT III certification, is what lets a modern autopilot fly and land the aircraft itself in weather too poor for a human to see the runway at all. It took roughly sixty years to get from a biplane over the Seine to that point, and the underlying principle never changed. The system executes; it doesn't decide.
The role it actually frees up
The genuine value of autopilot isn’t that it replaces pilots. It’s that it takes the repetitive, precision-demanding parts of straight-and-level flight off their hands so they can spend that attention on the things automation genuinely can’t do: monitoring weather developing ahead, cross-checking instruments against each other, communicating with air traffic control, and staying ready to intervene the instant something needs a human decision. On a long-haul sector, that’s the difference between a fatigued crew trying to hand-fly for ten hours and a crew that’s actually alert for the parts of the flight that matter most.
Sources & Further Reading
Common Questions
Frequently Asked Questions
Can a plane really land itself with no pilot input?
Yes, on a CAT III-certified aircraft, runway and crew — this is a real, routinely used capability for landing in fog and low visibility, not a novelty. But even then, pilots are required to actively monitor every stage and are ready to take over or go around at any sign of a fault.
Do pilots actually fly manually very much?
On a typical airline flight, hand-flying is usually limited to the first minute or so after takeoff and the final stretch before landing — the rest is autopilot-managed specifically to reduce fatigue and error on repetitive, precise tasks the automation does more consistently.
Is it true a plane can fly itself with no one in the cockpit?
Not in any commercial sense today. Autopilot requires a pilot to set and continuously supervise it; it has no independent decision-making authority, and current airliner certification assumes an actively monitoring flight crew at all times.
What happens if the autopilot itself fails?
It disconnects, typically with an aural and visual warning, and the pilot flying simply takes over manually — a scenario every airline pilot trains and is checked on regularly in the simulator.