Modern transport-category aircraft rely heavily on autopilot systems to reduce crew workload on long flights, maintain precise navigation paths, and execute instrument approaches with repeatable accuracy. Yet the autopilot is a tool, not a substitute for pilot judgment — and some of the most serious accidents in aviation history have involved crews who misunderstood what the autopilot was doing, failed to monitor its behavior, or were caught off guard when it disconnected unexpectedly. Understanding when, how, and why to engage or disengage the autopilot — and what happens in each case — is a core competency for any Airline Transport Pilot certificate holder.
The FAA's Instrument Flying Handbook (FAA-H-8083-15) and Instrument Procedures Handbook (FAA-H-8083-16) both address automation management as a fundamental instrument skill, while the Pilot's Handbook of Aeronautical Knowledge (FAA-H-8083-25) establishes the conceptual foundation. Airline-specific Standard Operating Procedures (SOPs) layer additional requirements on top of regulatory minimums, but the principles described here apply broadly across turbine autopilot-equipped aircraft.
How Autopilot Systems Work
An autopilot system senses aircraft attitude, airspeed, altitude, and navigational deviation through a network of sensors — inertial reference units, air data computers, radio navigation receivers, and GPS — and then commands the flight control surfaces to correct any deviation from the selected reference. Most transport-category autopilots operate through servos that are mechanically or electrically coupled to the control surfaces. When the autopilot is engaged, these servos resist manual control inputs and actively move the controls to achieve the commanded flight path.
Modern autopilots operate in modes. Pitch modes may include altitude hold, vertical speed, flight level change (FLCH), and glideslope capture. Roll modes include heading select, VOR/LOC tracking, LNAV (lateral navigation via FMS), and approach mode. Understanding which mode is armed (awaiting capture conditions) versus active (currently controlling the aircraft) is central to monitoring automation effectively. The Flight Mode Annunciator (FMA) displays this information prominently on the Primary Flight Display, and professional pilots are expected to cross-check the FMA after every mode change.
Engagement Procedures
Engaging the autopilot at the wrong moment — or without verifying the aircraft's state first — can produce an abrupt and dangerous maneuver. Before engagement, the standard practice is to confirm that the aircraft is in a stable, trimmed condition: wings level or at the intended bank, pitch attitude appropriate for the phase of flight, and airspeed within normal range. An autopilot engaged while the aircraft is significantly out of trim will command a sudden control surface deflection to correct the deviation, potentially startling the crew or exceeding structural limitations in an upset scenario.
Most transport-category autopilots impose a minimum engagement altitude after takeoff — commonly 400 feet AGL, though some SOPs specify higher values such as 1,000 feet AGL. This restriction exists because the autopilot is not certified for ground operations or the highly dynamic low-altitude environment immediately after liftoff, where prompt manual response to engine failure or windshear is essential. Pilots should always consult the Aircraft Flight Manual (AFM) for the specific minimums applicable to their aircraft.
The engagement sequence typically involves: (1) verifying the Flight Management System (FMS) or mode control panel (MCP) has the correct target values set for altitude, heading, and speed; (2) pressing the autopilot engage button, usually labeled CMD (command) or AP; (3) confirming the FMA annunciates the expected modes in green; and (4) briefly monitoring the aircraft's response to ensure it begins tracking the commanded reference without unexpected excursions. On aircraft with dual-channel autopilots, a second channel may be engaged for coupled approaches to CAT II or CAT III minima, following procedures specific to that approach type.
Disengagement Procedures
Autopilot disengagement falls into two categories: intentional (crew-initiated) and automatic (system-initiated). Each requires a different response and mindset.
Intentional Disengagement
Crew-initiated disconnects are performed using the autopilot disconnect switch, typically located on the control wheel or sidestick. A single press disconnects the autopilot and triggers an aural warning (a distinctive repetitive tone or cavalry charge sound) and a visual annunciation on the FMA. The pilot must silence the aural alert promptly — usually with a second press of the same switch or a dedicated disconnect button — to prevent it from becoming a distraction. At the moment of disconnect, the pilot takes manual control and cross-checks the aircraft's attitude on the PFD. If autothrottle is still engaged, it will continue to manage thrust unless also disconnected.
The key discipline at intentional disconnect is immediate cross-check of attitude and energy state. If the autopilot was in a pitch-down attitude due to a glideslope intercept or descent, the aircraft will tend to continue that pitch motion when the pilot takes over. Establishing hand-flying skills that account for the existing trim state prevents altitude excursions. The FAA Risk Management Handbook (FAA-H-8083-2) highlights automation complacency as a major human factors threat — regularly hand-flying at various phases of flight helps maintain proficiency and reduces the skill decay that makes disengagement dangerous.
Automatic Disengagement
The autopilot will disconnect automatically when certain system faults are detected, when the aircraft reaches limits beyond which the autopilot cannot safely operate (such as extreme pitch or bank angles), or when the crew applies a manual control force exceeding a threshold that the autopilot's override mechanism recognizes. These automatic disconnects are a primary cause of automation surprises. The National Transportation Safety Board has noted numerous incidents in which a crew, heads-down managing an FMS entry or addressing an unrelated alert, was suddenly confronted with an aircraft that had disconnected its autopilot and was deviating from the intended flight path.
The correct response to any autopilot disconnect — intentional or not — follows a consistent flow: aviate first (control the aircraft), then navigate (confirm the flight path), then communicate (advise ATC if needed), and finally diagnose the reason for the disconnect. Attempting to immediately re-engage the autopilot after an uncommanded disconnect without identifying the root cause can result in a second disconnect or, worse, engagement in an incorrect mode that worsens the situation.
Coupled Approaches and Disconnect Timing
During ILS coupled approaches, the autopilot typically remains engaged down to the decision altitude (DA) on a CAT I approach. For CAT II approaches, the autopilot may be maintained to approximately 100 feet RA, and for CAT III operations, autoland keeps the autopilot engaged through touchdown and rollout. The exact altitudes and procedures are aircraft-type specific and defined in the AFM. The Instrument Procedures Handbook (FAA-H-8083-16) emphasizes that mode awareness is critical during the approach — specifically, pilots must verify that APPR mode (or its equivalent) is active, that glideslope and localizer are captured (not merely armed), and that the aircraft is descending on the correct path before passing the final approach fix.
A common error is failing to confirm that localizer mode has transitioned from armed to active, allowing the aircraft to overfly the localizer centerline. Another is engaging a go-around mode prematurely, which can cause a pitch-up while still over the runway environment. Both errors are preventable by disciplined FMA monitoring and adherence to callouts.
Key Numbers and Rules
- Minimum engagement altitude: Typically 400 feet AGL after takeoff per many AFMs and SOPs; always verify the aircraft-specific value.
- FMA cross-check: Required after every autopilot or autothrottle mode change — confirm armed and active modes match intent.
- CAT I coupled approach minimum: Autopilot typically disengaged at or before DA (commonly 200 feet HAT for a standard CAT I ILS).
- CAT II/III: Dual-channel autopilot engagement required; autoland systems have their own certification requirements under 14 CFR Part 121 and the applicable operations specifications.
- Trim state check: Before engagement, verify aircraft is trimmed; an out-of-trim aircraft will produce a control transient on engagement.
- Aural alert response: Silence the autopilot disconnect tone promptly, but only after establishing positive manual control.
Memory Aid
Use the phrase "SAFE-D" for autopilot disengagement response:
- S — Silence the aural alert (after establishing control)
- A — Aviate — establish positive manual control of attitude and energy
- F — FMA — check the Flight Mode Annunciator for mode state
- E — Evaluate the reason for disconnect before re-engaging
- D — Decide whether to re-engage, hand-fly, or declare an abnormal situation
This flow keeps the priority where it belongs — on aircraft control first, diagnosis second.
Common Test Traps
- Confusing armed vs. active modes: A mode shown in white on the FMA is armed (waiting to capture); a mode shown in green is active. Assuming an armed mode is already controlling the aircraft is a frequent source of deviation.
- Re-engaging immediately after an uncommanded disconnect: The correct response is to hand-fly and diagnose first. Immediate re-engagement without understanding the cause can repeat the problem or introduce new modes incorrectly.
- Overlooking the trim state before engagement: Test questions describe scenarios where the autopilot is engaged in an out-of-trim condition and ask what happens — the answer is an abrupt pitch or roll transient as the autopilot commands correction.
- Assuming the autothrottle disconnects with the autopilot: These are typically independent systems. Disconnecting the autopilot does not automatically disconnect the autothrottle, which will continue to modulate thrust to maintain speed unless separately disengaged.
- Incorrect minimum altitude for CAT II/III autoland: ATP candidates sometimes confuse the DA values for CAT I, II, and III operations or fail to account for the requirement for dual-channel autopilot engagement before the final approach fix on CAT II/III operations.