Why the checkpoint still runs on people
Asking whether AI will replace transportation security screeners is really a question about tasks, not titles. A screener reads X-ray and computed tomography images of carry-on bags, which is machine-friendly work. The same screener also runs hand-wand checks and pat-downs, searches opened bags by hand, verifies IDs and boarding documents, and decides when a passenger needs a second look. That second half happens in public, face to face, under time pressure.
Judgment is the other sticking point. An algorithm can flag a dense shape in a bag. Someone still has to open that bag in front of its owner, explain why, resolve the alarm, and either clear the passenger or hand the situation to law enforcement. Screeners also keep the lane moving, direct confused travelers, test and calibrate equipment, and write up incidents that may end up in a legal file. Accountability for a missed threat sits with a trained person and the agency behind them.
The money and the headcount set the scale. About 50,290 transportation security screeners work in the US, with median pay of $66,770 (BLS, 2025). The Bureau of Labor Statistics projects employment down 4.2% between 2025 and 2035. That is the honest shape of the pressure here: automated lanes and better image detection trimming hours and entry-level openings, not checkpoints running themselves.
Machine work, assisted work, and the part that needs a person
Automated detection handles the image-reading end. Software scans bag scans for prohibited items and shapes, and document readers match credentials against travel records. Our coverage score tracks how much task time sits in that bucket: 0% of the work. If that reads as a small slice, it is because image review is only one step in a lane. You can see how we measure it on the coverage method page.
A bigger share is assisted work, where the system suggests and a screener decides. Automatic threat detection marks a region of an image and the screener adjudicates it. Camera analytics can time each step of passenger processing, but a supervisor acts on what that shows. Tasks where AI helps rather than acts account for 35% of task time.
The rest stays with people: 65% of task time. Pat-down and hand-wand screening, physical bag searches, questioning a passenger whose alarm will not clear, detaining someone and calling in police. None of that is a software problem. It needs hands, presence, and a person who can be held responsible for the call.
How strong is the evidence?
Thin, and we grade it that way. Our evidence grade for this job is D on an A to D scale, and a D means there is no published head-to-head test of AI against trained screeners on this job’s real tasks. Vendor detection rates and agency pilots are not the same thing as an independent comparison. So we publish no parity number here, because a number without a test would be a guess dressed up as data.
What would settle it: independent, published results comparing automatic detection and trained screeners on the same image sets, reporting both detection rates and false alarms, plus outcomes for the part that follows an alarm, where a person resolves it with the passenger. Until that exists, the sensible reading is assistance with image review, not substitution for the lane. Our rules for scoring a machine against a person sit on the quality parity page, and the wider method is set out in our methodology.
When the balance could shift
Most likely after 2042 (8 in 10 of our scenarios). What that window measures is explained on the replacement-year page.
Two things could pull it earlier. Automated screening lanes keep spreading, and each lane that reclaims bins, diverts bags and auto-clears images removes screener minutes. The cost gap also pushes in one direction: running detection software across a lane costs a fraction of staffing it, and the cost panel above this section shows how wide that gap is per hour of work.
Two things hold it back. The physical half of the job needs hardware at the dexterous humanoid tier, which is the hardest robotics step and nowhere near routine deployment at a public checkpoint. And security screening is a regulated, high-accountability setting. Changes need certification, auditing and public trust before anyone removes the person who does the pat-down. Our take on the hardware side is in the guide to humanoid robots and physical jobs.
How to stay needed in screening work
Lean into the work that does not move. Alarm resolution comes first: being the person who opens the bag, explains the step, and closes the case cleanly. Second, physical screening done well, including pat-downs and secondary inspection that hold up under review. Third, passenger handling in bad moments, the de-escalation and plain-language explaining that keeps a lane from seizing up.
Two skills raise your floor. One is equipment fluency: operating, testing and troubleshooting detection systems, and reading their output with healthy skepticism about false alarms. The other is incident writing and testimony, because documentation is what turns a judgment call into a defensible record. Both of those move you toward lead and supervisory work, where the agency still needs people in front of the machines.
What to do: Pick up the lane roles that pair equipment knowledge with passenger contact, since those are the hours automated detection does not take.
If you are weighing a move, nearby work includes security guards, customs and border protection officers and first-line supervisors of security workers. You can put any two of them side by side on our compare tool, see the wider other protective service workers family, check the government sector page, or browse the jobs that mostly need a person in our safest jobs list.