Why this job stays next to the student
Most of a special education aide’s day is spent within arm’s reach of a child. You help with feeding, toileting, lifting and moving between rooms. You sit beside one student and re-explain a math step the fourth time, in different words, because the third way did not land. Software can rephrase a worksheet. It cannot steady a hand on a pencil grip or walk a student to the nurse.
The second reason is reading people. A paraprofessional notices the jaw tighten, the foot tapping, the shift in breathing that comes before a meltdown, and moves early. That judgment is built on knowing one child over months. It is also legally bound: the supports in an IEP are delivered by named staff under a teacher’s direction, and someone has to be accountable in the room.
So the honest story here is erosion at the edges, not a job disappearing. Paperwork, progress logs and material prep are getting faster. The hands-on hours are not.
What AI does, what it helps with, and what stays with you
Our split puts 0% of task time in the group where AI can handle the work end to end. That group is mostly desk work: turning a lesson into a simpler reading level, generating extra practice problems, and drafting the routine wording in session notes before a human checks them.
Another 9% of task time sits in the group where AI assists but a person stays in charge. Think live captioning for a student with hearing loss, communication devices that predict the next word, translation for a family meeting, and tallying behavior data you have already recorded so the teacher can see a trend. The tool speeds the record. You still make the call about what happened and why.
That leaves 91% of task time with people. Personal care and physical assistance are in there. So is supervising students in hallways, the cafeteria and on the bus line. So is de-escalation, hand-over-hand prompting, and the small negotiation that gets a student back to the table. Our coverage score, which asks whether AI can do the work at all today, prints as 12 out of 100; the coverage method page explains how that is built.
What the evidence shows, and what it does not
There is no head-to-head test of AI against a trained special education aide on the work that matters: IEP goal progress, time on task, behavior incidents, or a safe transition between rooms. Our evidence grade for this job is D, and a D means the quality question has not been measured. For that reason we publish no parity number here.
What would settle it is specific. A trial in real classrooms, comparing AI-supported delivery with a trained paraprofessional on measured student outcomes across a term, with the same students and the same goals. Until something like that exists, claims about AI matching an aide are opinion. The method pages set out how we grade evidence and why a D never gets a score.
The labor data is clearer. BLS puts median pay for this occupation at $36,780 and projects employment change of -0.3% between 2025 and 2035 (BLS, 2025), against roughly 1.42 million jobs. That is a flat market, shaped mostly by school budgets and enrollment rather than by software.
When this could change
Most likely after 2035 (8 in 10 of our scenarios). For what that window measures and how it is built, see the replacement-year method.
Two things could pull it earlier. The first is cost: classroom AI tools sit far below the annual cost of staffing, so districts under budget pressure will keep adding software before they add people. The second is assistive technology getting better fast, especially communication devices, real-time captioning and reading supports that let some students work with less adult prompting.
Two things hold it back. The physical share of this job is large, and our robotics read puts the needed machine in the dexterous humanoid tier, which is not a near-term classroom product. And the rules matter: IEP services, supervision ratios and student safety all assume a named adult who can be held responsible.
How to stay needed
Lean into the parts of the job that nothing on a screen touches. Behavior support and de-escalation with students you know. Personal care and safe physical assistance. Live supervision in halls, at lunch and during transitions, where the judgment call is made in two seconds.
Two skills raise your floor. One is assistive technology fluency: setting up and troubleshooting AAC devices, captioning and switch access, so you are the person who makes the tool work. The other is clean data collection on IEP goals, written tightly enough that a teacher or a parent can act on it. Both are easier now with AI drafting, and more valuable because of it.
What to do: ask your district which AI tools it has already licensed, then volunteer to be the person who trains the rest of the team on them.
If you are weighing a move, nearby work includes general teaching assistants, special education teachers in elementary schools and preschool special education teachers. You can put any two of them side by side on our job comparison tool, see how the rest of the field scores on the other education support occupations page or across jobs in schools, and check where this kind of work sits on our list of jobs that mostly need a person.