Industries / Health / Healthcare / Behavioral health / Autism therapy
Therapists and supervision
Most ABA hours are delivered by technicians who work under a board-certified behavior analyst, who assesses each client, writes the plan and checks the work; this domain is how those people are trained, certified, licensed and supervised.
7
Milestones achieved, from 1983 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1983
Florida tests behavior analysts
In the early 1970s a state investigation found abuses carried out in the name of behavior modification at a south Florida institution for adolescents with intellectual disabilities, including beatings with a wooden paddle and public shaming. Florida set up a statewide peer review committee of behavior analysts and, in 1983, gave the first professionally developed examination for its Certified Behavior Analyst credential, with 59 written-answer and 95 multiple-choice questions. By the mid-1990s programs in California, Texas, Pennsylvania, New York and Oklahoma used the Florida exam.
Lessonleading behavior analysts of the time blamed the abuses on there being no definition of who was qualified to practice, so the field's first credential was built as consumer protection.
Credit
Florida Developmental Services Program Office; Henry Pennypacker, University of Florida; Jerry Shook
1998
A national certification board
Jerry Shook incorporated the Behavior Analyst Certification Board (BACB) as a Florida nonprofit on 5 May 1998 to offer two credentials: Board Certified Behavior Analyst (BCBA), at master's level, and Board Certified Assistant Behavior Analyst (BCaBA), at bachelor's level. It took over the Florida exams by an agreement signed on 15 January 1999. Growth followed the insurance laws: R. K. McBain, H. Yu and colleagues counted 1.34 BCBAs per 100,000 US children in 2003 and 29.88 in 2017, and found state autism insurance mandates were associated with about a 16% increase in BCBAs over that time. On 1 October 2026 the BACB counted 88,013 BCBAs and 4,931 BCaBAs.
Credit
Gerald (Jerry) Shook
2009
States begin licensing analysts
A licensure law makes a professional title, or the practice itself, legal only for people who hold a state license. By the BACB's table, Nevada and Oklahoma passed the first behavior analyst licensure laws in 2009, and by 2026, 40 states and the District of Columbia had one, the most recent Arkansas (2025) and Colorado (2026). Many were written alongside autism insurance laws, because some health plans would pay only licensed practitioners.
Lessonwhere a state put the license under its psychology board, as Arizona did, board rules on supervised experience held back the number of licensed analysts for several years until the rules were changed.
Credit
state legislatures; the BACB's model licensing act of 2009
2014
A credential for technicians
In 2014 the BACB created the Registered Behavior Technician (RBT) credential for the front-line staff who deliver most therapy hours. An RBT must be at least 18 and works under close supervision; the BACB requires that at least 5% of an RBT's monthly service hours be supervised, so a technician working with a child for 40 hours a week needs at least 2 hours of supervision a week. The count grew from 328 in 2014 to more than 100,000 in 2021, and on 1 October 2026 the BACB counted 259,577 RBTs, nearly three for every BCBA.
Credit
BACB
2022
A code that defines assent
A new Ethics Code for Behavior Analysts, and a separate RBT Ethics Code, took effect on 1 January 2022, replacing the Professional and Ethical Compliance Code of 2016. Its glossary defines assent as spoken or unspoken communication that can be taken to show willingness to take part in research or services by someone who cannot give informed consent, for example because of age or intellectual disability, and leaves it to research committees or service organizations to say how assent is judged.
Credit
BACB
2023
Certification goes national
In December 2019 the BACB announced that from 1 January 2023 it would accept new certification applications only from people living in the US and Canada, and that from 2022 it would test only in English. It said no major profession runs one set of global requirements and that it could not serve every applicant in their own language and law. Later exceptions have closed: UK residents could no longer apply from 1 January 2026, and new certifications in Australia stop on 1 January 2027, at the request of the Association for Behaviour Analysis Australia. National bodies took over: Ontario has regulated behaviour analysts through the College of Psychologists and Behaviour Analysts of Ontario since 1 July 2024, and the UK Society for Behaviour Analysis and the Association for Behaviour Analysis Australia run their own credentials.
Lessonthe BACB concluded that a credential built on one country's education and law could not fit every country's.
Credit
BACB board of directors; national associations
2025
Supervision mapped by county
Using BACB records for 1 January 2022, M. E. Yingling, M. H. Ruther and E. M. Dubuque found that 24.2% of the 3,138 US counties had neither a BCBA nor an RBT living in them, more than 90% of those counties rural. About half of the BCBAs qualified to supervise oversaw no RBTs, while 22 oversaw more than 100. Federal auditors found gaps inside clinics as well: in Indiana's Medicaid program, 7 of the ABA facilities sampled let technicians without the RBT credential deliver therapy to children.
Lessona national count of credential holders says little about whether supervised care exists where children live.
Credit
M. E. Yingling, M. H. Ruther and E. M. Dubuque; HHS Office of Inspector General
Watch for
Analyst-only video therapy
AnswersNow, a US company, delivers short ABA sessions entirely by video, with a BCBA working with the child and coaching parents and no technician. A study the company funded, run with Clemson University and co-written by an AnswersNow employee who holds its stock, reviewed records of 504 clients who received about 2.5 hours of services a week, against the 30 to 40 hours often recommended for comprehensive ABA. It reported gains on the Vineland-3, a scale of everyday living skills filled in by a parent, and high caregiver satisfaction, and the authors say the design cannot show that the therapy caused the gains.
Achieved whena group with no tie to the company compares this model with in-person, technician-delivered ABA in a controlled study and reports outcomes and dropouts for both.
Claimed by
William H. Edwards (Clemson University) and Brittany C. Wierzba (AnswersNow), June 2026
Open
01
Technician exits, counted
A yearly count of how many Registered Behavior Technicians leave the credential, nationally and for each US state, built from the BACB's published totals and any state counts it agrees to release, with the method published and the results set beside the turnover reported in workforce surveys.
The briefCredit · unclaimed02
Fidelity scored from video
A public set of short, scripted videos of discrete-trial teaching, the ABA format in which a technician gives an instruction, prompts if needed, rewards a correct answer and records the result, acted by consenting adults with no child or client present. Each script plants known errors, such as a missed prompt, a late reward or an unrecorded trial. The set is used to test whether AI models that read video can score procedural fidelity, the share of steps carried out as written, as accurately as BCBAs.
The briefCredit · unclaimed
Open 01 · the brief
Technician exits, counted
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A yearly count of how many Registered Behavior Technicians leave the credential, nationally and for each US state, built from the BACB's published totals and any state counts it agrees to release, with the method published and the results set beside the turnover reported in workforce surveys.
Why it is still open
Technician turnover has been measured only by surveys, such as a 2025 study of 36 technicians, and a 2025 survey of BCBAs says the BCBA turnover rate is unknown. The BACB publishes yearly totals: 196,579 RBTs at the end of 2024, 101,341 newly certified in 2025 and 246,109 at the end of 2025. By this file's arithmetic, 51,811 RBT certifications, 26% of the start-of-year count, ended during 2025, against 580 BCBA certifications (under 1%), and the same arithmetic gives 26% to 31% of RBTs a year from 2021 to 2025. The BACB's region tables give monthly totals by state but not new certifications or exits, and searches of Europe PMC and GitHub in November 2026 found no published estimate of exits.
Built with what exists
The BACB's annual data reports and region-specific monthly tables; US Census Bureau child population estimates; Python with pandas; a written request to the BACB for state counts of new and lapsed certifications.
First step
Enter the BACB's 2020 to 2025 annual figures in a spreadsheet, compute national exits for each year, and write to the BACB asking for state counts of new certifications.
Done looks like
A public table of national RBT exits for every year the BACB reports new certifications (2020 to 2025 on its site today) and state exits for at least 2021 to 2025, each figure with its arithmetic and source; a split between lapses, moves up to BCaBA or BCBA, and revocations wherever the BACB supplies it; and a comparison with at least three published turnover surveys.
Open 02 · the brief
Fidelity scored from video
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of short, scripted videos of discrete-trial teaching, the ABA format in which a technician gives an instruction, prompts if needed, rewards a correct answer and records the result, acted by consenting adults with no child or client present. Each script plants known errors, such as a missed prompt, a late reward or an unrecorded trial. The set is used to test whether AI models that read video can score procedural fidelity, the share of steps carried out as written, as accurately as BCBAs.
Why it is still open
Supervision is thin. The BACB's minimum is 5% of an RBT's hours, about a quarter of US counties have no BCBA or RBT, and in a 2025 survey BCBAs named lack of time, training and systems as barriers to collecting fidelity data. AI so far appears on the training side: a 2026 trial taught 23 university students a procedure with either a chatbot or virtual reality. Searches of Europe PMC and GitHub in November 2026 found no published test of AI scoring of staff fidelity from video.
pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov
Built with what exists
Phones on tripods; ELAN, free software for marking events on video; a discrete-trial fidelity checklist from the published research; video-reading AI models such as Google's Gemini and open-weight vision-language models; Zenodo to publish the clips and labels under an open license.
First step
Write 20 scripts from one published fidelity checklist, film them with two adult volunteers, one playing the learner, and score them by hand.
Done looks like
At least 200 clips covering 10 error types plus error-free clips, each scored step by step by two BCBAs who agree on at least 90% of steps; at least two AI models scored against the BCBAs' agreed answers, with per-step agreement and missed-error rates published.