Hiring and interviews

In mock job interviews with small groups of autistic adults without intellectual disability, raters who watch video or clips usually rate autistic candidates lower than non-autistic candidates. In a US study, raters who did not know anyone's diagnosis rated non-autistic candidates higher when they watched the interviews on video. When raters read transcripts of the same interviews, autistic candidates came out ahead.¹ In an Australian study, 255 non-autistic raters each judged 10-second clips of one autistic and one non-autistic man in a simulated interview, drawn from only two men of each kind; no women were filmed. They rated the non-autistic man more favorably on first impressions, employability and endorsement, and "hired" him more often; the authors caution that so few videos limit the finding.² A second US group, in Texas, filmed 21 autistic and 21 non-autistic adults in a mock interview for their ideal job and in a talk about a personal interest. Non-autistic students formed less favorable first impressions of the autistic adults, and in the interview more negative facial expression went with worse impressions, more so for autistic adults. Impressions of autistic adults improved more than those of non-autistic adults when they talked about a personal interest. Facial expression was coded by software developed on non-autistic faces.²²

Transcripts do not always remove the gap. In a UK study of 25 autistic and 25 non-autistic adults matched on age and IQ, four employment professionals read transcripts of answers to standard interview questions without knowing who was autistic, and rated the autistic adults' answers lower.³ The two transcript findings point in opposite directions, so how much of the gap lies in delivery and how much in the content of answers is not settled. A 2025 meta-analysis of 21 studies of first impressions, five of them using mock job interviews, found less favorable impressions of autistic people, usually moderate to large, in every format except written transcripts. For transcripts there was no clear overall difference and individual results were mixed. There was some evidence that observers who knew more about autism, or had better-quality contact with autistic people, formed better impressions.⁴ Disclosure and safety covers that research in detail.

Experiments on what a diagnosis label or autism training changes in these ratings are covered in Disclosure and safety. Whether to tell an employer is the subject of Disclosure at work.

In a UK online survey of 225 autistic, 64 otherwise neurodivergent and 88 neurotypical adults, nearly all had been through a job interview. Autistic participants rated interviews more negatively (2.21 on a scale of 1 to 4) than neurodivergent (2.75) and neurotypical (3.13) participants did, and rated group tasks lowest of all methods (1.60).⁵ People in all three groups were frustrated by the weight placed on social skills and wanted practical methods, such as work trials. Autistic participants described feeling that they had to hide autistic traits to get work, and many worried about discrimination if they disclosed. Most autistic respondents (87.3%) had never been able to give an employer feedback on the process, but so had most of the other two groups.⁵ The autistic sample was 64% women and 72% White, about 70% were in work, and 15% identified as autistic without a formal diagnosis.⁵

In interviews with 10 autistic adults in Australia, participants described camouflaging during job interviews, which took effort and left them stressed, anxious and exhausted.⁶ In a US study of 15 autistic job seekers and support staff, mainly about assistive technology, problems included deciding what to disclose and hearing little back from employers.⁷

The most direct test of changing the interview itself concerns the wording of the questions. In the UK transcript study, the same adults came back about six months later and answered adapted questions. These asked outright for specifics and examples, and broke multi-part questions into parts asked one at a time: "What are some of your strengths?" became a question about the main strengths and then a separate one about how they had been used at work. Interviewees also had a printed copy of the questions.³ At this second interview, ratings of both groups' answers were higher, and the autistic interviewees' rose more, though the design cannot separate the new questions from practice or the six-month gap. The gap in answer quality was no longer statistically significant, though a medium-sized difference remained.³ Employers' overall impressions improved on some items, such as communication skills and likeability, but not on confidence. Autistic interviewees were still rated lower on confidence, communication skills, likeability and how easy they would be to work with.³

Participants in that study suggested sending the questions in advance; the study did not test it.³ One employer program did it: a 2016 Deutsche Bank internship for autistic graduates in the UK sent candidates written questions with a week to answer, then interviewed them about their answers in rooms chosen for few sensory distractions.⁸ Participants in all three groups of the UK survey asked employers for more information about the hiring process and a choice of methods.⁵

Replacing the interview with another method does not guarantee a fairer result. In the UK survey, people rated methods they had been through. Autistic participants rated work trials 2.60 out of 4, about the same as the other groups (only 16 neurotypical participants had done one), and rated psychometric tests lower (2.13) than neurotypical participants did (2.96).⁵

Two studies of tests, using different kinds of test, reached different results. Using a hiring-technology company's records, 263 autistic job seekers, mostly men placed through an autism employment organization, scored generally similarly to 323 college-graduate job seekers, who were not matched to them, on game-based tests of cognitive ability, with some small differences.⁹ The first author was affiliated with HireVue, and the scores were archival data collected by "the recruiting-technology company"; the authors declared no conflict of interest.⁹ An online study published in 2026, with 107 employed autistic adults who reported their own diagnoses and 99 non-autistic adults, found score gaps unfavorable to autistic respondents on all three tools tested: ability tests, personality questionnaires and situational judgment tests. The judgment-test gap remained after allowing for ability, personality, gender and age.¹⁰ The authors call the findings preliminary, the personality scales were unreliable in the autistic sample, and they suggest the ability gap may be specific to their sample.¹⁰

Some autism programs build their own selection. Australia's Dandelion program used a four-week selection process that included training, and changed its hiring rules to rely less on past experience and interpersonal skills.¹¹

Maras and colleagues note that most earlier research on autistic job interviews tried to coach the interviewee rather than change the interview.³ In a small randomized feasibility trial, 71 autistic youth aged 16 to 26 in US school transition services got either their usual services or usual services plus a virtual interview simulator. The simulator group improved more in rated interview skills and in self-reported interview anxiety, and six months later 25% of them and none of the comparison group had a competitive job.¹² The comparison group was small (23 people), the estimate of the effect was very imprecise, and the two groups went to job interviews at similar rates.¹² The University of Michigan and the lead author receive royalties from the company that sells the tool, and three co-authors were its employees and shareholders.¹² In a later pilot trial led by a New Jersey group, with the simulator's lead developer as a co-author, 40 autistic youth aged 14 to 26 either took a 10-lesson online course on naming and describing their personal strengths or got usual services. The course group improved more in rated mock-interview skills; self-reported interview anxiety did not differ.²³ A larger trial across 16 schools, comparing the simulator with an active control, was set out in a 2024 protocol.²⁴

Large employers have run hiring programs aimed at autistic people since the 2000s, but published evaluations are thin. A 2017 systematic review found employment program studies limited by small samples, poor description of participants, and few randomized or controlled designs.¹³

The most studied program is Dandelion, a three-year program run by DXC Technology with Australian government departments. It places autistic adults in technology roles such as software testing and cyber security, with a support worker in each team.^11,14 A 2017 university report found that 30 of 38 people who started in government placements were still in the program; of those who left, some moved to other jobs or study.¹¹ A 2023 cost-benefit analysis, a modeled estimate for one program with no comparison group, of 56 participants, 86% of them men, found a high benefit ratio for government, and most participants moved into open employment afterwards; without a comparison group, how many would have done so anyway is unknown.¹⁴ DXC Technology and the Australian government departments responsible for defense and human services are listed as funders of that analysis.¹⁴

A 2016 Deutsche Bank internship in the UK for 8 autistic graduates, which arose from a partnership with the charity Autistica and was devised by staff within the bank, was followed by researchers through interviews before and after. Five interns had their contracts extended; some reported anxiety, difficulty judging communication, and confusion about office rules.⁸ One, who felt embarrassed to be on a scheme for autistic people, said: "I just want to be seen as an employee, not as the autistic employee".⁸

For other well-known programs the record is thinner. A 2008 case narrative in the journal Innovations, written by Specialisterne's founder Thorkil Sonne with Jonathan Wareham, describes the Danish firm that started this kind of hiring.¹⁵ No peer-reviewed outcome evaluation of SAP's or Microsoft's programs was found as of October 2026. A 2021 conceptual paper proposes a framework in which firms gain from hiring autistic people when top management backs it and the program brings the firm visibility and social approval.¹⁶

Programs are often sold on an "autism advantage", the idea that autistic employees bring superior skills such as attention to detail. A 2020 systematic review found two quantitative and four qualitative studies, and no strong evidence either for or against a workplace autism advantage.¹⁷ Its listed funders included DXC Technology, Australian government departments and ANZ Bank.¹⁷ In a survey of 59 Australian employers of autistic adults, employers reported benefits without additional costs.¹⁸ The review's authors had earlier warned that stressing stereotypes such as attention to detail can overlook differences between autistic people and downplay their support needs.¹⁹

Research on automated screening and autism is sparse. In a 2024 audit, GPT-4 was asked to rank a CV against the same CV plus four autism-related items (an award, a scholarship, a panel and a membership). It ranked the autism version first in none of 10 trials; a custom version instructed in principles of diversity, equity, inclusion and disability justice did so in 3 of 10. Across the six disability conditions tested, GPT-4 ranked the enhanced CV first in 15 of 60 trials.²⁰ With 10 trials per condition and one model version, the audit says little about other screening tools. A 2022 ethics paper argues that algorithmic fairness methods in AI recruitment have limits for disability, because disabilities vary so widely.²¹ No study found as of October 2026 has tested commercial AI scoring of video interviews with autistic candidates.

Autistic writers describe the standard interview as a poor test of the work. Megan Anna Neff writes that it "frequently depends on spontaneity and rapid processing", that a flat affect "might be misconstrued as disinterest", and that interviews "compel you to grapple with the dilemma of masking versus unmasking".^c1 Her suggestions include a scripted pause ("That's an insightful question; let me gather my thoughts for a moment") and looking between the interviewer's eyes.^c1 Daniel M. Jones writes that autistic candidates "often respond too honestly and literally": asked "Do you have flaws?", he would answer with his worst one.^c2 Eric Garcia writes that traditional interviews "often require unspoken rules and trick questions".^c6

Pete Wharmby, an autistic writer and former teacher, describes interviews that weigh "soft social skills like eye contact, small talk, bravado/confidence and a good, appropriate sense of humour" over competence. After his diagnosis he applied for a secondment to his school's senior leadership team, joked that he did not want to end up "a weird little hermit", and was turned down partly for being "too self-deprecating". His suggestions include a remote interview or a visit to the room beforehand, the questions in advance, and honest job adverts, since "We autistics take language seriously".^c9 Carly Jones, an autistic advocate, sums up the gap: "Our knowledge usually outweighs our confidence."^c10

Writers differ on how far to bend. Megan Griffith calls an interview "a negotiation", tells readers to ask the interviewer plenty of questions, and writes "please feel free to lie", since employers fudge salary and policy too.^c4 Devon Price suggests asking to see the questions in advance, and reminds readers that non-autistic candidates "round up" their experience, "so remind yourself that it's only fair for you to do so, too."^c5 Price cites the US study above as showing autistic applicants do better in "written interviews"; in that study the written versions were transcripts of spoken interviews.^c5,1 Daniel M. Jones found his way in through a work agency, arguing that agencies care more about filling posts than about credentials and that "Experience is greater than qualifications".^c3

Asking has costs too. Daniel Burford asked a recruiter for the questions in advance; after days of silence he received "Two vague sentences" three hours before the interview, and felt he had "asked for a thing at the moment when I most needed to look easy to hire". He still advises asking, but in writing, so that there is "a record, and a date".^c11 Luke Beardon, an autism educator, questions whether interviews predict job performance at all, writing that "the interview process can be so terrifying for autistic people that it severely decreases their chances of finding gainful employment"; at an employment forum he hosted, the autistic participants all said the fairest test would be to let a person do the job.^c12 In another book he tells, as a story drawn from real cases, of Idriss, who faltered as six panelists took turns asking questions and then answered all 18 well when one person asked them.^c13 Georgina, in a first-person account in a third Beardon book, is told by a training program to be assertive and hold eye contact, and sees the catch: "if I then get the job I will be expected to continue the charade every time I'm at work."^c14

On hiring programs, Garcia reports both the alternatives and the criticism. SAP has used a monthlong process in which candidates build Lego robots from instructions, and lets candidates choose formats such as walking interviews.^c6 Julia Bascom of the Autistic Self Advocacy Network told Garcia that some programs silo autistic staff into de facto sheltered workshops and avoid people with higher support needs, and ASAN withdrew from talks with Specialisterne in 2018; Specialisterne's North American head told Garcia the company had since progressed. John Marble described program hires as either "crashing and burning" or "growing and leaving".^c7 Price calls such hiring "a highly conditional form of acceptance", reserved for the apparently "high functioning" and those "willing to define their lives by their productivity".^c8