The answer depends on which part of the rising numbers you look at. Autism prevalence in the U.S. has climbed from roughly 1 in 1,667 children in 1994 to 1 in 31 children in 2022, according to CDC surveillance data. Some of that increase clearly reflects better detection of people who were always autistic but went unrecognized. Some of it reflects shifting diagnostic boundaries. And some of it, according to a growing number of clinicians, does reflect genuine overdiagnosis in ambiguous cases.
Why the Numbers Keep Rising
No single explanation accounts for the increase. Researchers point to at least seven overlapping factors: broader diagnostic criteria, greater public awareness, reduced stigma, diagnostic substitution (relabeling conditions that used to get a different name), the linking of specific services to an autism diagnosis, improved detection of milder presentations, and actual overdiagnosis. These factors interact with each other, making it difficult to isolate how much each one contributes.
One well-documented piece of the puzzle is diagnostic substitution. Between 1994 and 2003, autism prevalence in U.S. special education rose from 0.6 to 3.1 per 1,000 children. During the same period, diagnoses of intellectual disability and learning disabilities dropped by 2.8 and 8.3 per 1,000, respectively. That decline represented a sharp break from the trends of the previous decade, and it tracked closely with the autism increase. In other words, many children who would have been labeled with intellectual disability or a learning disability in the 1980s started receiving autism diagnoses instead.
What Clinicians Themselves Say
In a 2021 survey published in BMC Psychiatry, more than half of physicians (52.8%) said they believed that in over 10% of clinical assessments, an autism diagnosis was given despite an inconclusive evaluation. The conditions most commonly confused with autism included ADHD (cited by 68.4% of respondents), other mental health conditions (25.8%), and cerebral palsy or similar motor conditions (8.6%). Clinicians in the survey described autism and ADHD as the two neurodevelopmental conditions they considered most frequently overdiagnosed.
The reasons they gave were practical, not conspiratorial. Autism shares symptoms with several other conditions, particularly ADHD, social anxiety, and language disorders. When a child’s presentation is ambiguous, clinicians sometimes lean toward an autism diagnosis because it opens doors to services that other diagnoses do not. More than a third of physicians in the survey said that access to benefits was a key reason colleagues diagnosed autism in borderline cases. In Israel, where the study was conducted, children with autism qualify for 14 hours of weekly support services from infancy through age seven, while children with other developmental conditions largely do not. Similar dynamics exist in the United States, where an autism diagnosis can unlock school-based therapies, insurance coverage for behavioral interventions, and other supports that may not be available under a different label.
How Screening Tools Perform
The most widely used screening tool for toddlers, the M-CHAT (in its original and revised forms), catches about 83% of children who are autistic and correctly identifies about 94% of children who are not. That means roughly 17% of autistic toddlers are missed by the screener, and about 6% of non-autistic toddlers are flagged for further evaluation. A positive screen is not a diagnosis. It triggers a more thorough clinical assessment. But in a population where autism affects roughly 3% of children, even a small false-positive rate can mean that a meaningful number of families go through diagnostic evaluations that ultimately lead nowhere, or in some cases, lead to a diagnosis that doesn’t quite fit.
These numbers also reveal the opposite problem. Missing 17% of autistic children at the screening stage means tens of thousands of kids each year may not get flagged early enough for the interventions that work best in the toddler years.
Underdiagnosis Is Still a Major Problem
While overdiagnosis gets the headlines, underdiagnosis remains the bigger issue for several groups. The median age of autism diagnosis in 2024 was 6 years old, down only slightly from 7 years in 2015. For boys, the median dropped more noticeably, from 7 to 5. For girls, it barely moved, hovering near 8 years across the entire decade. That gap matters because early intervention is most effective during the preschool years.
Racial and ethnic disparities have shifted in recent years but haven’t disappeared. For the first time in CDC tracking history, autism prevalence among white children (24.3 per 1,000) is now lower than among Black (29.3), Hispanic (31.6), and Asian or Pacific Islander (33.4) children. This reversal likely reflects improved outreach and detection in communities that were historically underserved. But Black children with autism remain significantly more likely to also be classified as having an intellectual disability (50.8%) compared to Hispanic (34.9%) and white children (31.8%), suggesting that milder presentations in Black children may still go unrecognized while more severe cases are the ones getting diagnosed.
Girls with autism are diagnosed 3.8 times less often than boys. When they are diagnosed, they’re more likely to have a co-occurring intellectual disability (42.1% versus 36.9% for boys), which suggests that many girls without intellectual disability are flying under the radar entirely. Adult women represent a particularly large detection gap, with initial diagnoses in adulthood increasing among women in recent years as awareness of how autism presents differently across genders has grown.
The Role of Schools and Services
In the U.S., a medical autism diagnosis and a special education autism classification are technically separate processes. A child can be clinically diagnosed with autism but not qualify for special education if the condition doesn’t measurably affect their learning. Conversely, a school can classify a child under the autism category for educational services without that child having a formal medical diagnosis.
Children identified through the healthcare system receive their diagnosis significantly earlier. The median age of identification for children evaluated through health sources is 44 months, compared to 69 months for children identified only through schools. That 25-month gap isn’t just about schools using a general “developmental delay” label as a placeholder. Children in education-only settings were seen for any kind of evaluation later than those seen in healthcare settings, pointing to a broader access issue rather than a labeling preference.
Both Things Can Be True
The evidence points in two directions at once. In clinical settings where ambiguous cases are common and an autism label unlocks better services, some children are receiving diagnoses that don’t clearly fit the criteria. At the same time, girls, minorities, and adults without intellectual disability continue to be missed at high rates. The system simultaneously overdiagnoses some populations and underdiagnoses others.
One proposed solution, raised by clinicians in the BMC Psychiatry survey, is to decouple services from specific diagnoses. If support services were tied to a child’s functional needs rather than to a particular diagnostic label, the incentive to stretch a borderline case into an autism diagnosis would largely disappear. Children with social communication difficulties, sensory challenges, or executive function problems could receive targeted help regardless of whether their condition technically meets autism criteria. Until that shift happens, the diagnosis will continue to serve as both a clinical description and a key to accessing support, and those two roles will keep pulling the numbers in different directions.