US surveillance tells the headline story: CDC's ADDM network estimated autism in 1 in 150 eight-year-olds in 2000, 1 in 36 in 2020, and 1 in 31 (3.2%) in the 2022 data published in April 2025 — with older era estimates running around 1 in 2,000-10,000 under the narrow definitions of the 1960s-80s. Global figures cluster near 1% with wide variation. Three features of the surveillance matter before any interpretation: ADDM counts documented identification in health and education records, not a fixed biological measure; prevalence in 2022 ranged nearly four-fold across US sites — from about 1.5% in Laredo, Texas to 5.3% in California — under the same methodology; and groups historically underdiagnosed (girls, Black and Hispanic children, higher-IQ children, adults) account for disproportionate shares of the growth. Each is a fingerprint of ascertainment, not biology.

The rise decomposes into well-documented mechanisms. Definition: autism widened from Kanner's narrow syndrome through the additions of PDD-NOS and Asperger syndrome (1980s-90s) to DSM-5's merged spectrum (2013) — each revision recruiting milder presentations the earlier criteria excluded. Detection: universal screening recommendations, earlier evaluation, service-linked incentives for documentation, and collapsing stigma all raise identification without changing children. Substitution: diagnostic migration is directly measurable — as autism diagnoses rose, intellectual-disability and language-disorder classifications fell correspondingly in special-education data, the same children relabelled. Studies tracking autism traits rather than diagnoses — including large Swedish twin cohorts — find the underlying phenotype roughly stable across decades while registered diagnoses multiplied: the strongest single piece of evidence that the epidemic is substantially an epidemic of counting.

Honest analysis leaves room for some genuine increase the counting factors do not fully absorb. Established contributors with real but modest effects: advancing parental age (both parents), improved survival of extremely preterm infants, and possibly perinatal factors under study. What the evidence excludes: vaccines — the original MMR claim rested on a fraudulent, retracted study whose author lost his licence, and has since been contradicted by cohort studies covering millions of children across countries, including specifically in high-risk siblings; thimerosal's removal from childhood vaccines changed nothing in autism trends. Politically prominent claims of single environmental causes (most recently around paracetamol/acetaminophen in pregnancy, where major sibling-controlled studies find no causal signal) have repeatedly outrun the evidence. Genetics remains the dominant known etiology, with heritability around 80% in twin and family studies.

The April 2025 ADDM release landed in a charged environment: US health leadership framed the numbers as a preventable epidemic driven by environmental toxins and announced investigations promising rapid answers, while the CDC's own report attributed the rise substantially to identification differences — the four-fold site variation tracking service availability being hard to explain any other way. Points of genuine substance inside the political noise: severity is not vanishing into mild cases (a large share of ADDM-identified children have co-occurring intellectual disability, and profound autism deserves dedicated research and services); prevalence gradients by race have inverted as access equalised, ending the era when autism was diagnosed mainly in white and affluent families; and demands for etiologic research are legitimate even when their favoured hypotheses are not. The failure mode on both sides is the same: treating a surveillance artefact question as an identity or ideology question.

Whatever the decomposition, the service reality is unambiguous: millions more identified people need what identification promises — early intervention with demonstrated benefit, school supports, adult services that mostly do not exist, and for the profoundly affected, lifelong care whose burden falls on families. The adult-diagnosis wave — people recognising themselves in expanded criteria after lifetimes of unexplained difficulty — brings genuine relief and access alongside real questions about where clinical thresholds sit when self-identification and brief online assessments enter the pathway. And the neurodiversity reframing — autism as variation to be accommodated rather than solely deficit to be treated — has improved lives and public understanding while sitting in unresolved tension with the realities of profound autism; serious advocacy now spans both, and pretending either half of the spectrum represents the whole misleads.

For parents: rising prevalence numbers say nothing about your child's individual risk — the figures track counting more than biology; act on development, not headlines: if concerns exist, seek evaluation early, because intervention works best young and waiting lists are long. Vaccinate on schedule — that question is closed by evidence at the scale of millions. For adults wondering about themselves: formal assessment through qualified clinicians beats online tests, and a diagnosis is worth pursuing when it would change support, understanding or accommodations. For readers of the political fight: demand that epidemic claims engage the site-variation and trait-stability evidence, and that reassurance claims engage profound autism's unmet needs — the test both sides routinely fail.