Autism Spectrum Disorder and Early Detection: The Emerging Role of AI-Assisted Diagnostic Aids

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by differences in social communication, social interaction, and patterns of restricted or repetitive behaviors and interests. Research has consistently shown that many signs of autism can be recognized during the first few years of life, allowing opportunities for early identification and intervention. Despite growing awareness and improvements in screening practices, many children continue to experience critical delays between the first appearance of developmental concerns and the completion of a formal diagnostic evaluation.

The consequences of delayed diagnosis extend far beyond the timing of a clinical label. Early identification is critical because the developing brain demonstrates substantial neuroplasticity during infancy and early childhood. Timely diagnosis can facilitate access to evidence based interventions, including speech and language therapy, occupational therapy, behavioral support services, educational accommodations, and family centered resources. When diagnosis is delayed, children may miss important developmental opportunities that could improve communication, adaptive functioning, social participation, and long term outcomes. Families may also experience prolonged uncertainty, increased stress, and difficulty navigating healthcare and educational systems while searching for answers.

Several factors contribute to the persistent diagnostic bottleneck in autism assessment. One of the most significant challenges is the remarkable heterogeneity of autism spectrum disorder itself. Autism is not a single presentation but rather a spectrum encompassing a wide range of developmental profiles, strengths, and support needs. Some young children display clear and recognizable features, including delayed language development, reduced eye contact, limited joint attention, atypical social reciprocity, repetitive movements, sensory sensitivities, and restricted interests. In these cases, concerns may emerge early and prompt referral for specialist evaluation.

However, many children present with subtler characteristics that are more difficult to recognize. Some demonstrate age appropriate language development while exhibiting nuanced social communication differences that may not become apparent until social demands increase. Others develop compensatory strategies that mask underlying challenges, particularly in structured environments. These variations can delay recognition by parents, educators, and healthcare professionals.

Diagnostic disparities are particularly evident among certain populations. Girls with autism are frequently diagnosed later than boys because their symptoms may differ from traditional diagnostic expectations. Many girls display stronger social imitation skills, fewer overt repetitive behaviors, or greater ability to camouflage social difficulties. Similarly, children with higher cognitive abilities or stronger verbal skills may not meet conventional assumptions about autism presentation, leading to delayed referral and assessment. Children from racial, ethnic, cultural, and socioeconomically disadvantaged groups also face disproportionate barriers to diagnosis due to differences in healthcare access, cultural perceptions of development, provider bias, and systemic inequities.

Healthcare infrastructure limitations further contribute to delays. In many regions, there is a shortage of developmental behavioral pediatricians, child psychologists, child psychiatrists, speech language pathologists, and multidisciplinary autism diagnostic teams. Demand for comprehensive evaluations often exceeds available capacity, resulting in lengthy waiting lists that can extend for months or even years. Rural and geographically underserved communities face additional challenges, as families may need to travel long distances to access specialized services. Financial barriers, inadequate insurance coverage, language differences, and limited health literacy can further complicate the diagnostic pathway.

Against this backdrop, artificial intelligence has emerged as a promising tool to enhance autism screening and assessment. The rationale for applying AI in autism detection is compelling because diagnosis relies heavily on identifying complex behavioral patterns across multiple domains of development. Machine learning algorithms excel at analyzing large volumes of multidimensional data and identifying subtle relationships that may not be readily apparent through traditional assessment methods alone.

Recent advances have enabled AI systems to evaluate a variety of behavioral and developmental indicators associated with autism. These include patterns of visual attention and gaze tracking, facial expressions and affect recognition, motor movements, vocal characteristics, speech patterns, caregiver reported developmental concerns, and structured clinical observations. By analyzing these data streams, AI models can identify patterns that may be associated with autism risk and generate objective measures to support clinical evaluation.

One area of particular interest involves automated analysis of eye tracking and social attention. Research has shown that differences in visual attention to faces, social interactions, and environmental stimuli may emerge early in development. AI driven systems can quantify these patterns with a level of precision that would be difficult to achieve through observation alone. Similarly, advances in computer vision technology allow for detailed assessment of facial expressions, gesture use, and social engagement behaviors during naturalistic interactions.

Speech and language analysis represents another promising application. Machine learning algorithms can examine vocal features such as pitch variation, speech rhythm, prosody, and language use patterns. These characteristics may provide valuable insights into communication development and potentially contribute to earlier identification of children who would benefit from comprehensive assessment.

Importantly, AI technologies may also help address some of the healthcare access challenges that currently contribute to delayed diagnosis. Digital screening platforms, smartphone based assessments, telehealth integrated tools, and automated behavioral analysis systems could increase access to preliminary evaluations in underserved communities. Such approaches may facilitate earlier referral of high risk children while reducing demands on limited specialist resources.

Despite these promising developments, the role of artificial intelligence in autism diagnosis must be approached with appropriate caution. Autism is a complex neurodevelopmental condition that exists within diverse social, cultural, and developmental contexts. Diagnostic assessment requires careful consideration of developmental history, family perspectives, clinical observations, cognitive functioning, language abilities, and coexisting medical or psychiatric conditions. These elements require professional judgment that extends beyond algorithmic analysis.

AI systems are therefore best viewed as decision support tools rather than replacements for clinical expertise. They can improve consistency, enhance efficiency, identify patterns, and assist in risk stratification. However, they cannot independently establish a diagnosis, interpret developmental nuances, or replace the comprehensive evaluation performed by trained healthcare professionals. Concerns related to algorithm bias, data quality, generalizability across diverse populations, privacy, and ethical implementation must also be carefully addressed before widespread adoption.

As the field continues to evolve, the integration of artificial intelligence into autism assessment has the potential to transform early detection and improve access to care. The greatest value of these technologies will likely come from their ability to complement rather than replace clinicians, supporting earlier recognition, more efficient referral pathways, and more consistent assessment processes. By combining advanced computational tools with expert clinical judgment, healthcare systems may be better positioned to reduce diagnostic delays and ensure that children and families receive timely access to the services and support they need.

Current Clinical Standard for Early Identification

The early identification of Autism Spectrum Disorder (ASD) is a critical component of pediatric and developmental healthcare. Timely recognition allows children and families to access appropriate interventions, educational support, and developmental services during periods of greatest neuroplasticity, when therapeutic interventions are most likely to improve long term outcomes. Because autism presents with a wide range of behavioral, social, communicative, and developmental features, detection requires a systematic and longitudinal approach that integrates developmental surveillance, standardized screening, comprehensive clinical assessment, and consideration of alternative or co-occurring conditions.

Developmental surveillance serves as the foundation of autism detection and should be incorporated into every well-child visit throughout early childhood. Surveillance is an ongoing process rather than a single assessment, recognizing that developmental trajectories evolve over time and that autism-related characteristics may emerge gradually or become more noticeable as social and communication demands increase. During routine encounters, clinicians should actively inquire about caregiver concerns regarding development, behavior, communication, social interaction, and learning. Parents and caregivers are often the first to recognize subtle developmental differences, making their observations a valuable component of early detection.

In addition to eliciting caregiver concerns, clinicians should directly observe the child’s social communication behaviors and developmental functioning. Key areas of assessment include eye contact, response to name, use of gestures, joint attention, social reciprocity, emotional engagement, and the ability to initiate and maintain interactions with others. Developmental surveillance should also include a review of language acquisition, motor development, adaptive functioning, and play behaviors. Particular attention should be paid to imaginative and symbolic play, as well as the child’s ability to share interests and experiences with others. Clinicians should also ask specifically about developmental regression or developmental plateau, especially the loss of previously acquired language, social, or adaptive skills, as these may represent important early indicators of autism or other neurodevelopmental disorders.

While developmental surveillance is essential, it should be complemented by standardized screening tools that improve the likelihood of identifying children who may require further evaluation. The American Academy of Pediatrics recommends universal autism-specific screening at both 18 and 24 months of age, in addition to ongoing developmental surveillance at all health supervision visits. Screening is particularly important because some children with autism may not exhibit obvious symptoms during brief clinical encounters, and structured tools can help identify concerns that might otherwise be overlooked.

Among available screening instruments, the Modified Checklist for Autism in Toddlers, Revised with Follow-Up (M-CHAT-R/F) remains one of the most widely used and extensively validated tools in primary care settings. The questionnaire is designed to identify children at increased risk for autism by assessing behaviors related to social communication, attention, and interaction. The follow-up interview component helps reduce false-positive results and improves the tool’s predictive value. However, clinicians should recognize that screening instruments are designed to identify risk rather than establish a diagnosis.

A positive autism screening result indicates the need for further developmental assessment and diagnostic evaluation rather than confirming the presence of Autism Spectrum Disorder. Similarly, a negative screening result should not automatically exclude the possibility of autism. Continued developmental concerns expressed by caregivers, clinical observations suggestive of social communication difficulties, developmental delays, or evidence of regression warrant further investigation regardless of screening outcomes. Clinical judgment remains essential when interpreting screening results and determining the need for referral.

The diagnosis of Autism Spectrum Disorder remains fundamentally clinical and requires a comprehensive, multidisciplinary assessment whenever possible. Diagnostic evaluation involves gathering detailed developmental and medical histories, obtaining caregiver reports, conducting direct behavioral observation, and evaluating the child’s functioning across multiple settings and developmental domains. Diagnostic decisions are guided by the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR), which require persistent deficits in social communication and social interaction, along with restricted, repetitive patterns of behavior, interests, or activities that result in clinically significant functional impairment.

A thorough developmental history should explore prenatal and perinatal factors, early developmental milestones, language acquisition, social engagement, adaptive functioning, educational performance, family history, and the timing and progression of symptoms. Direct observation allows clinicians to assess social reciprocity, communication patterns, sensory behaviors, repetitive movements, and responses to environmental stimuli. The impact of symptoms on daily functioning, academic performance, family life, and peer relationships should also be carefully evaluated.

Several standardized diagnostic instruments can support the assessment process. The Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), is widely regarded as one of the most valuable tools for structured behavioral observation. Similarly, the Autism Diagnostic Interview-Revised (ADI-R) provides a detailed, standardized caregiver interview that explores developmental history and autism-related behaviors. While these instruments contribute valuable information, they should not be viewed as standalone diagnostic tests. No single assessment tool can independently confirm or exclude autism. Rather, diagnostic conclusions should be based on the integration of multiple sources of clinical information within the broader context of a comprehensive evaluation.

An essential aspect of autism assessment involves consideration of alternative explanations for developmental and behavioral symptoms. Many conditions can mimic, overlap with, or coexist alongside autism, making differential diagnosis a critical component of clinical decision-making. Hearing impairment should always be considered, particularly in children presenting with delayed language development or reduced responsiveness to social cues. Formal audiologic evaluation is often necessary to exclude hearing loss as a contributing factor.

Other neurodevelopmental and psychiatric conditions may also present with features similar to autism. Language disorders can affect communication skills without the broader social communication deficits characteristic of autism. Intellectual disability and global developmental delay may result in developmental impairments that require careful differentiation from autism-related social deficits. Attention-deficit/hyperactivity disorder frequently co-occurs with autism and may contribute to social difficulties, impulsivity, and behavioral challenges. Anxiety disorders can affect social functioning and communication, while trauma-related symptoms may produce behaviors that resemble social withdrawal or emotional dysregulation.

Medical conditions should also be considered during the diagnostic process. Sleep disorders, epilepsy and other seizure disorders, genetic syndromes, metabolic disorders, and neurological conditions can influence developmental functioning and behavior. Genetic conditions such as Fragile X syndrome and other chromosomal abnormalities may present with autism-like features and warrant consideration when clinically indicated. Importantly, the presence of one diagnosis does not exclude the possibility of autism. Many children and adults with Autism Spectrum Disorder have co-occurring developmental, psychiatric, or medical conditions that require concurrent identification and management.

Ultimately, accurate autism detection and diagnosis require a comprehensive, patient-centered approach that combines developmental surveillance, standardized screening, detailed clinical assessment, and thoughtful differential diagnosis. Early recognition remains crucial for optimizing developmental outcomes, but diagnostic accuracy is equally important to ensure that children receive appropriate interventions tailored to their specific needs. By integrating caregiver perspectives, clinical expertise, validated assessment tools, and a thorough evaluation of co-occurring conditions, healthcare professionals can provide timely and accurate diagnoses that support effective treatment planning and long-term developmental success.

Why AI Is Being Studied

AI-assisted tools are being studied because the current diagnostic system is slow, specialist-dependent, and unevenly available. A well-designed tool could help clinicians prioritize referrals, standardize parts of assessment, and identify children who need developmental support sooner.

AI may be useful when it measures behaviors that clinicians can observe but cannot easily quantify during a brief visit. Eye-tracking can measure social visual engagement. Computer vision can quantify gaze, gestures, facial expression, repetitive movement, motor patterns, and response to social bids. Natural language processing can analyze prosody, rhythm, vocabulary, turn-taking, and pragmatic language.

Machine-learning systems can also combine several inputs. For example, a model may incorporate caregiver questionnaires, clinician observations, and structured home videos. This type of multimodal assessment may provide more useful information than a single behavioral measure.

Still, performance in a research dataset is not the same as performance in routine practice. Enriched samples, narrow age ranges, controlled video quality, specialty-clinic populations, and limited demographic diversity can inflate accuracy. Clinicians should look for prospective validation, external validation, prespecified thresholds, transparent reporting of indeterminate results, and performance across sex, race, ethnicity, language, and developmental level.

FDA-Authorized and FDA-Cleared Diagnostic Aids

Only a limited number of AI-assisted or technology-assisted tools have FDA authorization or clearance as pediatric autism diagnostic aids. This distinction is important because many developmental apps or behavioral platforms have not undergone FDA review for diagnostic use.

Canvas Dx, formerly the Cognoa ASD Diagnosis Aid, is an FDA-authorized prescription diagnostic aid for children 18 through 72 months who are at risk for developmental delay based on parent, caregiver, or clinician concern. It combines caregiver input, clinician input, and video-based behavioral information. It is intended to aid diagnosis, not to function as a stand-alone diagnostic replacement.

EarliPoint is an FDA-cleared eye-tracking-based diagnostic aid. It measures a child’s visual response to social information and is intended to aid qualified clinicians in the diagnosis and assessment of autism within its cleared age range and intended-use population.

Clinicians should verify the current device labeling before use. Software version, age range, intended setting, contraindications, warnings, and performance data may change over time.

Table 1. Regulatory Categories for AI-Assisted ASD Tools

Category	Clinical interpretation
FDA-authorized or FDA-cleared diagnostic aid	May be used within its labeled indication as an adjunct to clinician assessment.
Research tool	Promising but not ready for routine diagnostic use unless validated and authorized for that purpose.
Wellness or educational app	May help families organize observations, but diagnostic claims require caution.
AI risk score	Should support triage or decision-making, not replace clinical diagnosis.
Unclear regulatory status	Do not rely on diagnostic claims until status and validation are confirmed.

Evidence Base for AI-Assisted Detection

The evidence base is growing, but it remains uneven. Some AI-based studies report high sensitivity, specificity, or classification accuracy. Many studies, however, are limited by small samples, retrospective designs, referral-enriched populations, or lack of external validation.

This matters because autism screening and diagnosis are highly dependent on pretest probability. A tool that performs well in a specialty clinic may perform differently in primary care. A tool trained on one demographic group may perform less well in another. A tool validated in a carefully controlled study may perform differently when videos are recorded at home with variable lighting, background noise, and caregiver prompting.

Eye-tracking approaches are among the better-developed technology-assisted strategies. Social visual engagement differs in many young children with autism, and eye-tracking can quantify these patterns objectively. In clinical use, however, gaze behavior can be influenced by attention, fatigue, cooperation, developmental level, visual impairment, and testing context.

Speech and language AI tools are earlier in clinical translation. Acoustic and language features may differ in autism, including prosody, rhythm, pragmatic language, and conversational reciprocity. These tools may eventually support assessment, but they should not be described as reliable stand-alone diagnostic biomarkers in infants or toddlers without stronger evidence.

The most important evidence gap is not whether AI can classify patterns in datasets. It can. The harder question is whether AI improves real-world outcomes. Does it shorten time to diagnosis? Does it improve referral accuracy? Does it increase access for underserved families? Does it reduce disparities? Does it help children receive effective support sooner? These questions require longitudinal and implementation studies.

Where AI Fits in Clinical Workflow

AI should be placed inside a structured clinical pathway. It should not be used as a shortcut around clinical evaluation.

Table 2. Mobile-Friendly Clinical Workflow

Step	Clinician action
Surveillance	Ask about caregiver concerns, observe social communication, review milestones, and assess regression or plateau.
Standardized screening	Use validated screening tools at recommended ages and whenever concerns arise.
AI-assisted tool	Use only within the validated or labeled population. Treat output as adjunctive.
Clinical interpretation	Combine AI output with history, examination, hearing status, language level, adaptive function, and comorbidities.
Referral or diagnosis	Refer complex or uncertain cases. Diagnose only when the clinical evidence supports DSM-5-TR criteria.
Follow-up	Do not wait for diagnostic certainty before addressing developmental needs. Connect families to early intervention when indicated.

This workflow protects against a common error: treating a device result as the endpoint. The endpoint is not a score. The endpoint is an informed clinical plan.

A child with language delay, global developmental delay, or concerning social communication differences may need services even before an autism diagnosis is finalized. Early intervention should be based on developmental need, not only on diagnostic certainty.

Potential Benefits

The strongest potential benefit of AI-assisted detection is improved access. If validated tools help primary care clinicians identify children who need further evaluation, families may reach appropriate services sooner.

AI may also improve consistency. Human raters vary in training, experience, fatigue, and interpretation. A structured algorithm can apply the same measurement rules repeatedly. This may be useful for video review, eye-tracking metrics, and standardized caregiver questionnaires.

Another benefit is documentation. AI-generated measurements may support communication among pediatricians, developmental specialists, psychologists, speech-language pathologists, schools, and families. Better documentation may also help justify referrals and services.

These benefits are conditional. A tool that identifies risk but does not lead to evaluation or services may increase anxiety without improving care. Implementation must include referral pathways, family counseling, and access to developmental support.

Risks and Limitations

AI tools can be wrong. False-positive results may lead to anxiety, unnecessary referral, mislabeling, or delayed recognition of another developmental condition. False-negative results may falsely reassure clinicians and families, delaying needed evaluation. Indeterminate results can create confusion unless the workflow specifies the next step.

Algorithmic bias is a major concern. Autism has historically been underrecognized or diagnosed later in girls, children from minoritized racial and ethnic groups, children from lower-resource settings, and children whose primary language differs from that of the healthcare system. If AI tools are trained on nonrepresentative datasets, they may reproduce these inequities.

Data quality also matters. Video lighting, camera angle, background noise, caregiver prompting, language, child fatigue, and device access can all affect performance. A tool validated in a controlled study may perform differently in a community clinic or home-recorded setting.

Clinicians should also be cautious about black-box outputs. A probability score without explanation may be difficult to integrate into care. The most useful tools will provide interpretable outputs, clear limitations, and practical next steps.

Table 3. Safety and Equity Checks Before Acting on an AI Result

Issue	Practical response
Positive AI result	Confirm clinically. Evaluate developmental profile and differential diagnosis.
Negative AI result	Do not dismiss persistent caregiver or clinician concern. Repeat surveillance or refer.
Indeterminate result	Treat as unresolved risk, not reassurance. Arrange follow-up.
Age outside label	Do not extrapolate performance beyond the validated or cleared age range.
Language mismatch	Use interpreter support and culturally informed assessment.
Limited follow-up access	Do not screen without a plan for referral, counseling, and developmental services.
Possible bias	Monitor tool performance across sex, race, ethnicity, language, and insurance status.

Pharmacotherapy and Drug-Safety Considerations

AI detection tools do not treat autism. No medication is FDA-approved to treat the core social-communication features of autism spectrum disorder.

Pharmacotherapy may be appropriate for selected associated symptoms or co-occurring conditions. Risperidone and aripiprazole have FDA-labeled indications for irritability associated with autistic disorder in pediatric patients within specified age ranges. These medications may reduce aggression, severe tantrums, self-injury, and marked irritability in selected patients.

These drugs require careful monitoring. Clinicians should monitor weight, body mass index, fasting glucose or A1c, lipids, sedation, extrapyramidal symptoms, akathisia, prolactin-related effects when relevant, and overall risk-benefit balance. Medication should not replace behavioral, developmental, educational, environmental, or family-centered interventions.

The article should avoid implying that AI-based early detection directly enables pharmacogenomic medication selection in routine autism care. That remains investigational or context-specific and requires stronger evidence before being presented as a standard clinical pathway.

Table 4. Medication Clarification for ASD Articles

Topic	Clinically accurate wording
Core autism symptoms	No medication is FDA-approved to treat the core social-communication features of ASD.
Irritability	Risperidone and aripiprazole may be used for irritability associated with autistic disorder in labeled pediatric age groups.
Monitoring	Monitor metabolic effects, sedation, movement symptoms, and overall functional benefit.
AI relationship	AI detection does not determine medication need. Medication decisions require separate clinical assessment.
Pharmacogenomics	Do not present as routine ASD detection or treatment guidance without stronger evidence.

Ethical and Regulatory Considerations

AI-assisted autism detection raises ethical questions beyond accuracy. Families should understand what the tool does, what data it collects, how videos or behavioral data are stored, who can access the information, and whether data may be used for product improvement or research.

Pediatric data require special caution. Children cannot provide independent informed consent in the same way adults can. Parent or caregiver consent should be clear, specific, and understandable.

Clinicians also need clarity about responsibility. A device output does not remove the clinician’s duty to interpret the result, consider the differential diagnosis, communicate uncertainty, and arrange follow-up. Documentation should state that the AI result was used as an adjunct, not as an autonomous diagnosis.

Regulatory oversight is evolving. Software updates can change performance. Health systems adopting these tools should monitor real-world outcomes, false-positive and false-negative patterns, demographic performance, referral completion, time to services, and family experience.

Implementation in Practice

Before adopting an AI-assisted autism tool, a practice should answer several questions. Is the tool FDA-authorized or FDA-cleared for the intended use? What age range is covered? What population was used for validation? What happens after a positive, negative, or indeterminate result? Who explains the result to families? Is there a pathway to early intervention, audiology, speech-language evaluation, developmental pediatrics, psychology, or neurology when indicated?

Training is necessary. Clinicians need to know how to select appropriate patients, explain the tool, interpret outputs, and avoid overreliance. Staff need workflows for collecting videos or questionnaires, protecting data, and documenting results.

The most realistic model is hybrid care. Primary care practices can use validated tools to support triage and early identification. Specialists remain essential for complex presentations, diagnostic uncertainty, co-occurring neurologic or psychiatric symptoms, and children with substantial developmental complexity.

Table 5. Practice Adoption Checklist

Question	Why it matters
Is the tool FDA-authorized or FDA-cleared?	Confirms whether diagnostic claims have regulatory review.
Is the child within the labeled age range?	Avoids extrapolating beyond available evidence.
Was the tool validated in a similar population?	Helps assess generalizability.
Are false-positive and false-negative pathways defined?	Prevents confusion and delayed care.
Is follow-up available?	Screening has limited value without services.
Can results be explained clearly to families?	Reduces misunderstanding and anxiety.
Is performance monitored across groups?	Helps detect bias and inequity.

Readability and Communication With Families

Clinicians should avoid language that makes AI sound magical or definitive. Families need clarity, not hype. A better explanation is simple: “This tool may help us decide whether your child needs a more detailed autism evaluation. It does not replace clinical judgment, and we will interpret the result along with your concerns and your child’s developmental history.”

This wording does two useful things. It respects the technology, and it preserves clinical responsibility.

When a result is positive, clinicians should explain the next step. When a result is negative but concerns remain, clinicians should not close the case. When the result is indeterminate, families should be told that the result does not answer the question and that follow-up is still needed.

References

Abbas, H., Garberson, F., Glover, E., & Wall, D. P. (2018). Machine learning approach for early detection of autism by combining questionnaire and home video screening. Journal of the American Medical Informatics Association, 25(8), 1000-1007. doi:10.1093/jamia/ocy039

Abbas, H., Garberson, F., Liu-Mayo, S., Glover, E., & Wall, D. P. (2020). Multi-modular AI approach to streamline autism diagnosis in young children. Scientific Reports, 10, 5014. doi:10.1038/s41598-020-61213-w

American Academy of Pediatrics, Council on Children With Disabilities, Section on Developmental and Behavioral Pediatrics, Hyman, S. L., Levy, S. E., & Myers, S. M. (2020). Identification, evaluation, and management of children with autism spectrum disorder. Pediatrics, 145(1), e20193447. doi:10.1542/peds.2019-3447

American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.). American Psychiatric Association Publishing. doi:10.1176/appi.books.9780890425787

Bone, D., Bishop, S. L., Black, M. P., Goodwin, M. S., Lord, C., & Narayanan, S. S. (2016). Use of machine learning to improve autism screening and diagnostic instruments: Effectiveness, efficiency, and multi-instrument fusion. Journal of Child Psychology and Psychiatry, 57(8), 927-937. doi:10.1111/jcpp.12559

DailyMed. (Current label). Aripiprazole tablet prescribing information. National Library of Medicine. Retrieved June 2026.

DailyMed. (Current label). Risperidone tablet prescribing information. National Library of Medicine. Retrieved June 2026.

Food and Drug Administration. (2021). De Novo classification request for Cognoa ASD Diagnosis Aid: DEN200069. U.S. Food and Drug Administration.

Food and Drug Administration. (2023). 510(k) premarket notification K230337: EarliPoint. U.S. Food and Drug Administration.

Food and Drug Administration. (2025). 510(k) premarket notification K243891: EarliPoint System. U.S. Food and Drug Administration.

Jones, W., Klaiman, C., Richardson, S., et al. (2023). Development and replication of objective measurements of social visual engagement to aid in early diagnosis and assessment of autism. JAMA Network Open, 6(9), e2330145. doi:10.1001/jamanetworkopen.2023.30145

Jones, W., Klaiman, C., Richardson, S., et al. (2023). Eye-tracking-based measurement of social visual engagement compared with expert clinical diagnosis of autism. JAMA, 330(9), 854-865. doi:10.1001/jama.2023.13295

Lord, C., Rutter, M., DiLavore, P. C., Risi, S., Gotham, K., & Bishop, S. L. (2012). Autism Diagnostic Observation Schedule, Second Edition. Western Psychological Services.

Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism Diagnostic Interview-Revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24, 659-685. doi:10.1007/BF02172145

Megerian, J. T., Dey, S., Melmed, R. D., Coury, D. L., Lerner, M., Nicholls, C. J., Sohl, K., Rouhbakhsh, R., Narasimhan, A., Romain, J., Golla, S., Shareef, S., Ostrovsky, A., Shannon, J., Kraft, C., Liu-Mayo, S., Abbas, H., Gal-Szabo, D. E., Wall, D. P., & Taraman, S. (2022). Evaluation of an artificial intelligence-based medical device for diagnosis of autism spectrum disorder. npj Digital Medicine, 5, 57. doi:10.1038/s41746-022-00598-6

Robins, D. L., Casagrande, K., Barton, M. L., Chen, C. M. A., Dumont-Mathieu, T., & Fein, D. (2014). Validation of the Modified Checklist for Autism in Toddlers, Revised with Follow-Up. Pediatrics, 133(1), 37-45. doi:10.1542/peds.2013-1813

Shaw, K. A., Williams, S., Patrick, M. E., et al. (2025). Prevalence and early identification of autism spectrum disorder among children aged 4 and 8 years: Autism and Developmental Disabilities Monitoring Network, 16 sites, United States, 2022. MMWR Surveillance Summaries, 74(SS-2), 1-22. doi:10.15585/mmwr.ss7402a1

Siu, A. L., & U.S. Preventive Services Task Force. (2016). Screening for autism spectrum disorder in young children: U.S. Preventive Services Task Force recommendation statement. JAMA, 315(7), 691-696. doi:10.1001/jama.2016.0018

Tariq, Q., Daniels, J., Schwartz, J. N., Washington, P., Kalantarian, H., & Wall, D. P. (2018). Mobile detection of autism through machine learning on home video: A development and prospective validation study. PLOS Medicine, 15(11), e1002705. doi:10.1371/journal.pmed.1002705

Thabtah, F. (2019). Machine learning in autistic spectrum disorder behavioral research: A review and ways forward. Informatics for Health and Social Care, 44(3), 278-297. doi:10.1080/17538157.2017.1399132

Zuckerman, K. E., Lindly, O. J., & Sinche, B. K. (2015). Parental concerns, provider response, and timeliness of autism spectrum disorder diagnosis. Journal of Pediatrics, 166(6), 1431-1439.e1. doi:10.1016/j.jpeds.2015.03.007