For speech-language pathologists
How should SLPs evaluate speech therapy apps, AI and other technology?
Published · Updated
The short answer
Judge speech therapy technology the way you would any treatment. Ask for published evidence in children like yours, accuracy on children’s error productions rather than overall, and a clear account of what it records and keeps. Then check the privacy law that applies (in the US, HIPAA and a business associate agreement in health care, FERPA in schools, COPPA for children under 13), accessibility, cost, and whether it fits the treatment plan. ASHA says efficacy claims for AI tools must be backed by evidence.
Evidence: tested on whom, and by whom?
ASHA defines evidence-based practice as bringing together clinical expertise, the best available evidence, and the client’s and caregivers’ perspectives (ASHA). Its guidance on evaluating products asks clinicians to use a product with populations for which there are accuracy data, ideally reflecting the racial, ethnic and linguistic diversity of the people they serve. Without validity data, you need to be able to justify using it and note the limits (ASHA). For AI tools, ASHA says “any claims about the efficacy of an AI tool must be substantiated with evidence” (ASHA).
- Is there peer-reviewed research showing benefit to clients? ASHA calls replicated, peer-reviewed research the most credible, and says claims should rest on more than anecdotes and client reviews (ASHA).
- Who took part: what ages, diagnoses, target sounds, languages and dialects? Do they look like your caseload?
- Who ran the research? ASHA calls researching the developer’s background an important step (ASHA).
- What does the wider evidence say? A review of 132 children’s speech therapy apps rated none as excellent quality (Furlong et al., 2018), and in a small 2026 study, four children aged 4 to 6 who practiced alone with a speech app didn’t become more accurate on their target sounds (Strömbergsson et al., 2026).
- Watch for ASHA’s warning signs: claims that sound too good to be true, outcomes described in vague “feel-good” terms, and dismissing findings that question the product (ASHA).
For what’s being studied in AI and wearables, see AI and wearables in speech therapy. For teletherapy evidence, see online speech therapy for kids.
Accuracy with children’s speech
An overall accuracy figure can hide the number that matters. In a systematic review of 32 studies and 18 automated speech analysis tools, tools met an 80% agreement threshold with human judges for intelligibility, severity or error category, but accuracy was typically below 80% on words containing mispronunciations (McKechnie et al., 2018). Speech recognition is also significantly less accurate for children than for adults, partly because there’s far less child speech data to train on (Shivakumar & Georgiou, 2020).
- Is accuracy on error productions reported separately from accuracy on correct ones?
- Was it compared with several SLP raters, and how well did those raters agree with each other?
- Was it tested with children of the ages, dialects and error types you treat?
- Were the figures from lab recordings or real therapy sessions?
- What does it do with productions it’s unsure about, and can you override it?
A 2025 tutorial gives clinicians a framework for appraising clinical AI for validity, reliability, and ethical and equitable use (Benway & Preston, 2025). More in automated speech scoring for SLPs.
Privacy in health care: HIPAA and business associate agreements
In the US, if you or your practice is a HIPAA covered entity, a vendor that creates, receives, maintains or transmits protected health information (PHI) for you is generally a business associate, and HIPAA requires a business associate agreement (BAA) before you share PHI with it. HHS’s examples include a cloud service that stores electronic PHI, an app vendor that provides transcription services, and a third-party AI chatbot on a patient portal (HHS).
The BAA has to describe how the vendor may use and disclose PHI and bar other uses. The Security Rule also requires the vendor to report security incidents, including breaches, and the vendor needs BAAs with any subcontractors that handle the PHI (HHS).
- Ask: will the vendor sign a BAA, and does it cover every service that touches client data, including outside AI or transcription services?
- Teletherapy platforms. ASHA recommends video with end-to-end encryption confirmed by the clinician, a BAA, and secure storage, and notes that everyday social media and video call apps often lack adequate security (ASHA). HHS’s relaxed pandemic-era enforcement for telehealth is over: its transition period expired on August 9, 2023 (HHS).
- Consumer apps and devices outside HIPAA. The FTC says makers of health apps and connected devices that aren’t covered by HIPAA must follow its Health Breach Notification Rule, and that a breach includes sharing health data without the person’s authorization, not only hacking (FTC, 2024).
- Recordings. Does it record audio or video? Does it keep recordings or only scores, where, for how long, and are they used to train its AI? ASHA says you should know what personal data a product collects and stores so you can inform clients (ASHA).
In schools: FERPA, and COPPA for children under 13
In most cases, HIPAA’s Privacy Rule doesn’t apply to an elementary or secondary school. Student health records at a FERPA-covered school, including those kept by a contracted provider acting for the school, are education records under FERPA instead. Even where HIPAA doesn’t apply, FERPA and IDEA require written parental consent before a school shares a student’s information with Medicaid for billing (US Department of Education [ED] & HHS, 2019).
- Vendors and the school official exception. A school can share personal information from education records with an online service only under conditions including that the provider is under the school’s direct control and uses the records only for authorized purposes (ED, 2014).
- Free tools and click-through terms. Clicking “accept” on terms of service creates a contract, and ED recommends that free online services go through the same approval process as paid ones, without staff bypassing it (ED, 2014).
- Retention. ED recommends contracts include data destruction requirements so student information doesn’t stay on the provider’s systems after the contract ends. Metadata stripped of identifiers may be used for other purposes unless the contract forbids it (ED, 2014).
- COPPA. For online services covered by COPPA, an audio file of a child’s voice is personal information. Operators generally need verifiable parental consent, a written retention policy, and may keep children’s data only as long as reasonably necessary (FTC, 2025). A school can consent on parents’ behalf only when data is collected for the school’s use and benefit and for no other commercial purpose, and it can ask for the data to be deleted (FTC COPPA FAQ).
More on COPPA in children’s voice data privacy. This is general information, not legal advice: check with your district, employer or a lawyer for your situation.
Accessibility, cost and fit with the treatment plan
- Accessibility. In the US, ADA Title II covers public schools and other state and local government entities (ADA.gov). A 2024 rule requires the web content and mobile apps they provide, including through contracts, to meet WCAG 2.1 Level AA; a 2026 rule moved the deadlines to April 26, 2027 for entities with a population of 50,000 or more and April 26, 2028 for smaller ones (28 CFR 35.200; ADA.gov). For teletherapy, ASHA points to real-time captioning for clients who are deaf or hard of hearing and to ADA accommodations (ASHA).
- Cost. ASHA suggests asking whether a paid version really offers better data security or features than a free one, whether you need it at all, and regularly asking whether a tool is necessary, beneficial or simply optional for your setting (ASHA).
- Your competence. ASHA’s Code of Ethics says certificate holders may offer only services they’re competent to provide, and should use a procedure only when benefit can reasonably be expected (ASHA).
- Fit with the plan. ASHA says AI can’t replace the SLP and encourages individualized treatment plans (ASHA). Can you set the targets, the amount of practice and the feedback schedule? See feedback frequency in motor learning and speech sound therapy dosage.
- Data you can use. Does it show you what the child actually practiced between sessions? See tracking practice between sessions.
Disclosure
Saylo, which publishes this page, is a speech-practice wearable in development. It hasn’t been tested with children yet, and we’ll publish validation results before making any claims about accuracy or outcomes.
Questions SLPs ask
Do I need a BAA with a speech therapy app?
If you’re a HIPAA covered entity and the app creates, receives, maintains or transmits PHI for you, it’s generally a business associate and you need a BAA. Check with your compliance team or a lawyer about your situation.
Does HIPAA apply to school-based SLPs?
Usually not to student records. In most cases, health records at a FERPA-covered school are education records under FERPA, so FERPA’s rules on consent and vendors apply instead.
Can a school consent to an app collecting a child’s voice under COPPA?
Only in limited cases. The FTC says a school can consent for parents when data is collected for the school’s use and benefit and for no other commercial purpose.
What accuracy figure should I look for?
There’s no single threshold. Ask for agreement with several SLP raters on children’s error productions specifically, in children like the ones you treat, because overall figures can hide weak performance on mispronounced words.
Sources
- ASHA: Evidence-based practice (EBP).
- ASHA: Evaluating procedures, products, or programs: Considerations for audiologists and speech-language pathologists.
- ASHA: Generative artificial intelligence for clinicians.
- Furlong, L., Morris, M., Serry, T., & Erickson, S. (2018). Mobile apps for treatment of speech disorders in children: An evidence-based analysis of quality and efficacy. PLOS ONE, 13(8), e0201513.
- Strömbergsson, S., Edlund, E., Pettersson, M., Phan, N., & Kurimo, M. (2026). Self-supervised app-based speech training for children with speech sound disorder: A single-case experimental design study. International Journal of Language & Communication Disorders, 61(1), e70163.
- McKechnie, J., et al. (2018). Automated speech analysis tools for children’s speech production: A systematic literature review. International Journal of Speech-Language Pathology.
- Shivakumar, P. G., & Georgiou, P. (2020). Transfer learning from adult to children for speech recognition: Evaluation, analysis and recommendations. Computer Speech & Language, 63, 101077.
- Benway, N. R., & Preston, J. L. (2025). Equipping speech-language clinicians for the critical appraisal of an artificial intelligence-driven, evidence-based future. Language, Speech, and Hearing Services in Schools, 56(3), 442–468.
- US Department of Health and Human Services, Office for Civil Rights: Business associates (content last reviewed July 30, 2026).
- ASHA Practice Portal: Telepractice.
- US Department of Health and Human Services, Office for Civil Rights: HIPAA and telehealth.
- Federal Trade Commission (2024). Complying with FTC’s Health Breach Notification Rule.
- US Department of Education & US Department of Health and Human Services (2019). Joint guidance on the application of the Family Educational Rights and Privacy Act (FERPA) and the Health Insurance Portability and Accountability Act of 1996 (HIPAA) to student health records (December 2019 update).
- US Department of Education, Privacy Technical Assistance Center (2014). Protecting student privacy while using online educational services: Requirements and best practices.
- Federal Trade Commission (2025). Children’s Online Privacy Protection Rule: Final rule. Federal Register 90(76), 16918–16983 (April 22, 2025).
- Federal Trade Commission: Complying with COPPA: Frequently asked questions.
- US Department of Justice, ADA.gov: Fact sheet: New rule on the accessibility of web content and mobile apps provided by state and local governments.
- 28 CFR 35.200: Requirements for web and mobile accessibility (as amended April 20, 2026).
- ASHA: Artificial intelligence (AI): Additional resources and guidance.