Speech technology

AI and wearables in speech therapy: what’s real and what’s hype?

Published 21 October 2026 · Updated 21 October 2026

The short answer

AI and wearable tools for speech therapy are early-stage. The most promising research, AI-assisted practice for the R sound, has been tested in small groups of older children alongside a clinician, while a small study of young children practising with an app on their own found no improvement. Wearables that give real-time feedback have been studied mainly in adults. ASHA says AI can’t replace a speech-language pathologist (SLP) and that claims about AI tools need evidence.

What’s being studied

AI-assisted practice for R. In a feasibility study, 5 people aged 10 to 19 who could already make R did ten 40-minute sessions where a classifier judged their R during drill practice. Listeners heard improvement after 30 minutes of practice without a clinician, and all five improved from before to after treatment (Benway & Preston, 2024).
Speech apps without a therapist. Four Swedish children aged 4 to 6 practised with a speech app on their own for three weeks. None became more accurate on their target sounds; the authors concluded the app wasn’t sufficient in its current form (Strömbergsson et al., 2025).
Wearable feedback. Adults with healthy voices wore a voice monitor that gave feedback through the day. Feedback changed their behaviour while it was on, and a daily summary held up better afterwards than constant feedback (Van Stan et al., 2017).
Wearables for adults with Parkinson’s. In a study of 34 people, both a standard voice therapy and the SpeechVive wearable increased loudness compared with no treatment. The device’s inventor was an author and shareholder, as the paper discloses (Richardson et al., 2022).
Online (telepractice) therapy. A comparison of ASHA outcome data for children aged 6 to 9½ found no significant difference between in-person and telepractice speech sound therapy, and a review called the evidence for school-age children limited but promising (Coufal et al., 2018; Wales et al., 2017).

Why children’s speech is hard for AI

Speech recognition is still significantly less accurate for children than for adults, because children’s voices and pronunciation vary more and there’s far less child speech data to train on (Shivakumar & Georgiou, 2020). That matters most for exactly the sounds a child is still learning. More in automated speech scoring for SLPs.

How to spot hype

It claims to replace speech therapy or a therapist.
It says “clinically proven” without pointing to a published study.
It quotes an accuracy figure without saying who it was tested on: ages, dialects, and what kinds of speech errors.
Its evidence comes only from adults, or from the company that makes it.
It doesn’t say whether it records your child, or where recordings go.

ASHA’s guidance on AI says these tools can’t replace an SLP, that claims about them must be backed by evidence, and that regulatory oversight is limited (ASHA).

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.

Common questions

Can an app replace speech therapy?

No. In one small study, children who used a speech app on their own didn’t improve, and ASHA says AI tools can’t replace an SLP. Apps may help with practice an SLP sets up.

Is online speech therapy as good as in person?

For school-age children working on speech sounds, one large comparison found no significant difference in outcomes, and a review called the evidence limited but promising.

Are there speech wearables for kids?

The wearable research we found was in adults. Saylo is one children’s wearable in development, and it hasn’t been tested with children yet.