If you've spent time on a waitlist for speech therapy, you already know the shortage is real. In some parts of the country, families wait six months or more for an initial evaluation. Researchers at Stanford and other institutions are now exploring whether AI can close that gap.

The diagnostic applications are still developing. Language models aren't yet reliable for diagnosing speech disorders on their own, but fine-tuning techniques are showing enough promise that Stanford researchers described a model capable of assisting SLPs with diagnostic tasks as something that "will potentially allow them to help more children." That's the goal: not to replace the speech-language pathologist, but to make it possible for one SLP to reach more families without shortchanging any of them.

Where AI is already making a difference

For the SLPs who are currently working with your child, the most immediate change isn't happening in the therapy room. It's happening after hours. Speech therapists, like most healthcare providers, spend a significant chunk of their time on documentation, progress notes, and session planning. When those tasks take two hours, that's two hours not spent preparing something meaningful for the next child.

Some clinicians have started using generative AI to move through that work faster. A 2025 Baltimore Banner report found SLPs using AI to draft session notes, generate home practice materials, and create visual supports tailored to individual children. Every clinician quoted was clear: AI doesn't replace their judgment. But it does give them time back, and time in speech therapy translates directly into what your child gets in each session.

A different kind of technology: letting children see their own speech

One development that's showing real results doesn't involve generative AI at all, but it reflects the same shift toward giving children better feedback about their own communication. Researchers at Montclair State University, funded by the NIH, are studying visual biofeedback software with more than 100 children ages 8 to 18. The software shows children a real-time visual display of how their speech compares to a target sound, so they can see and hear when they're hitting it.

For a child who has practiced the same sound hundreds of times without knowing exactly what "right" feels like, that kind of immediate visual feedback changes the experience of learning. Early results suggest children correct speech sounds faster when they have that information. It's a reminder that the most useful technology for kids in therapy isn't always the most complex.

What stays human

It's worth being clear about what AI doesn't change, because some parents worry that technology creeping into therapy means less of the relationship that makes therapy work. It doesn't. The things that most determine your child's progress still require a trained SLP who knows them:

• Diagnosing and classifying the specific speech or language disorder
• Reading a child's emotional state in real time and adjusting accordingly
• Building the therapeutic relationship that makes a child willing to try hard things
• Advocating for services and documenting needs in language insurers will accept

AI may support some of those tasks eventually. For now, if AI reduces the administrative burden on your child's SLP, your child benefits. More of each session goes to them. The work happening in research labs right now is aimed at making that happen at scale, for the families who are still waiting.