AI for Speech Therapy: The Best AI Tools for SLPs in 2026

AI for speech therapy means using artificial intelligence to help with three jobs: making therapy materials, writing documentation, and giving clients extra practice between sessions. In 2026 the good tools genuinely save prep and paperwork time, but none of them replaces clinical judgment — and most materials generators never check whether their words actually contain the target sound, so every output still needs an SLP’s review.

This guide is the map of that landscape: which AI tools SLPs actually use, what each one is good for, what it costs, and where the risks sit. It links out to our deeper guides on safety, ethics, and accuracy, so you can start here and go as deep as you need.

Full disclosure before we start: Ga-loo is our product. It appears in the materials-generation section below, and we explain exactly where it is strong and where another tool is the better choice. Every price here was checked on official pricing pages or app store listings (July 2026; SpeechLP in September 2026) and can change; always confirm before you buy.

AI for speech therapy at a glance

How we compared these tools

We grouped tools by the job they actually do, because “AI for speech therapy” covers three very different jobs — and general chatbots cut across all of them:

• Making therapy materials — worksheets, word lists, stories, pictures.

• Documentation — session notes, SOAP notes, progress summaries.

• Client-facing practice — apps a child or adult uses between sessions.

For each tool we looked at the official feature list, current public pricing, the free tier or trial, and one question most roundups skip: does anything check the output for accuracy before you put it in front of a student?

Best AI tools for speech therapy: price comparison

Which AI tools make speech therapy materials?

Ga-loo — phonetically checked articulation worksheets (our tool)

Ga-loo does one narrow job: it generates printable articulation worksheets from your parameters — target sound, word position, age, theme — and then checks every word against the CMU Pronouncing Dictionary before you see the result. If you ask for initial /r/, a deterministic validator counts the /r/ occurrences, flags competing sounds, and shows you the report. You approve or regenerate; nothing goes to print without your review.

That validation step is the reason Ga-loo exists, and as far as we can tell it is still unique — see the section on accuracy below. The trade-off is scope: Ga-loo will not write your IEP goals or plan a literacy unit. It makes articulation worksheets, quickly and verifiably. You can read how the generator works or try it on a free trial without a card. Pricing: credit packs from $9 for 10 worksheets, or a $15/month subscription.

MagicSchool AI — for the school side of the job

MagicSchool is built for teachers, not SLPs, but school-based clinicians use its tools for the paperwork around therapy: parent emails, accommodation language, leveled texts. The permanent free plan is generous; Plus is $12.99/month (or $8.33/month billed annually). For therapy materials themselves it has the same limitation as any general model: no speech-specific checking.

Canva Magic Media — pictures, not word lists

Canva’s AI image generator is handy for custom scenes — a “what’s wrong with this picture” card, a barrier-game board. Qualifying educators get Canva Pro free through Canva for Education. Treat the output as clip art, not clinical content: an AI image of “a rabbit and a robot” supports your /r/ session, but no tool at Canva knows why you asked.

Can you use ChatGPT for speech therapy?

General-purpose chatbots are the most common way SLPs use AI today, and for good reason: they are cheap (both have free tiers; ChatGPT Plus and Claude Pro are each $20/month), and they will draft a story about dinosaurs loaded with /s/ words in seconds.

The catch is that nothing verifies the draft. Large language models work with text, not sound, and they routinely include words that only look right. Ask for “initial K words” and you may get “knee” and “know” — spelled with k, pronounced without it. Practitioner resources that recommend these tools say the same thing: review every word before it reaches a student, and keep client information out of the chat, because a consumer chatbot is not a HIPAA or FERPA environment. We wrote more about this in Can You Trust AI-Generated Speech Therapy Materials?

We tested that failure mode in detail — with documented examples and a practical checking workflow — in ChatGPT vs. a phonetically validated generator.

Which AI tools help with SLP documentation?

Documentation assistants are a different product category: they listen to (or read about) a session and draft the note. The AI risk profile is different too — the concern is accuracy of the record and privacy of the recording, not phonetics. Three tools come up repeatedly among SLPs:

• SLPFlow ($20/month, Professional $49/month, 14-day trial without a card) generates SOAP, DAP, BIRP and IEP progress notes from recordings or written summaries, and is built specifically for SLPs.

• Ambiki is a full practice-management platform for SLPs, OTs and PTs (plans from $25/month per user); its Tenalog add-on generates session notes at $1 per session and can describe articulation errors from your input.

• SPRY is a clinic-grade EMR with an AI scribe trained on SLP terminology, from $150/month per provider — sized for clinics, not individual school SLPs.

All three position the AI draft as exactly that: a draft the clinician reviews and signs. That is the right model, and it is worth noticing that the documentation market already treats human review as non-negotiable — a standard the materials-generation market has not caught up to.

AI speech therapy apps for kids

This is the newest and least proven category: apps a child uses at home, with speech recognition that listens to productions and reacts to them.

Better Speech’s “Jessica” markets itself as an AI speech therapy helper that clients practice with between sessions, using speech recognition to react to productions of sounds like /r/ and /s/. It is deliberately positioned as a supplement to a human SLP, and a registered clinical trial comparing hybrid AI-plus-therapist care with traditional therapy is still under way — so treat outcome claims as pending, not proven. Pricing is not public.

SpeechLP is an articulation game app for children aged 3–9 whose “AI listening” gives sound-by-sound feedback during play. It is free with a limited number of daily levels; paid plans on the App Store run from $15 to $59 a month.

Sara, short for Speech Articulation Reinforcement App, came out of Columbia’s 2025 startup challenge. It targets children aged 4–12, detects errors such as a distorted /r/, and gives personalized feedback.

Speech recognition for children’s speech is genuinely hard, and outcome claims in app store listings are marketing until published trials of the app itself appear. The sensible use today is supervised carryover: the SLP sets the target, checks that the app judges productions the way a clinician would, and treats the app’s scores as practice data, not assessment.

The accuracy gap no tool likes to talk about

Here is the pattern across every materials generator we reviewed: none of the general-purpose tools discloses any methodology for verifying that generated words actually contain the target sound. The tools that are careful about quality — the documentation assistants — solve it with mandatory human review of a draft. The materials tools mostly leave verification to you, silently.

ASHA’s guidance on generative AI says clinicians should independently substantiate claims and that AI cannot replace clinical judgment. In practice that means one of two workflows:

• Generate with a general tool, then check every word yourself against pronunciation — which gives back some of the time the AI saved you.

• Use a tool where verification is built in and visible, and spend your review time on clinical fit instead of phonetics.

The second workflow is the one Ga-loo was built for: the validator is deterministic (dictionary lookup, not another AI guessing), the report is shown to you, and the worksheet still requires your explicit approval. We think that combination — machine checking plus clinician sign-off — is where all of these tools should end up.

If you want the clinical reasoning behind this — why a single off-target word matters when practice trials are the budget — see why phonetic validation matters.

Is AI safe for speech therapy?

Mostly yes, when it is used for support rather than authority. ASHA is explicit that “AI cannot replace the audiologist, speech-language pathologist, or assistant,” and that clinicians must evaluate any technology they use (ASHA). In practice, safe use comes down to four questions, each covered in its own guide:

• Is the task low-risk? Brainstorming and generic materials are; diagnosis and unreviewed clinical decisions are not. See AI for speech therapy: what’s safe, what’s not.

• Do clients need to know? When AI is part of their care, ASHA’s Code points toward informed consent. See ASHA ethics and AI: informed consent, explained.

• Can you trust the output? A fluent worksheet is not proof of accuracy. See our five-part test in Can you trust AI-generated speech therapy materials?

• Is the target actually checked? Spelling is not sound. See why phonetic validation matters and ChatGPT vs. a phonetically validated generator.

What about student privacy and HIPAA?

Accuracy is only half of the responsibility question; the other half is privacy. The three tool categories handle it very differently, and it is worth being deliberate:

• Consumer chatbots (ChatGPT, Claude, free tiers especially) are not HIPAA or FERPA environments. The safe workflow is strict: describe the task (“a story for a 6-year-old working on initial /s/”), never the student.

• Documentation tools live or die by compliance, and the serious ones say so plainly — SLPFlow, Ambiki and SPRY all advertise HIPAA compliance, and clinic-grade tools will sign a BAA. Verify the paperwork before you record a session.

• Materials generators should not need protected information at all. Ga-loo is built zero-PII by design: a worksheet is defined by sound, position, age band and theme — there is deliberately no field for a student’s name.

The fine print matters here. Both big chatbot makers sign a Business Associate Agreement only for enterprise and API offerings: Anthropic for Claude Enterprise and its API, not for Claude Free, Pro or Max (Claude Help Center); OpenAI for its API and sales-managed ChatGPT Enterprise or Edu accounts, not for ChatGPT Business or consumer plans (Paubox). An individual SLP on a $20 plan should assume there is no BAA.

A good rule of thumb for any AI tool: if the workflow tempts you to type a child’s name into it, the workflow is wrong, whatever the privacy policy says.

Which AI tool should you pick?

• You need articulation worksheets for a specific target → Ga-loo; it is the only tool here that verifies the phonetics for you.

• You need ideas, stories, or goal language drafts → ChatGPT or Claude free tiers, with manual checking and no client data.

• You drown in school paperwork → MagicSchool AI.

• Your notes eat your evenings → SLPFlow for a lightweight start; Ambiki or SPRY if you need whole-practice management.

• You want home practice between sessions → a speech-recognition app such as SpeechLP, set up and monitored by you.

• You want a static materials library instead of generation → that is a different category; see our guide to SLP Now alternatives.

If you spot something outdated, tell us at support@ga-loo.com and we will fix it.

Frequently asked questions

What is the best AI tool for SLPs?

There is no single best tool, because the jobs differ. For articulation materials, pick a generator that verifies the target sound (Ga-loo is built for that). For notes, a documentation assistant made for SLPs, such as SLPFlow or Ambiki. For open-ended drafting, a general chatbot like ChatGPT or Claude, used without client data.

Can AI replace speech therapists?

No. ASHA states that AI “cannot replace the audiologist, speech-language pathologist, or assistant” (ASHA). AI can draft materials and notes and support home practice, but assessment, diagnosis, goal setting, and the clinical decisions around them remain the SLP’s job.

Is ChatGPT good for speech therapy?

For brainstorming, parent letters, and story drafts, yes. For articulation word lists it is unreliable, because it works from spelling, not sound: words like “knee” end up on an initial /k/ list. Check every word aloud, and never type client information into a consumer plan. More in ChatGPT vs. a phonetically validated generator.

Are there free AI tools for speech therapy?

Yes. ChatGPT and Claude have free tiers, MagicSchool AI has a permanent free plan, Canva Pro is free for qualifying educators, SpeechLP has a free tier, and Ga-loo offers a free trial without a card. Documentation tools usually offer trials rather than free plans.

Is AI HIPAA compliant for SLPs?

It depends on the plan, not the brand. Consumer chatbot plans are not covered by a BAA; enterprise and API offerings can be. Documentation tools built for clinicians — SLPFlow, Ambiki, SPRY — advertise HIPAA compliance, but confirm the BAA before recording a session. The safest option is a tool that needs no protected information at all.

Are there AI speech therapy apps for kids?

Yes — apps such as SpeechLP, Sara, and Better Speech’s “Jessica” use speech recognition to give feedback on a child’s productions. They are best treated as supervised home practice set up by an SLP, not as therapy on their own, because independent evidence for individual apps is still thin.

Free, verified word lists that need no AI at all

A large share of articulation work runs on a plain word list. These are free, checked by pronunciation, and organized by sound and position.

• All articulation word lists by sound and position.

• Initial R sentences, initial S sentences and initial TH sentences for sentence-level drill.

• ST blend words, STR blend words and final ST words for cluster targets.

Sources and further reading

• ASHA — Generative Artificial Intelligence (AI) for Clinicians

• Claude Help Center — Business Associate Agreements for commercial customers

• Paubox — How do I get a BAA with OpenAI?

• SpeechLP — App Store listing

• Columbia Magazine — Can AI fill a gap in childhood speech therapy?

Keep reading

• Best apps and digital tools for SLPs 15 tools sorted by the job they do, with prices.

• SLP Toolkit alternatives Free and paid options for the data side, including AbleSpace.