If you are a school-based speech-language pathologist looking for AI that fits your week, six tools in our directory earn their place: Glint for communication boards and visual schedules, Diffit for leveling text to a language goal, FrenalyticsEDU for goal progress data, Read&Write for carryover across the school day, Kami for read-aloud on any worksheet, and MagicSchool for drafting the prose around a report. All six work on the paperwork. None of them should touch the eligibility call.
That last line stopped being a philosophical preference this month. On September 2, 2026, the DC Office of the State Superintendent of Education published its LEA AI Model Policy for Staff Use (OSSE), a red, yellow and green system in which, as reported, AI use for determining individualized education programs is prohibited outright, alongside grading, discipline and teacher evaluation. In July, K-12 Dive reported that four more states, Idaho, Maryland, Oklahoma and Virginia, now require districts to adopt AI policies, with Oklahoma's law barring AI as the primary basis for grading, discipline or other high-stakes decisions (K-12 Dive). You are the staff member those clauses land on hardest, because eligibility, present levels and goal-writing are your job.
How this list was built
The pool was the 28 tools in the directory's special education and accessibility category, plus tools from other categories that do a job an SLP recognizes. I judged them on fit for a caseload rather than a classroom, what the vendor documents about student data, honest pricing, and whether the tool leaves the clinical decision with the clinician. Vendors pay nothing for inclusion and cannot buy a place on this list.
Four tools I looked at hard and left out, which tells you more than the six I kept. TeachScribe captures observations by voice, which sounds ideal for session notes, but our record shows FERPA and COPPA both declared as No, it is scoped to early years, and it is hosted in the UK. Classroom Companion listens continuously to teacher-student interactions and sorts them into praise, warnings and support, which is the always-on classroom capture the new DC policy treats as prohibited surveillance. Vocina runs spoken assessment well, but its record states it shares data with third parties and sets a 13-plus minimum, which rules out most elementary caseloads. EchoGuide Pro looked promising for spoken practice and then published no price and made no clinical claim.
Pricing and privacy details verified September 10, 2026.
The finding nobody selling you software will mention
Search our directory of 126 tools for "therapy" and you get zero results. Search it for "speech" and you get four, none built to deliver or measure speech therapy. Glint comes closest, and its own listing calls it a tool for special education and speech pathology, but what it makes is materials, not sessions and not data.
So the honest answer to "what is the best AI tool for SLPs" is that there is no such category yet. What exists is a set of general education tools, a few of which do one slice of your job well. Every page currently ranking for this query is published by a company selling one of those slices.
The rest of the numbers are worth knowing before you request anything. Of the 28 tools in the special education category, 14 declare FERPA compliance and 12 declare COPPA. Eleven host data in the US, seven in the UK, four in the EU, and five do not say where the data sits at all. Only 5 of the 28 carry an independent security assessment. Those are vendor declarations, not audits, and a blank means our record is empty, not that a tool has failed anything.
Comparison
| Tool | Best for | Pricing | Grades | Min age | Privacy (declared) | Data hosted | Independent security assessment |
|---|---|---|---|---|---|---|---|
| Glint | Communication boards | Freemium | All | 18+ | FERPA, COPPA, GDPR | US | No |
| Diffit | Leveling text | Freemium, $14.99/mo | Not stated | Not stated | Vendor page: FERPA, COPPA | Not disclosed | No |
| FrenalyticsEDU | Goal progress data | Free plan, then quote | K-12, post-sec | No minimum | FERPA, COPPA | US | No |
| Read&Write | Carryover support | Free for K-12 teachers | Not stated | Not stated | Vendor page: FERPA, COPPA | Not disclosed | Yes |
| Kami | Read-aloud worksheets | Freemium, $149/yr | All | No minimum | FERPA, COPPA, GDPR | US | Yes |
| MagicSchool | Drafting report prose | Free, Plus $8.33/mo | Not stated | Not stated | Vendor page: FERPA | US processing | Yes |
The privacy column reports what each vendor declares, not the result of an audit. The last column is the one independent check in the table.
Best for communication boards and visual supports: Glint
Glint is the best fit for building AAC boards, visual schedules and social stories because it is the only tool in our directory whose vendor writes for speech-language pathologists directly.
AAC, augmentative and alternative communication, covers anything a student uses to communicate instead of or alongside speech, from a laminated core board to a speech-generating app. Glint takes a prompt and returns a board, a schedule or a step-by-step social narrative, with natural voice output across around 50 languages, and the vendor's site carries a section headed for therapists and SLPs (Glint). What sets it apart from building the same materials in Canva is the structure: you describe a routine and get a board back rather than laying out squares yourself.
Watch out for: the account minimum is 18-plus, so this is a tool you use to make things, not one a student logs into. Neither the directory record nor the vendor site sets out how the symbol set was chosen or validated, which matters if your district has a core vocabulary standard.
Best for producing visual supports at volume. Not for delivering therapy or collecting session data.
Pricing: a free tier and a free trial; the vendor publishes no price for paid plans, so treat those as quote-based (checked September 10, 2026, Glint).
Safety: our record shows the vendor declaring FERPA, COPPA and GDPR compliance, data hosted in the US, no student data collected, a minimum age of 18-plus, and no independent security assessment. Because the account is adult-only and no student data goes in, the FERPA question here is mostly about what you type into a prompt. Your district still decides whether it wants a signed data agreement first.
Best for leveling text to a language goal: Diffit
Diffit is the best fit for turning any passage into the reading level a language goal actually targets, because it rewrites the same content at several levels and builds the vocabulary and question set with it.
Paste a URL, a PDF, a video link or your own text, choose a level, and you get an adapted passage plus vocabulary, summary and questions, exportable to Docs, Slides or PDF. For a student working on inferencing at a second-grade level inside a fifth-grade science unit, that is the prep problem solved in a minute. What sets it apart from MagicSchool's text leveler is scope: MagicSchool returns the leveled passage, Diffit returns the passage plus the scaffolds around it.
Watch out for: the free plan keeps only content you created in the last 90 days (Diffit), so anything you want next year needs exporting or paying for. Our directory record for Diffit is blank on privacy, which sent me to the vendor.
Best for language and comprehension goals inside general education content. Not for anything involving a student's record.
Pricing: Basic is free with 90-day retention; an individual teacher subscription is $14.99 a month or $149.99 a year; school pricing is a flat annual rate tiered by enrollment and quoted on request (checked September 10, 2026, Diffit pricing).
Safety: Diffit's privacy page states it is FERPA and COPPA compliant, that it signs the National Data Privacy Agreement and the Student Data Privacy Consortium's statewide agreements, and that it collects no student data (Diffit privacy). It does not disclose which country hosts the data. Our directory record is empty on all of this and should be updated.
Best for goal progress data: FrenalyticsEDU
FrenalyticsEDU is the best fit for the progress-monitoring half of your caseload because it is built around IEP and MTSS goals rather than around lessons.
MTSS, multi-tiered systems of support, is the tiered intervention structure most US districts run. Frenalytics gives each student a profile, records skill-based sessions against goals, and produces the progress reports that otherwise eat your Friday. Tracking twelve articulation goals and eight language goals is a data problem, and this is a data tool. What sets it apart from tracking the same goals in a spreadsheet is that the reports come out formatted for an IEP meeting rather than for you.
Watch out for: it collects a lot. Our record lists name, date of birth, learning and performance data, usage analytics and location data, a longer list than any other tool here, which makes this a district decision rather than a personal one. It also declares GDPR compliance as No, which matters for international schools.
Best for SLPs and case managers who own progress reporting. Not for a solo clinician who wants something light.
Pricing: a free plan at $0 covers 10 student profiles, one staff account and one session per student per day; paid EDU plans are listed from $1,499, with enterprise pricing quoted on request and a two-week trial. The pricing page does not make the billing period plain, so confirm it before you budget (checked September 10, 2026, Frenalytics pricing).
Safety: the vendor declares FERPA and COPPA compliance, hosts in the US, collects student data, offers deletion on request, sets no minimum age, and holds no independent security assessment. Because student names and dates of birth go in, this one needs a signed district data agreement before a single profile is created.
Best for carryover across the school day: Read&Write
Read&Write is the best fit for making a language goal survive outside your room, because it puts text-to-speech, word prediction and vocabulary support on every document a student opens, and it is free for K-12 teachers.
Everway, formerly Texthelp, gives teachers full premium access at no cost (Everway pricing). For a student whose goal is expressive vocabulary, word prediction in the middle of a history assignment delivers more carryover practice than a weekly session can. What sets it apart from Kami is reach: Kami works on documents you put inside Kami, Read&Write follows the student across Windows, Mac, Chrome, iOS and Android.
Watch out for: it supports the student rather than measuring them, so it gives you no goal data. The trust page does not say where data is hosted, and our directory record is blank on FERPA and COPPA.
Best for carryover, access and independence between sessions. Not for progress monitoring.
Pricing: free for K-12 teachers; district licenses at $2.75 per seat, group licenses at $18.30 per seat with a 150-seat minimum, and an individual license at $184 a year (checked September 10, 2026, Everway pricing).
Safety: Everway's trust page displays FERPA and COPPA badges and ISO/IEC 27001 certification, and offers a data protection addendum for download (Everway trust). It states no hosting region. Read&Write is one of only five tools in this category carrying an independent security assessment, which is the strongest procurement signal on the list.
Best for read-aloud on any worksheet: Kami
Kami is the best fit for making the materials you already use accessible in the moment, because it turns any PDF into something a student can hear, annotate and answer in, and it collects no student data doing it.
Drop in a worksheet and a student can have it read aloud, record a spoken answer instead of writing one, or annotate it. For a student on a receptive language goal who cannot get through the directions, that is the barrier removed without a referral. What sets it apart from Read&Write is the spoken response: Kami lets a student answer by voice inside the document, which is a useful artifact for an expressive goal.
Watch out for: the free tier gives you around ten tools and no LMS integration (Kami pricing), so the version that fits a caseload workflow is the paid one. It also lives in the browser, so it does not follow a student into a paper packet.
Best for accessible materials and spoken responses. Not for anything that has to happen offline.
Pricing: a free Basic plan; a Teacher plan at $149; Pro at $12 a month or $85 a year; school and district plans quoted on request (checked September 10, 2026, Kami pricing).
Safety: the vendor declares FERPA, COPPA and GDPR compliance, hosts in the US, collects no student data, sets no minimum age, and holds an independent security assessment. That combination is rare in this category and it is why Kami is on the list.
Best for drafting the prose around a report: MagicSchool
MagicSchool is the best fit for the writing that surrounds an evaluation, the parent letter, the plain-language summary, the meeting agenda, because it drafts fast and its privacy policy says student data is never used to train models.
The caution is that MagicSchool also ships an IEP generator, and that is the feature the new state rules are about. Use it to phrase what you have already decided. Do not use it to decide. What sets it apart from ChatGPT for this job is that commitment in writing: MagicSchool's policy states it will not use student data to train, fine-tune or improve AI models (MagicSchool privacy). Most general chatbots on a personal account make no such promise.
Watch out for: the free plan caps monthly generations, and the pull toward pasting a student's present levels into a prompt is strong. Under the DC model policy and Oklahoma's statute, an AI-determined IEP is the thing explicitly ruled out.
Best for the writing around the decision. Not for the decision, the eligibility call, or anything with a student's file in it.
Pricing: free forever with limited monthly generations; Plus at $8.33 per user per month billed annually, or $12.99 monthly; enterprise quoted by volume (checked September 10, 2026, MagicSchool pricing).
Safety: the privacy policy acknowledges FERPA obligations, states that student data is not used to train models, and describes the service as designed for educators, with students reaching it only through a school agreement. Hosting is described as processing in the US rather than a named region. MagicSchool carries an independent security assessment.
The big names, briefly
ChatGPT is capable and, on a personal account, has no district data agreement behind it, so it belongs to your own writing and never to a student's file. NotebookLM is the best of the general tools for reading a stack of evaluation reports and asking questions of them, because it answers only from what you upload. Khanmigo is a student tutor built for academic content and will not serve a communication goal. Quizlet drills vocabulary and nothing more: useful for word retrieval, irrelevant to articulation.
How to actually choose
Four questions, in this order, will settle it faster than any feature comparison.
Does a student's name have to go in? If yes, check whether your district has a signed data agreement with that vendor before anything else. Of the six here, only FrenalyticsEDU needs student records to function. The rest work on materials.
Is this the doing or the deciding? Writing a social story is doing. Judging whether a student meets eligibility criteria is deciding. Outsource the first, never the second. That line is now written into state policy, and it was the right line before it was the law.
Who signs? A tool a clinician adopts alone is a tool the district has not vetted. If the answer to the first question was yes, the answer here has to be someone above you.
What happens when the student transfers? Ask whether you can export the goal data and whether the vendor deletes on request. Twenty-one of the 28 tools in this category offer deletion on request; seven do not say.
If a tool passes all four, pilot it with one student for a marking period before it touches your caseload. Speech and language therapists in the UK and Australia face the same four questions, with the ICO and GDPR in place of FERPA.
What I would do this week
Pick one student whose visual supports are out of date. Rebuild their schedule and one social story in Glint, which needs no student data and no procurement conversation, and time yourself. Then take the worksheet that student struggles with most and run it through Diffit at their instructional level. Two tools, one student, one afternoon, and you will know within a week whether either belongs in your rotation. Leave FrenalyticsEDU alone until someone in the district office has read its data agreement.
The bottom line
No AI tool has been built for school-based speech-language pathology yet, so the practical move is to borrow six good general tools for the parts of your week that are paperwork and keep every clinical judgment where it belongs. Glint and Diffit save the most time this month with the least procurement friction. Kami and Read&Write help the student between sessions. FrenalyticsEDU is a district decision, not a personal one. MagicSchool writes around the report and never inside it.
ASHA notes that while median caseloads have held near 48 students, the workload wrapped around them has grown dramatically (ASHA). That growth is paperwork, and paperwork is what a machine should be doing. Your clinical judgment is the part nobody can automate, and now, in a growing number of states, the part nobody is allowed to. Compare the rest of the category in the special education and accessibility directory, or browse all 126 tools.
Your caseload is waiting. Spend the saved hour on them.
Dan Fitzpatrick is a Forbes contributor, three-time bestselling author and founder of The AI Educator, and has trained more than 150,000 educators across 30 countries. More from Dan.
