top of page
Search

Assistive Technology for Dyslexia: Checking the Claims

Writer: Andrea Chernin
Andrea Chernin
Oct 18
24 min read

Assistive Technology for Dyslexia: Checking the Claims


You make decisions about assistive technology for dyslexia more often than you make decisions about programs, and with far less to go on. A program comes with a manual, a dosage specification and, sometimes, a clearinghouse rating. A tool comes with a feature list, a marketing page and a free trial. The parent in front of you has already downloaded three of them.


So this is an audit of the claims rather than a product roundup: what the federal evidence record actually holds, what the single most-cited meta-analysis actually reported, what two vendors actually say in print, and what each accessible-format library actually requires before a student may use it.


The last of those turns out to be the part that will cost you an afternoon, and nobody writes about it.

The short version


  • The phrase names at least three different objects: a procurement and service category, a class of commercial products, and an accommodation decision. Evidence attaches to none of them as a whole.

  • "Assistive" appears zero times in all three National Center on Intensive Intervention tools charts, zero times in the federal reading intervention practice guide, and zero times in both federal writing practice guides.

  • The meta-analysis everyone cites reports three different numbers, and the only statistically significant moderator in it was whether the study used a between-subject or a within-subject design.

  • The product differentiator you will hear most often — human narration beats synthetic narration by about half — traces to a row in that meta-analysis which its own authors group with the non-significant moderators.

  • The specialised dyslexia font question has a clean published answer, and the answer is no.

  • Three accessible-format libraries require a professional to certify eligibility. One publishes a list of professions, one publishes a list of twenty-five job titles, and one defines no profession at all but asks the certifier for a licence number. "Educational therapist" appears on none of the lists.

Three different objects wear the same phrase


Start with the word, because the word is doing three jobs at once.


The National Center on Accessible Educational Materials, run by CAST, draws the distinction the field actually uses. Accessible educational materials are "print- and technology-based educational materials," including textbooks and core materials, "designed or enhanced in a way that makes them usable across the widest range of learner variability."


Assistive technologies, by contrast, "are designed to address specific barriers learners with disabilities may face," and the examples CAST gives are the three you already know: "text-to-speech, screen readers and speech recognition."


Then CAST says the useful part out loud:


The line between what is accessible and assistive technology can become blurred.

It blurs, on CAST's own account, because of "built-in accessibility features on the consumer devices many learners and families already own." The read-aloud function in a phone's browser and the read-aloud function in a licensed literacy suite are the same function. One of them is a product with a research page and a purchase order attached; the other is a setting.


That is the first thing to hold onto when a claim arrives. The claim is almost always made about a product. The research, where there is any, is almost always about a function.

What the clearinghouses hold on it


If you have read the five evidence-rating systems compared, you know the drill: find out whether the thing you are being sold is in a clearinghouse at all before you argue about its rating.


For assistive technology, the answer is that it is not in them.


The National Center on Intensive Intervention publishes three tools charts, and they are the closest thing this field has to an independent product register. Searched on 18 October 2026, the word "assistive" appears zero times and the phrase "text-to-speech" appears zero times across all three — the Instructional Intervention Tools chart, the Academic Progress Monitoring chart and the Academic Screening chart.


The subject filters are Reading, Mathematics, Behavior and, on the progress-monitoring chart only, Spelling and Written Expression. There is no access or technology domain, and there is no rubric for rating one.


That is not an oversight. NCII rates interventions and measures against technical standards, and a read-aloud function is neither. But it means the practitioner looking for an independent rating of a tool is looking somewhere that does not rate tools.


The federal practice guides are quieter still


Four practice guides bear on this directly. The word counts below were taken from the published PDFs on 18 October 2026.



This is the same shape found in the oral reading fluency norms and in the standard error of measurement: the number or the practice everyone attributes to a federal document is not in that document.


The one federal recommendation that names a technology


There is exactly one, and it is worth reading closely.


Recommendation 3 of the 2012 elementary writing guide is "Teach students to become fluent with handwriting, spelling, sentence construction, typing, and word processing." The panel rated it Moderate Evidence.


The guide states that this evidence "is drawn from nine studies of instruction in handwriting, spelling, sentence construction, and word processing," and then itemises them: three tested handwriting instruction, three tested explicit instruction in phonological awareness and spelling, two tested sentence-construction interventions, and "one study tested the effectiveness of practice using a word processor."


So the only federal recommendation in this area that names a technology names typing and word processing, rates the bundle Moderate, and rests the technology part of it on a single study. The recommendation is also about teaching a student to use the tool fluently, not about giving the student the tool.


And the guide's own definition of writing is more generous than you would expect


The same 2012 guide defines its terms in a way that matters here. Writing, it says, "can be done through paper and pencil, typing, audio recording, or speech synthesis," and "writing also can include students dictating ideas to an adult or peer for transcription."


A student dictating to a scribe is writing, on the federal panel's own definition. But the guide draws a second line immediately: "Writing from dictation, correcting grammatical errors on a worksheet, and combining two sentences generated by a teacher do not qualify as authentic writing, because students are not generating the content themselves."


The distinction is who generates the content, not who moves the pen. That is a cleaner principle for thinking about speech-to-text than anything on a vendor page, and it comes from a document that never once uses the phrase "assistive technology."

The meta-analysis everyone cites, and the three numbers inside it


One study does most of the work in this area. If you have seen a slide claiming that text-to-speech improves comprehension for students with reading disabilities, you have probably seen a number from it.


The study is Wood, Moxley, Tighe and Wagner (2018), "Does use of text-to-speech and related read-aloud tools improve reading comprehension for students with reading disabilities? A meta-analysis," in the Journal of Learning Disabilities, 51(1), 73 to 84 (doi 10.1177/0022219416688170). It is open access, so you can read it yourself rather than reading somebody's summary of it, which is the whole argument of this section.


It reports three numbers, and the one you will be shown is the first.


  • The headline: a random effects model yielded an average weighted effect size of d = 0.35, 95 percent confidence interval 0.14 to 0.56.

  • The adjusted figure: a trim and fill analysis "suggested that four studies were missing on the left side of the mean effect size," and the adjusted estimate was d = 0.24, 95 percent confidence interval 0.02 to 0.45. A lower bound of 0.02 is a confidence interval that has nearly reached zero.

  • The split that matters most: the authors report that "only a single, significant moderator emerged, namely whether the study design was a between- or within-subject study." Between-subject studies averaged d = 0.61, 95 percent CI 0.39 to 0.83. Within-subject studies averaged d = 0.15, with a confidence interval spanning zero.


Read that third bullet again, because it is the most practitioner-relevant sentence in the paper and it never appears on a product page.


In the one meta-analysis the field relies on, the thing that best predicted whether read-aloud support showed an effect was not the student, the grade or the tool. It was how the study was built.

When the same students were measured with and without audio, the average effect was 0.15 and the interval included zero. When different groups of students were compared, it was 0.61. Neither of those is a measurement of your caseload, and the honest reading is that the design question is unresolved rather than that one number is right.


What the authors said about their own evidence base


They were explicit, and it is worth quoting them rather than characterising them. The paper notes that "currently, there is a lack of randomized control trials," that the sample of studies was "relatively small," that "study quality remains a concern with many studies missing key components such as proper controls," and that "placebo effects and motivation have yet to be explored or explicitly controlled for."


Intervention intensity could not be examined at all because of inconsistent reporting across studies — the same reporting problem that turns up in the dosage literature.


Non-significant moderators, on the authors' own list, "included: what was read aloud, type of oral modality, reading material difficulty." Grade level did not moderate the effect either.


Hold on to "type of oral modality." It is about to matter.

The claim that sells the product


The most common product differentiator in this market is that human narration outperforms synthetic narration. Learning Ally's own efficacy paper states it plainly: the Wood meta-analysis, it says, "also suggests that human-read audio had a 50% greater effect on student comprehension versus synthetically narrated audio."


Here is where that figure comes from. Table 2 of the meta-analysis contains a moderator called Delivery Method, with two modes: Recorded Voice at 0.29 and Human Reader at 0.44. The ratio of those two point estimates is 1.52. That is the "50 percent greater" claim.


And "type of oral modality" is on the authors' own list of non-significant moderators.


So the claim is arithmetic performed on two point estimates from a comparison the authors did not find statistically significant, inside a meta-analysis whose adjusted overall effect was 0.24 and whose only significant moderator was study design.


What you may not conclude from this. You may not tell a family that synthetic narration is as good as human narration. A non-significant moderator in a small meta-analysis is not evidence that two conditions are equivalent; it is evidence that this body of studies could not separate them.


The defensible statement is narrower and more useful: the published comparison of human against synthetic narration is a non-significant difference between two point estimates, so if a budget decision is being made on the strength of it, the strength is not there yet. The same discipline applies here as in the working memory training literature — the boundary goes in the same sentence as the finding, every time.

One vendor research page is more careful than the field's shorthand


It is worth saying when a vendor gets it right, because it happens and because practitioners are routinely told the opposite.


Everway, the company that publishes Read and Write, maintains a research page on text-to-speech that lists seven sources, including the Wood meta-analysis and Keelor and colleagues' 2023 study in Annals of Dyslexia, 73(3), 469 to 486 (doi 10.1007/s11881-023-00281-9). Four of its statements are more conservative than most practitioner-facing summaries of the same literature:


  • On the role of the tool: text-to-speech "supports access, not replacement of instruction," and studies "consistently describe TTS as an access support rather than a substitute for reading instruction."

  • On test scores: in testing contexts students with reading-based learning disabilities "often show larger score increases when text is read aloud," but "higher scores under audio conditions reflect improved access to content" and "should not be interpreted as growth in underlying reading skills."

  • On the product versus the practice: "the tool alone does not drive outcomes."

  • On the feature everyone demonstrates: comparing audio alone against audio plus synchronised word highlighting, "adding highlighting did not consistently change comprehension scores," and whether highlighting adds value beyond the audio "remains an open question."


The page also states that the effect sizes reported for text-to-speech are "generally considered modest."


That last one is the sentence a practitioner needs and the sentence a slide deck drops. If a tool's own publisher describes the effect as modest and the mechanism as access rather than instruction, a referral conversation that describes it as a reading intervention has overstated it — and the overstatement did not come from the vendor.

The audiobook efficacy paper, read the way you would read a technical manual


Learning Ally's eligibility pages describe the Learning Ally Audiobook Solution as "a proven accommodation reserved for people with reading-based learning differences like dyslexia" and as "a proven multi-sensory reading accommodation." Its efficacy paper sets out what "proven" rests on, which is more than most vendors publish and is therefore worth reading carefully rather than dismissing.


The paper presents five district or school analyses. Every one of them compares students who used the product "with fidelity" against students who did not:


  • Denver Public Schools. Over 1,000 students; the district's Department of Accountability, Research and Evaluation is quoted as concluding that "students tended to see a higher increase in performance with increased usage of Learning Ally."

  • East Windsor Regional School District. On the STAR Reading assessment, "students not using the Audiobook Solution demonstrated an annual growth of 0.62 STAR Reading levels," while "students who used the Audiobook Solution with fidelity grew 1.25 STAR Reading Levels."

  • Baltimore Public Schools. Thirty-five schools; "teachers submitted evaluations of 166 students" based on Brigance word recognition and comprehension results, reported as grade-level gains.

  • Archdiocese of Denver. A study conducted by the Zarlengo Foundation, reporting that fidelity users "typically achieved about double the rate of reading growth" against an expected 0.4 grade levels, "in just 50 days."

  • State assessments. An internal evaluation comparing how often schools using the product appeared in the top band of state reading scores. The section heading says "top 25%" and the body says "top 35%" twice.


Two observations, neither of which is a criticism of the product.


First, a usage comparison is not a randomised comparison. In every one of these analyses the contrast is between students who used the tool as intended and students who did not. Students who use a tool as intended differ from students who do not in ways that also predict reading growth, attendance among them — the Denver analysis separately reports that fidelity users "exhibited lower rates of chronic absenteeism."


The district's own verb is "tended." That hedge is doing real work, and it survived into the marketing document, which is to the authors' credit.


Second, the word "proven" is the vendor's, and the strongest independent number in the whole area is the 0.24 adjusted effect above. Both things can be true: the product can be well made, widely used and worth recommending for access, and the published evidence can still be thinner than the adjective implies. The same gap between a label and its adjudication runs through this entire field.

The font claim, which has a clean answer


Most questions in this area end in "it depends." This one does not.


Wery and Diliberto tested the specialised dyslexia font OpenDyslexic in Annals of Dyslexia, 67(2), 114 to 127. The design was a single-subject alternating treatment design with twelve students in grades 3 through 6 identified as having dyslexia, comparing OpenDyslexic against Arial and Times New Roman (doi 10.1007/s11881-016-0127-1) across three tasks: letter naming, real word reading and nonsense word reading. Results were analysed by visual analysis and by improvement rate difference.


The abstract's finding sentence is as plain as research sentences get:


Results from this alternating treatment experiment show no improvement in reading rate or accuracy for individual students with dyslexia, as well as the group as a whole.

The authors add that while some students commented that the font was "new" or "different," "none of the participants reported preferring to read material presented in that font," and conclude: "These results indicate there may be no benefit for translating print materials to this font."


Twelve students in a single-subject design is a small study, and it measured rate and accuracy rather than comprehension or endurance. But it is a direct test of the specific claim the font developers make, it is open access, and it is more than the claim itself has behind it.


If a parent asks whether to buy the font or reformat the materials, you can answer that question from the published record in one sentence — which is rare enough in this area to be worth knowing.

The gate nobody writes about: who may certify eligibility


Here is the part that actually blocks the work.


The three largest accessible-format libraries in the United States — the National Library Service for the Blind and Print Disabled, Bookshare and Learning Ally — do not sell books to anyone who wants them. Each operates under a provision of federal copyright law, and each requires a professional to certify that the reader has a qualifying condition before the collection opens.


Learning Ally names the provision on its own eligibility page: "a special provision of the U.S. Copyright law (17 U.S.C. § 121, et seq., Marrakesh exemption)."


Which means that somewhere in your intake, someone has to sign something. The question practitioners ask, usually after a parent has already asked them, is whether that someone can be you.


Three systems. Three different answers. And a pattern worth noticing: the more precise the list, the less likely your title is on it.


The National Library Service publishes a list of professions


NLS is the most specific of the three. Its frequently asked questions answer "Who can certify people as eligible?" like this:


Eligibility must be certified by one of the following: doctor of medicine, doctor of osteopathy, ophthalmologist, optometrist, psychologist, registered nurse, therapist, or professional staff of hospitals, institutions, and public or welfare agencies (such as an educator, social worker, case worker, counselor, rehabilitation teacher, certified reading specialist, school psychologist, superintendent, or librarian).

NLS also confirms directly that "individuals are eligible if a certifying authority determines that their perceptual or reading disability prevents them from reading regular print materials."


Read the structure of that list rather than scanning it for your job. There are seven named professions, and then a category — "professional staff of hospitals, institutions, and public or welfare agencies" — whose parenthetical examples, including "certified reading specialist," are examples of institutional staff.


The word "therapist" sits unqualified in the first group. Nothing on the page says whether it means a therapist of any kind or a therapist of the clinical kinds the rest of the list describes, and nothing on the page addresses a practitioner in independent private practice.


Bookshare publishes twenty-five job titles


Bookshare is the most generous of the three about who counts, and the most explicit. Its criterion is capability rather than credential:


Bookshare eligibility is determined by a professional with appropriate expertise who has either evaluated the individual OR has reviewed evaluations that were conducted with the individual by others with appropriate expertise in Bookshare's qualifying conditions.

Note what that second clause does. A professional who has reviewed someone else's evaluation can certify. If a psychoeducational report has crossed your desk, you have done the reviewing part already.


Bookshare then publishes a list of "common job titles for professionals who may have the appropriate expertise," which "include, but are not limited to," arranged in two columns.


  • In schools: Special Education Teacher, General Education Teacher, Teacher of the Visually Impaired, Assistive Technology Professional, Dyslexia Specialist, Learning Disability Specialist, Reading Specialist, Resource Specialist, Librarian, Physical or Occupational Therapist, Speech Language Pathologist, School Psychologist.

  • Outside of schools: Doctor of Medicine, Doctor of Osteopathy, Neurologist, Ophthalmologist, Optometrist, Orientation and mobility specialist, Osteopathy, Psychiatrist, Psychologist, Psychiatrist again, Professional staff of Hospitals or Healthcare Institutions, Professional staff at Public or Social Welfare Agencies, Registered Nurse.


Twenty-five entries. Four observations:


  • "Educational therapist" is not one of them, in either column.

  • "General Education Teacher" is, which tells you how low the bar is meant to be. This is not a list built to exclude anyone.

  • The out-of-school column is almost entirely medical. There is no column for independent practice, and a practitioner who is not school personnel and not hospital or agency staff fits neither heading.

  • The list is informal in construction: "Psychiatrist" appears twice and "Osteopathy" is listed as a job title alongside "Doctor of Osteopathy." A list with duplicates in it is a list of examples, not a schedule of authorised signatories — which is consistent with "include, but are not limited to," and is the strongest textual argument that the controlling phrase is "appropriate expertise."


Bookshare also confirms that "IEP or 504 plans are not required to access Bookshare," and that it accepts membership in NLS or Learning Ally as verification of eligibility.


Learning Ally defines nobody — and its form asks for a licence number


Learning Ally has the widest published criterion of the three and the narrowest form.


On its eligibility pages, "eligibility may be determined by an educator or other qualified professional with experience identifying reading deficits, visual impairments, blindness, or physical disabilities," and "determinations may be based on classroom observations, student work samples, and informal or formal assessments." Elsewhere: "Qualified Educators can determine eligibility through assessments, observations and review of student work." Elsewhere again: "a Qualified Educator or Qualified Evaluator with experience making such determinations."


"Qualified Educator" and "Qualified Evaluator" are capitalised like defined terms and are not defined anywhere on the page. The membership application, revised October 2025, defers in the same direction rather than closing the loop: "Please have the following certification completed by a qualified evaluator.


Qualified evaluators may differ according to the condition. For more information visit: Eligibility Guide" — and the Eligibility Guide is the page that says "an educator or other qualified professional."


Then comes the form itself, and this is the finding that matters most for an independent practitioner. The Certifier Statement reads:


I attest to the reading deficit, blind or visual impairment, or other physical disability that limits the applicant's ability to effectively use standard print. I also attest to my competency to make this certification.

And the fields beneath it ask for: Name of Qualified Evaluator, Title/professional specialty, Licensing authority, License no., Employer, Address, Phone, Email, Signature.


A licensing authority and a licence number. If you hold AET's ET/P or BCET, or a structured literacy credential, or a university certificate, you hold no licence and there is no licensing authority to name — which is the whole subject of what learning specialist certification actually is and of the credential ladder. The prose invites an educator; the form has a field an unlicensed educator cannot complete.


This is the same lesson the supervision paperwork taught: when an organisation publishes both a policy and a form, the form is where the requirement is actually applied, and the two do not always agree. Word-count the form, not only the page.


The three systems certify each other, which is the practical way through


Here is the loop. Bookshare accepts NLS or Learning Ally membership as verification. Learning Ally's application offers a second option: "Check if you receive services from Bookshare and are a Bookshare member. Learning Ally will confirm your membership with their organization."


So the certification is transitive. One determination, accepted by the library with the broadest published criterion, propagates to the others — and the library with the broadest criterion and the most explicit list is the one that says "appropriate expertise" and includes general education teachers in its examples.


The honest answer to "may I certify this?" therefore has three parts, and none of them is a yes or a no:


  • Bookshare's published criterion is expertise, and reviewing an existing evaluation satisfies it. Your title is not on the list of examples; neither is the list exhaustive, by its own terms.

  • Learning Ally's prose would admit you and its form may not. Ask them in writing before you promise a family anything, and keep the reply.

  • NLS names seven professions and a category of institutional staff. If you are in independent practice, the published text does not tell you that you qualify, and the published text is all there is.


And whichever route you take, you are attesting to your own competence to make the determination — Learning Ally's form says so in terms. That attestation is yours under AET's own competence clauses, not the library's: I.II.A requires members to "provide only those professional services for which they have been adequately trained," and I.I.B binds practice to "the boundaries of their competence based on their education, training, supervised and professional experience."


Certifying that a student cannot effectively use standard print is a determination about a student's reading, which is squarely inside the work. Certifying something you have not assessed and have not reviewed is not.

What AET's own documents say about any of this


Nothing. And the word counts are worth having, because this is the body that credentials you.


The AET Code of Ethics, Revision #9, 2025 runs to roughly 3,000 words. Counted on 18 October 2026:


  • "assistive" — zero. "Assistive technology" — zero.

  • "accommodation" — zero. The word does not appear in any form.

  • "software" — zero. "computer" — zero. "digital" — zero. "audiobook" — zero. "text-to-speech" — zero.

  • "technology" and its variants — fifteen, and every one of them is about e-practice, data handling or artificial intelligence rather than about a tool a client uses.

  • "device" — three, in the same register: telecommunication devices used for e-practice, electronic devices used "to gather, manage, and store client information," and the disposal of electronic devices without data leaks.


The nearest thing to a clause on this is II.I.D, "recognize that technology is a rapidly evolving field." Its five sub-items ask members to:


  • "strive to stay current in knowledge and practice related to technology."

  • "use published materials ethically and only for the purposes intended."

  • "adhere to and educate clients about copyright laws, attribution protocol, and citation style, and guidelines pertinent to the use of technology, including artificial intelligence (AI)."

  • work to educate clients that AI sources "are not always credible, equitable, or inclusive."

  • "empower clients to discern the appropriate use of technology, including credible sites for learning and research."


Every one of those is about the practitioner's conduct with technology. Not one is about a tool in a student's hands.


There is a quiet irony in the one clause that does touch this. The accessible-format libraries exist because of a copyright provision, and AET's only instructional technology clause is about copyright — pointing at attribution and permitted use rather than at access.


The clause that actually governs a tool decision is the plainest one in the document, and it is not about technology at all. II.I.E: "select and use appropriate instructional methods, curricula, materials, and other resources." A read-aloud tool is a resource. The standard is appropriateness, and the Code leaves the judgement to you.


The AET Fact Sheet is quieter still: "assistive," "technology," "accommodation" and "software" each appear zero times. The fact sheet lists what educational therapists do and do not do — the DO NOT list has four items, the second of which is "administer cognitive, intelligence, or psychological tests (unless otherwise qualified to do so)," parenthetical included — and technology appears nowhere in either column.


This is the same silence found in the group size documentation and in the session notes codes: the professional body states a standard of appropriateness and leaves the specification to the practitioner.


Not named is not forbidden. It is also not guidance. If you want AET's position on certifying library eligibility, the published record does not contain one, and asking in writing is the only way to get one.

Four factual axes for reading an assistive technology claim


The four axes that work for comparing instruments transfer here with one change. For instruments the axes are what it measures, who supplied the norms, the age band and the qualification level. For tools, ask these instead.


Axis one: what function does it perform? Not what it is called. Speech output, speech input, word prediction, format conversion, text simplification, visual tracking. A single subscription usually bundles five or six functions, and research, where it exists, examined one.


Axis two: whose data is behind the claim, and is it about the function or the product? The Wood meta-analysis is about read-aloud presentation in general, not about any product.


A vendor that cites it is citing evidence for the function it performs, which is legitimate and is also not the same as evidence for its own software. A district usage analysis is about the product, and about the students who chose to use it.


Axis three: which task was measured? Comprehension of a passage read aloud in a research session is one outcome. Independent reading growth over a school year is another. A test score under an audio condition is a third, and the publishers say themselves that it reflects access and not skill. These do not substitute for one another — the same distinction that makes a benchmark and a percentile different objects.


Axis four: what gate does using it involve? Eligibility certification by a professional, a subscription or licence, a device the family must own, a school account the family cannot get at home, or nothing at all because the function ships with the operating system. This axis is the one that decides whether the tool reaches the student, and it is the one no efficacy paper discusses.

A decision sequence, not a decision


You will not find a recommendation here about which tool to use, for the same reason the instrument comparisons do not tell you which test to buy. What follows is the order of the questions.


  • Name the barrier in one clause. Decoding limits access to text the student can understand when heard. Transcription limits output the student can compose orally. Spelling limits output the student can compose by hand. Each of those points at a different function; "she needs assistive technology" points at nothing.

  • Check whether the function is already on the device. Read-aloud, dictation and spell-check ship free on every current phone, tablet and laptop. If the free function meets the barrier, the question of which product to license does not arise, and the family has nothing to buy.

  • Separate access from instruction, in writing. A tool that reads text aloud does not teach decoding, and a structured literacy program does not confer access to this week's history chapter. Both can run at once. Saying which is which in your plan prevents the conversation in which audio is mistaken for intervention — a distinction the adolescent intervention guides force on you anyway once a student is reading years below grade placement.

  • Decide what you will record. If the tool is in the plan, write down what it is for and what would tell you it is working. Your progress monitoring measures reading skill, not access. Decide in advance what evidence of access looks like — pages completed, assignments submitted, independent reading attempted — and record it as you would any other plan decision.

  • Settle the eligibility question before the family does. If an accessible-format library is the route, find out now whether the library will take your signature. Ask in writing, keep the answer, and if it will not, identify who in the student's life can sign and say so plainly rather than leaving a parent to discover the form.

  • Say what you do not know. The published evidence supports access; it does not establish that a tool raises reading skill. That sentence belongs in your intake conversation, and it is the informed consent discussion doing its job.

What is out of scope here, and where it belongs


Nothing here is legal advice. This post reports what organisations and federal research documents publish. It does not address, and you should not read it as addressing, any of the following:


  • Whether a student is entitled to a device, a licence or a service, or who has to pay for it. That is statute, regulation and the terms of a school district's obligations, not ethics and not published research.

  • Whether a tool may be used during a standardised test. That is the test publisher's own administration manual first, and the testing programme's accommodation rules second. The publisher-permission pattern is the right one to follow: read what the publisher permits for its own instrument, quote it, and stop. Changing the presentation of a normed test changes what the score means, which is the norms problem in a different coat.

  • Eligibility for special education, or the identification of a disability. A library's eligibility certification is not a diagnosis and does not establish one. Certifying that a reader cannot effectively use standard print says nothing about what a screener can and cannot tell you.

  • State dyslexia or reading-plan requirements, which differ by state and change.


For the first and the last of those, the people to ask are an attorney licensed in your state and your state education agency. For the second, the test publisher. Nothing here is legal advice. A practitioner who needs a ruling rather than a reading needs one of those three.

Where the published record runs out


Being explicit about the holes is more useful than filling them.


  • No independent clearinghouse rates these tools. NCII rates interventions, screeners and progress-monitoring measures; the What Works Clearinghouse holds no intervention report on text-to-speech software as a category. There is no equivalent of the five evidence-rating systems for an access tool, so a vendor's own research page is often the most organised summary available — which is why reading it closely, rather than dismissing it, is the skill.

  • The research is about functions, and the market sells bundles. Nothing published separates the contribution of speech output from highlighting, pacing, dictionary lookup and vocabulary supports inside one product. The one feature comparison discussed above found that adding highlighting "did not consistently change comprehension scores," and its publisher calls the question open.

  • Speech-to-text has almost no practitioner-facing evidence base at all. The federal elementary writing guide mentions speech recognition once, in a biography; its secondary counterpart does not mention it; neither mentions word prediction. A research literature on dictation and speech recognition for students with learning disabilities does exist, but it is not summarised anywhere a practitioner would find it, and this post does not assert what it shows.

  • Nobody publishes the independent-practice case. Every eligibility list is organised around institutions — schools, hospitals, agencies. None of the three libraries addresses a private practitioner directly, and all three were asked the same question by this post and answered it only by implication.

  • Learning Ally's and Bookshare's full policy documents are not public in the way their FAQs are. What is quoted above is what is published. If a determination is load-bearing for a family, email the organisation and keep the reply, the way you would with any CE or renewal requirement.

Frequently asked questions


Can an educational therapist certify a student's eligibility for Bookshare or Learning Ally?


Bookshare's published criterion is "a professional with appropriate expertise" who has either evaluated the student or reviewed someone else's evaluation, and its list of common job titles is explicitly non-exhaustive — but "educational therapist" is not on it.


Learning Ally's criterion is "an educator or other qualified professional," which is broader still, while its application form asks the certifier for a licensing authority and a licence number. Ask the organisation in writing and keep the answer; do not promise a family you can sign before you have it.


Does text-to-speech improve reading comprehension?


The meta-analysis the field relies on reported an average weighted effect of 0.35, falling to 0.24 after adjustment for possible publication bias, and found that the only statistically significant moderator was whether a study used a between-subject or a within-subject design, at 0.61 and 0.15 respectively.


Its authors noted the absence of randomised controlled trials and that study quality remains a concern. The defensible statement is that read-aloud presentation supports access to text; it is not established that it raises underlying reading skill.


Is human narration better than synthetic narration?


The published comparison is a two-mode moderator in that same meta-analysis, at 0.44 for a human reader against 0.29 for recorded voice, and the authors list type of oral modality among the non-significant moderators. The widely quoted "50 percent greater" figure is the ratio of those two point estimates. It is not evidence that the two are equivalent either; it is evidence that this body of research could not separate them.


Do specialised dyslexia fonts help?


The direct test of OpenDyslexic, a single-subject alternating treatment design with twelve students in grades 3 to 6 identified as having dyslexia, found "no improvement in reading rate or accuracy for individual students with dyslexia, as well as the group as a whole," and no participant preferred reading in the font. The authors concluded there "may be no benefit for translating print materials to this font."


Does AET's Code of Ethics say anything about assistive technology?


No. The words "assistive," "accommodation," "software," "computer" and "digital" each appear zero times in Revision #9, 2025, and its fifteen uses of "technology" concern e-practice, data handling and artificial intelligence. The clause that governs a tool decision is II.I.E, which requires members to "select and use appropriate instructional methods, curricula, materials, and other resources" and leaves the judgement to the practitioner.


Is using a read-aloud tool an accommodation or an intervention?


Everway, which publishes one of the best-known literacy tool suites, states in its own research summary that text-to-speech "supports access, not replacement of instruction," and that higher test scores under audio conditions "should not be interpreted as growth in underlying reading skills." Treat it as access, record it as access, and run instruction alongside it.

Keep the practitioner-facing work coming


This site publishes one practitioner-facing article a day — credential landscapes, instrument comparisons, and the documents behind the claims the field repeats. Related reading: educational therapist versus learning specialist versus executive function coach, dysgraphia and writing intervention tools, and is educational therapy evidence-based.


If that is useful, join the Illuminate educator list and the next one will reach you directly.

Recent Posts

See All
Standard Error of Measurement: What That Band Means

Every score report prints a confidence interval and almost no practitioner acts on it. What the standard error of measurement is, why the band's width is a software setting rather than a property of t

 
 
 

Comments

bottom of page