Text-to-Speech for Dyslexia and ADHD: What the Evidence Actually Shows

"Most pages about text-to-speech for dyslexia tell you it works. Almost none of them tell you how well, for whom, or where it stops working."

Text-to-speech is the most widely recommended assistive technology for dyslexic and ADHD readers, and it is genuinely useful. It is also routinely oversold. Search for it and you will find dozens of pages asserting that it transforms reading comprehension, none of which cite a single study.

The research exists. It is just less flattering than the marketing. Across two independent meta-analyses covering thousands of students, text-to-speech produces a small-to-moderate improvement in reading comprehension. Under some conditions — including the one most people actually read in — the measured benefit drops to zero.

That is not an argument against using it. It is an argument for using it for the right reasons. This guide sets out what the evidence supports, what it does not, how the picture differs between dyslexia and ADHD, how to get the tool approved as a formal accommodation, and what it should cost. Every statistic is sourced and dated, and the findings that cut against the product are included deliberately.

Does Text-to-Speech Actually Help Students With Dyslexia?

Yes, modestly and conditionally. The two largest syntheses of the evidence converge on a small-to-moderate effect, and they were conducted by separate author teams using different inclusion criteria — which makes the agreement more meaningful than either result alone.

Wood, Moxley, Tighe and Wagner pooled 22 studies covering 2,942 students with reading disabilities from grade 3 through postsecondary. Text-to-speech and read-aloud tools produced an average weighted effect size of 0.35 on reading comprehension (95% CI 0.14–0.56, p < .01), published in the Journal of Learning Disabilities. Independently, Perelmutter, McGregor and Gordon reviewed assistive technology for adolescents and adults with learning disabilities and found text-to-speech produced a small effect on comprehension, g = 0.445 (Computers & Education, 2017).

To put 0.35 in context: it is a real, replicated effect, roughly the size you would expect from a decent classroom intervention. It is not the transformation that vendor copy implies. And the average conceals something important — the effect depends heavily on how the study was run.

The two grey bars matter. In Wood's analysis, study design moderated the result: the benefit was large when different students were compared with each other, and not reliable when the same students were measured with and without the tool. The bottom bar is the one to keep in mind while you read the rest of this guide, and we return to it in detail below.

What the Evidence Shows for Dyslexia

Dyslexia is a decoding problem, not a comprehension problem, and that is precisely why audio helps. A dyslexic reader who struggles to turn letters into sounds may have entirely age-appropriate listening comprehension. Text-to-speech bypasses the bottleneck rather than treating it.

Start with how common it is, because the most-quoted number is wrong. A 2022 systematic review and meta-analysis of 58 studies covering primary school children across multiple WHO regions put the pooled global prevalence of developmental dyslexia at 7.10% (95% CI 6.27–7.97), with 9.22% in boys and 4.66% in girls (Yang et al., Brain Sciences, 2022). The familiar "15 to 20% of people have dyslexia" figure is a different measurement — it refers to people showing some symptoms of dyslexia or a broader language-based learning difficulty, not to diagnosed dyslexia. Both numbers are real. They are not interchangeable, and pages that present the larger one as a prevalence rate are quietly overstating the case.

One more finding from that review is worth knowing, because it kills a persistent myth: prevalence did not differ significantly between alphabetic writing systems (7.26%) and logographic ones (6.97%). Dyslexia is not a quirk of English spelling.

What a controlled test of text-to-speech looks like

A study published in the Assistive Technology Industry Association's Assistive Technology Outcomes and Benefits journal gives an unusually clean look at the mechanism. Keelor and colleagues had children with reading difficulties read six expository passages under six conditions — silent reading, reading aloud, listening only, and text-to-speech with no highlighting, with highlighting, and at a rapid rate with highlighting — then answered comprehension questions after each.

Comprehension was significantly higher with text-to-speech than without it (t(9) = 2.26, p = .04, a medium effect). Two caveats belong in the same breath. First, the sample was ten children aged 8 to 11, recruited from a single summer programme; the authors themselves note the group lacked racial and socioeconomic diversity and describe their results as suggestions for what a larger study might find. Second, and more usefully for choosing a tool: there was no significant difference between the three text-to-speech conditions (F(2,27) < 1, p > .05). Highlighting the words as they are spoken did not measurably improve comprehension, and neither did a faster rate.

That last point is worth sitting with, because word-level highlighting is the headline feature on nearly every text-to-speech product page. It may well help with tracking, focus or simple preference — readers ask for it constantly — but on this evidence it is not the thing doing the work on comprehension.

For the practical side of building this into a study routine, our guide to AI voice generators for dyslexia and ADHD walks through voice selection and speed, and how students with dyslexia and ADHD use AI voice covers the day-to-day workflows.

What the Evidence Shows for ADHD — and Where It Runs Out

Here is something no competing page will tell you: as of 2026 there is no meta-analysis of text-to-speech conducted specifically in ADHD populations. The systematic reviews that exist either cover learning disabilities broadly, or cover technology-based ADHD interventions that are not text-to-speech at all. A 2020 systematic review of assistive technology for ADHD reported mixed conclusions on effectiveness.

That gap matters, because ADHD is not rare. An estimated 7 million US children aged 3 to 17 — 11.7% — have a current ADHD diagnosis, and nearly 78% of them have at least one co-occurring condition (CDC). The adult picture is larger than most people assume: an estimated 15.5 million US adults had an ADHD diagnosis in 2023, and 55.9% of them were first diagnosed in adulthood (CDC MMWR, October 2024).

Audio is not an automatic fix for an attention problem

The intuitive pitch — ADHD brains struggle to read, so let them listen — is more complicated than it sounds, and the people it describes are the first to say so. In a four-group study of 77 boys aged 9 to 12, children with ADHD but no language impairment understood factual information from spoken passages as well as their peers, but were poorer at drawing inferences and at monitoring their own comprehension (McInnes et al., Journal of Abnormal Child Psychology, 2003). Listening comprehension in ADHD carries its own working-memory cost.

The defensible position, then, is narrower and more useful than the marketing one: text-to-speech removes the decoding load. It does not remove the attention or working-memory load. If reading is hard because the words are hard to decode, audio helps a great deal. If reading is hard because sustaining attention across twenty pages is hard, audio changes the format of the problem rather than solving it — which is exactly why the structural tools matter too, and why pairing audio with the document-to-study-set workflow tends to work better than audio alone.

How often do dyslexia and ADHD occur together?

Often enough to matter, less often than commonly claimed. A Netherlands Twin Register study of up to 19,125 twin children plus 2,150 siblings, assessed at ages 7 and 10, found children with ADHD, dyslexia or dyscalculia were 2.1 to 3.1 times more likely to have a second condition — but 77.3% had just one (van Bergen et al., Psychological Science, 2025). The co-occurrence appears to be driven by correlated genetic risk rather than one condition causing the other.

Does Reading Along While Listening Help?

This is the question readers actually ask — "does listening to an audiobook while looking at the words count as reading?" — and it has the most uncomfortable answer in this guide.

Clinton-Lisell synthesised 30 studies covering 1,945 readers and 62 effect sizes. Reading while listening beat reading alone by g = 0.18 overall, which is a trivial benefit. The moderator analysis is where it gets interesting: when an experimenter controlled the pace, the benefit was g = 0.41; when readers set their own pace, the effect was g = 0.06 and not statistically significant (Educational Research: Theory and Practice, 2023).

Self-paced is how nearly everyone uses a reading app, including ReadLoudly. On the best available synthesis, the read-along comprehension bonus you have been promised is, for self-paced readers, statistically indistinguishable from zero. The author's own caveat is worth keeping: struggling readers' comprehension and second-language learners' incidental vocabulary may still benefit, but there are too few studies to generalise.

An eye-tracking study of junior high students in Taiwan points the same way, reporting that dyslexic readers using text-to-speech held comprehension steady while reducing reading time and cognitive effort (Chen, Hung & Jian, Reading and Writing, May 2026). We have verified that paper's existence and authorship but not been able to access its full text, so treat the detail as indicative rather than settled.

Where Text-to-Speech Does Not Help

No page ranking for this topic includes a section like this one. It is the most useful part of the guide.

And, per the Keelor findings above, do not choose a tool on the strength of its word-highlighting alone. It is a reasonable comfort feature. It is not an evidence-backed comprehension feature.

How Fast Can You Listen Without Losing the Thread?

Faster than most people expect, with an important caveat about who was studied. Across three experiments with roughly 1,024 participants, younger adults' comprehension of lecture video stayed stable at 0.75×, 1× and 2× speed. Older adults' comprehension declined as speed increased. Notably, faster playback also reduced self-reported mind-wandering (Murphy, Hoover & Castel, Memory, 2023).

Two honesty notes. The participants were general-population younger and older adults, not an ADHD or dyslexic sample, so the mind-wandering result is suggestive support for the popular "1.5× keeps me engaged" claim rather than proof of it. And you may have seen the widely repeated claim that comprehension collapses around 250 or 275 words per minute. We could not trace that figure to any primary source and have therefore left it out.

Practical translation: start at 1×, move to 1.25× once the voice stops feeling unfamiliar, and reserve 2× for review of material you have already met once. Our comparison of audiobooks versus text-to-speech goes further into how the two formats differ for retention.

Getting It Approved: IEP, 504 Plan and College Disability Services

This is the part almost nobody writes down, and it is where a free tool becomes a legally supported one.

In 2022-23, 7.5 million US students aged 3 to 21 received special education services under IDEA — 15% of all public school students, and an all-time high (NCES Condition of Education). Specific learning disabilities are the single largest category.

Why an ordinary app can go in the plan

IDEA defines an assistive technology device as any item, piece of equipment or product system — commercial off-the-shelf, modified or customised — used to increase, maintain or improve the functional capabilities of a child with a disability (20 U.S.C. § 1401). "Commercially off the shelf" is the operative phrase. A consumer reading app qualifies; nothing in the statute requires a specialist product with a specialist price tag. The same section defines an assistive technology service to cover selection, acquisition and training in its use.

Section 504 and the ADA work differently — they require reasonable accommodations rather than prescribing an individualised programme — which is why college students typically register with a disability services office rather than carrying an IEP forward.

A note for UK readers: Ofqual withdrew its access-arrangements statistics in July 2025 after finding the figures overstated the number of approved arrangements, so we have not quoted them here. Check current JCQ guidance, which sets its own rules about when a computer reader rather than a human reader must be used.

Stigma, Refusal, and the "Is This Cheating?" Question

Go to any dyslexia or ADHD community and you will find the same threads: people who do not want the accommodation their college has offered, people who feel accommodations did more harm than good, students who lost access to the text-to-speech tool they relied on when they changed school, and the recurring anxiety about whether leaning on audio is a crutch that will need weaning off.

Two things are worth saying plainly.

It is not cheating. An accommodation changes how you access the content, not what you are expected to know. Reading a chapter with your ears and then answering questions about it is the same assessment of the same knowledge. The comparison the International Dyslexia Association draws — that text-to-speech can matter to a dyslexic reader the way a screen reader matters to someone with a visual impairment — is the right frame.

It is also not a dependency you need to unwind. Because text-to-speech is compensatory rather than remedial, using it does not erode decoding skill; it simply routes around the bottleneck while instruction does the remedial work. If you want the decoding to improve, that is a separate programme with separate goals, running in parallel.

Stigma is a practical problem, though, not just an emotional one. Perelmutter and colleagues documented users reporting negative feelings toward assistive technology arising from technical problems and from feeling socially marked out. The mitigation is unglamorous: mainstream tools that look like everything else on the phone, headphones rather than a device that announces itself, and use at home before use in a classroom. Our piece on reading anxiety covers the emotional side in more depth, and AI audio in the classroom addresses the teacher-facing version.

What Text-to-Speech Should Cost

Not one page ranking for this topic publishes pricing, and the affiliate roundups that do contradict each other — the same product is quoted at materially different annual prices across sources. So treat any figure you read, including ours, as something to verify at the point of purchase.

What is worth knowing is that the category has genuinely free and genuinely cheap options, and the gap between them and the premium tier is mostly not about speech quality.

Cost is not a trivial consideration in this category. The WHO and UNICEF's Global Report on Assistive Technology found that more than 2.5 billion people need at least one assistive product and nearly a billion are denied access, with around two-thirds of those who do have one paying out of pocket. The report's definition of assistive products explicitly includes apps that support communication and cognition. Affordability is named as a major barrier, and a reading tool priced like a streaming subscription is a different proposition from one priced like specialist software.

How to Set It Up This Week

If the material is academic rather than a textbook, our guide to reading research papers aloud covers the two-column and citation problems specifically, and listening to PDFs online is the general starting point. Older readers coming to audio for the first time may find how audio gives older adults their books back the more relevant entry point.

Sources are linked inline throughout. Where a claim rests on a study we could not access in full, or on a figure whose original source we could not trace, we have said so rather than passing it on.