Choosing a reading app for dyslexia

An app store search for dyslexia returns hundreds of results, and the honest ones are not labelled. The badges, the star ratings and the word scientifically do not separate them. What separates them is a short list of questions you can put to any reading app in about ten minutes, mostly without downloading anything. Here is the list we use — and because fairness demands it, our own app goes through it last.

The questions are not technical. They are about claims: what the app says it will do for a dyslexic reader, and whether that claim is one anybody could check. A reading app can be honestly useful without a single study behind it, provided it claims accordingly. The trouble is almost never the tool. It is the sentence on the label.

1. Does it claim to treat, cure or retrain anything? A product that does has told you most of what you need to know about it.
2. Does it measure comprehension, or only speed? A speed figure with no comprehension figure beside it is the industry’s oldest sleight of hand.
3. Are spacing, motion and pace controls you hold, or decisions the app has made on your behalf and called a philosophy?
4. Is there text to speech — and does the voice run on your device, or does your text travel to somebody’s server to be spoken?
5. Where does your reading data go, who can see it, and does the app still work with the network switched off?

Does it claim to treat anything?

Dyslexia is a specific learning difficulty affecting decoding and spelling, and the interventions with real evidence behind them are structured literacy instruction — delivered by people, over months. No app that reformats, paces or voices text is a treatment, and the good ones say so in plain sight. If a product’s page uses the language of therapy — retrains the brain, corrects, cures — walk away, whatever else it does well. Not because the features are necessarily bad, but because a company willing to make that claim where you can see it will be making others where you cannot.

There is a softer version of the same failure: the testimonial wall, where the treatment claim is outsourced to screenshots of grateful users. The experiences are real and worth something. They are also the cheapest evidence there is, which is why every product in every market has them. The question to hold onto is whether the claim survives being written down plainly: this app will make your dyslexia better. If the company will not write that sentence, the testimonials should not be allowed to whisper it.

Does it measure comprehension, or only speed?

Words per minute is the easiest number in reading to improve, because you can raise it indefinitely by understanding less. Any app that reports your speed owes you a comprehension figure next to it, and an app that celebrates speed alone is measuring the one thing that means nothing on its own. The trade-off is not controversial: beyond roughly 400–500 words per minute on unfamiliar prose, comprehension falls away, and what remains is skimming — a legitimate skill, but not the one being advertised.

The test need not be elaborate. A short recall check after a passage — did these words appear, did these ideas — is enough to keep the speed number honest. Better still, the check itself is one of the better-evidenced things a reading app can contain, because retrieval practice improves retention more than re-reading does. If there is no check at all, treat every number the app shows you as decoration.

Are spacing and motion yours to control?

For dyslexic readers the best-supported ordinary visual intervention is spacing: pushing letters and lines apart reduces visual crowding, and increased spacing has been found to improve reading speed and accuracy in dyslexic children with no training attached. So look for the dials. Letter spacing, line spacing, text size — and, if anything on the screen moves, one control that scales all motion down to zero. An app confident in its defaults will still hand you the dials. An app that will not is telling you whose comfort it was designed around.

Motion deserves its own sentence. Animation can carry information, but for some readers it is purely a cost, and an app that respects the operating system’s reduced-motion preference without being asked has thought about the people most likely to be using it. That is a detail, and details of that kind predict the others.

Is there text to speech, and where does it run?

Because dyslexia is a difficulty with decoding rather than with thinking, comprehension by ear is typically intact, and text to speech is the single most useful feature an app can offer this audience — the case is set out fully in text to speech as a reading tool. Its presence is close to non-negotiable. The follow-up question is newer: does the voice run on your device, or does your text travel to a server to be spoken? Cloud voices upload what you read. For a contract, a medical letter or a half-written thesis, that is not a detail.

On-device used to mean robotic; that has stopped being true, and small neural voices now run entirely inside a browser. Where the voice runs is therefore a genuine choice, and an app should be able to answer it in one sentence. If the answer takes a privacy policy to express, that is also an answer.

Where does your reading go?

Reading is close to the most personal data there is — a log of what you are trying to understand. An app for dyslexic readers holds something more delicate still: the fact of the difficulty itself, plus every score you ever post. So ask where sessions are stored, what leaves the device, and whether an account is required at all for the app to function. The blunt test is the network. If the app works with the connection off, your reading is staying with you. If it will not start, you know where your reading lives.

The trouble is almost never the tool. It is the sentence on the label.

The checklist, compressed

Our own app, against the list

Fairness requires that we go last and answer plainly. Treatment claims: none, as policy — this is not a diagnosis, a therapy or a treatment, and we will never describe it as one; we keep a public list of the claims we refuse to make. Comprehension: a session can end in a recall test that mixes words which genuinely appeared with plausible distractors, and the score sits next to the speed, which keeps us honest too. Motion: one slider scales all animation to zero, and the app obeys the system’s reduced-motion setting unasked. Voice: on-device — a small neural voice running in the browser, so nothing you read is spoken by a server. Data: the whole app is a single file that works offline; nothing is uploaded.

And now the honest failures against our own list. The presentation is unusual — one word at a time at a fixed point — which removes crowding and line-tracking but also removes re-reading, and re-reading is precisely what many readers lean on when a sentence fails to land. Conventional spacing controls barely apply when there is only one word on screen, so a reader who mainly wants a better page, rather than a stream, is better served elsewhere; the survey in reading tools for dyslexic adults covers those options without favour. Above all: we have not tested the app with dyslexic readers. A clean mechanism is not a demonstrated result, and we decline to pretend otherwise.

Ten minutes, before you download

All five questions can be answered from a product page, a privacy policy and one trial passage, in roughly the time it takes to make tea. Ask them in order, because the order is a filter: a wrong answer to the treatment question ends the interview, a missing comprehension measure demotes the app to entertainment, and the remaining three decide whether it deserves your documents. Any dyslexia reading app worth keeping survives all five without special pleading. Most do not, which is not cynicism — it is ten minutes well spent, and it leaves you with the short list that was always the point.