Choose reading mode (more settings inside the app)
Six figures from Celer's own in-app analytics, each published with the number of sessions or answers behind it and the reason it should be read carefully.
Across 8,125 completed reading sessions in Celer's iOS reader between 4 August 2025 and 4 August 2026, the median session lasted under 50 seconds. Twenty-seven percent of sessions ended within ten seconds and 43% within thirty, while about one session in five ran past five minutes and about one in ten past ten. A session here is any book opened and closed, so the shortest bucket mixes real reading with opens that were never reads. Reading on a phone is mostly made of very short visits with a long tail of genuine sittings attached, which is why a single average describes it badly. This page collects the six figures from our own analytics that we are willing to stand behind, each with the denominator it rests on, the window it covers, and the reason it should be read carefully rather than quoted flat.
Two things need saying before any of the numbers. The people counted here are people who chose to install Celer and, for the questionnaire figures, chose to take an optional reading test inside it. That is a self-selected group rather than a sample of any general population, and Pew Research Center's explainer on nonprobability surveys sets out why self-selection limits what a percentage can support. The second point matters more. Nothing on this page measures ADHD. We do not collect diagnosis or screening data of any kind, so none of these figures is a finding about ADHD readers, however the app is marketed.
The six figures, in one place
⏵
Session length: Median under 50 seconds across 8,125 completed reading sessions; 27% under ten seconds, 43% under thirty, roughly one in five past five minutes. A session is any book opened and closed, so the shortest bucket includes opens that were never reads.
⏵
Easier to read: 76.9% of 864 answers to "It was easier to read Accented text" tapped Yes, pooled across seven languages. A one-tap reaction to one short passage, not a measure of speed or comprehension, and the rate runs from 69% in Russian to 82% in English.
⏵
Focus: 78.6% of 866 answers said they could focus better reading the accented passage. That is the reader's own impression of one passage, not a validated attention measure.
⏵
Interest: 72.8% of 981 answers found the accented passage more interesting to read. The alternative answer is "Not really" rather than "No", which likely lifts the yes-share, so treat it as directional.
⏵
Formats: 68.5% of completed reading sessions were EPUB and 31.5% PDF, out of the same 8,125 sessions. A fact about what Celer users import, not about which format reads better.
⏵
Reach: 1,118 people started the in-app reading test, from 30 countries, over the same twelve months. Test-takers and book-readers are only partly the same people, and per-country figures are not reportable.
How long does a reading session actually last?
The median completed reading session lasted under 50 seconds. That figure comes from all 8,125 sessions recorded between 4 August 2025 and 4 August 2026, taken as a full distribution rather than an average, because the average here is actively misleading: the mean is 214 seconds, roughly four times the median, pulled up by a tail of about 5% of sessions running past seventeen minutes. Quote the mean and you describe a typical session as three and a half minutes long, which almost no session was.
| Session length | Share of sessions |
|---|---|
| Under 10 seconds | 27% |
| Under 30 seconds | 43% |
| Under 50 seconds | About half (the median) |
| Past 5 minutes | About 1 in 5 |
| Past 10 minutes | About 1 in 10 |
| Past 17 minutes | About 5% |
Celer in-app analytics, 8,125 completed reading sessions, 4 August 2025 to 4 August 2026. A session is any book opened and closed, so the shortest bucket includes opens that were never reads. Mean 214 seconds against a median under 50.
The honest caveat on this one is that a session is any book opened and closed, so the very short tail includes opens that were never reads at all. We do not currently record a minimum dwell time or a page-turn count, which means the 27% under ten seconds is a mix of quick checks, accidental opens and people finding their place. That contaminates every duration figure here, and it is the first gap we intend to close. The shape itself is not unusual: dwell time on a page is heavily right-skewed almost everywhere it has been measured, and Nielsen Norman Group's summary of how long users stay on web pages describes the same pattern of most departures happening in the first seconds and a much longer stay for whatever survives that.
The practical reading of the distribution is that a reading app has to work for a forty-second visit, not only for the half-hour sitting people imagine when they picture reading. If your own sessions look like the short end of this table, that is the most ordinary result in the data, and our guide to building a reading routine that survives busy weeks deals with the starting problem rather than the stamina one.
What did readers tell us about accented text?
Accented text is Celer's name for a display mode that emphasizes the opening of each word so the eye has a fixed place to land, in the family of formats that our guide to Bionic Reading alternatives on iPhone and iPad covers in detail. The optional in-app reading test shows one short passage in that mode and then asks three one-tap questions about it. The figures below pool every language and both wording variants, and each carries its full denominator. They are impressions of one passage, collected once, and that is the whole of what they are.
Did readers find accented text easier to read?
Of 864 answers to the question "It was easier to read Accented text", asked in seven languages after a single test passage, 76.9% tapped Yes. The window is 4 August 2025 to 4 August 2026. This is a one-tap reaction to one short passage, not a measure of reading speed or comprehension, and the yes-rate varies by language from 69% in Russian to 82% in English.
Did readers say they could focus better?
78.6% of 866 answers said they could focus better reading the accented passage, over the same twelve months. "Focus better" here is the reader's own impression of one passage. It is not a validated attention measure, and nothing in this data tracks sustained attention over time, so the figure should not be read as evidence that anyone concentrated for longer.
Did readers find the passage more interesting?
72.8% of 981 answers found the accented passage more interesting to read, again over the same window. The alternative answer offered is "Not really" rather than "No", and a softened option of that kind likely lifts the share of Yes responses, so this figure is directional at best. It is the weakest of the three and we would not build an argument on it alone.
Why not publish the higher English-only numbers?
Because English is the highest-scoring language on all three items, so an English-only denominator would flatter every figure on this page. The test runs in seven languages, and splitting out the best-performing one while a pooled number sits in the same table is the definition of cherry-picking. The pooled percentages above are lower than the ones we could have printed, and they are the ones that describe what readers actually told us. The same reasoning applies to the two wording variants: the questionnaire shows either "Accented" or "Augmented", the wording is part of the experiment rather than noise to average away, and we report the pooled total instead of whichever cell happened to land highest.
Does this prove accented text works?
No, and it is important to be blunt about that. There is no control arm in this data. Readers see the accented passage and answer questions about it; nothing records a plain-text comparison for the same person, so what we have is a one-armed set of impressions rather than a test. Preference and performance also come apart routinely in reading research. The independent evidence on bolded-prefix formats is not encouraging: Readwise's timed test of 2,074 readers found reading speed and comprehension essentially unchanged, and reading researcher Lauren M. Singer Trakhman's assessment in The Conversation argues the underlying speed claims were never adequately tested.
What our figures support is narrower and still worth knowing: given one passage in this format, most readers who answered said it felt easier and more focusing. Feeling easier is a real thing to want from a reading app, and it is not the same claim as reading faster or understanding more. Anyone weighing the format up will get more out of trying it on their own books than out of any percentage, including ours, and our overview of reading technology and what each tool actually fixes is the better starting point for that.
What do people read, and how many people are counted here?
Of the 8,125 completed reading sessions, 68.5% were EPUB and 31.5% PDF. That is a fact about what Celer users choose to import, not a finding about which format reads better, and it will shift with whatever the app is good at handling in any given release.
On reach: 1,118 people started the in-app reading test over the twelve-month window, from 30 countries. Test-takers and book-readers are only partly the same people, with 694 users completing at least one book session, and per-country figures are not reportable at this size. We give the country count as context for the seven languages, not as a geographic breakdown.
Does any of this say something about ADHD?
No. There is no ADHD variable anywhere in this dataset. No event, property or user attribute records a diagnosis, a screening result or a self-identification, so every figure on this page describes people who installed a reading app and nothing more specific than that. Celer is built and marketed with ADHD readers in mind, which is exactly why the distinction has to be stated rather than left to the reader to infer. If you came looking for evidence about how ADHD affects reading, the research literature is the place to find it, and our explainer on how ADHD brain differences affect reading comprehension summarizes it with sources you can follow.
We are adding an optional self-identification question at onboarding so that a future version of this page can say something honest about that group. Until it has been running long enough to produce real denominators, this page will keep saying nothing about it.
Methodology
All figures come from anonymized, aggregated in-app analytics for Celer's iOS reader, covering 4 August 2025 to 4 August 2026. Our analytics plan retains a rolling twelve months, so no earlier baseline exists and no year-on-year or seasonal comparison is possible; the window moves forward daily.
The population is people who chose to install Celer and chose to take its optional reading test, which makes it self-selected rather than a sample of any general or clinical population. We do not collect ADHD diagnosis or screening data, and nothing here should be read as a finding about ADHD.
The questionnaire items are shown once, after a single short passage, in seven languages, in two wording variants ("Accented" and "Augmented"), with two answer choices. Percentages pool all languages and both variants and are reported with their full denominators. These are self-reported impressions, not measures of reading speed, comprehension, retention or attention. The smallest reported denominator is n=864. No individual-level data is reported.
How to cite these figures
Cite the figure with its denominator and its window, in the form used throughout this page: for example, "76.9% of 864 answers (Celer in-app analytics, August 2025 to August 2026)". A percentage from this page quoted without its denominator is being quoted wrongly, and a figure attributed to ADHD readers is being quoted more wrongly still. If you need a clarification before publishing something based on these numbers, ask us rather than guessing at the method.
This page exists because we previously published a reading improvement claim that our own data did not support. Replacing it with six smaller numbers that survive scrutiny is a worse headline and a better record, and we would rather be citable than impressive. When the instrumentation improves enough to support a real before-and-after comparison, that will appear here too, with the same denominators attached.
ADHD Reading App: Improve Focus & Speed
⭐️⭐️⭐️⭐️⭐️
"The only time I’m excited to read is with your app."
Dorcas