NF
Nora Fink
Co-founder · Research & ProductAI researcher and mother of a child with dyslexia. Nora leads the scientific framework, item design and transparent model development.
AGES 7+ · AT HOME · IMMEDIATE RESULTS
See in about 10 minutes where reading is getting stuck — and what can help next.
Dyslexia99 combines six targeted screening areas covering reading, speech sounds, processing speed, handwriting, and sentence comprehension. Directly after, you get a detailed profile with strengths, flagged patterns, and actionable next steps.
- Six targeted task areas instead of a generic questionnaire
- Detailed PDF report with individual results for every single area
- Age-adjusted content in German, English, and Spanish
US$49 one-timeUS$49Includes full screening and detailed report
Secure parent area · Instant results upon completion · One-time payment
Screening only. Not a medical, psychological, or educational diagnosis.
Dyslexia99
Good test qualityAlex · age 10
Overall screening viewSeveral areas deserve a closer look
Word recognitionRelative strength
Speech soundsWatch closely
Visual speedRelative strength
Read aloudWatch closely
SentencesRelative strength
Writing sample patterns
Our research & reviews are represented across leading scientific platforms
17 ORCID-indexed research works as well as additional public datasets, research records, and peer-review activity for PLOS ONE.
arXiv infrastructure
Peer-review activity
Research preprints
Research preprints
Open-science records
Research & datasets
Research preprints
Research datasets
Verified researcher profile
Public datasets
Academic profilePublications, preprints, datasets and researcher profiles — directly verifiable on the respective platforms.
~10 minTo initial clear results
6 areasHolistic rather than one-dimensional
7+Age-adjusted tasks
1 reportStrengths, patterns and next steps
A calm, guided screen from setup to report
Every task is short and clearly explained. Content changes with age, while the overall flow stays familiar.
01
Quick tech check
Check the camera, microphone and room noise before the timed tasks begin.
02
Writing sample
Photograph a recent, uncorrected piece of writing with at least five sentences. We describe visible patterns; we do not diagnose from handwriting.
03
Words and made-up words
After one practice item, classify 20 age-appropriate letter strings as real or invented.
04
Speech sounds
Short picture tasks cover rhyme, syllables, first and last sounds, deletion, substitution and segmentation.
05
Visual speed and sequencing
A no-audio, RAN-inspired task measures fast visual search, sequencing, confusable symbols and short-term order memory.
06
Read aloud for one minute
Read an age-appropriate text. The system looks at pace, word accuracy, pauses, repetitions and self-corrections.
07
Sentence understanding
After one practice item, answer 10 short, age-appropriate true-or-false sentences.
Separate consent for the screen, research use and marketing emails
Results stay in a secure parent area and can be downloaded as PDF
Model-assisted interpretation is shown with uncertainty and raw module data
A future reading app can start from the same learner profile
BORN FROM A REAL FAMILY CHALLENGE
Three screenings. Still no clear answer.
Nora's son Lukas was visibly struggling with reading. Yet his dyslexia remained unrecognized across three well-intentioned screenings. The family received isolated scores and general feedback — but no clear picture of where the difficulties lay or what would actually help next.
This is precisely why Dyslexia99 was built: a screening that goes beyond a single total score to make individual reading component skills visible.
Our goal is to help families understand earlier which patterns stand out, where strengths lie, and when further support makes sense.
Parent experience Open research Responsible AI
The team behind the mission
Technology, research and lived family experience come together in one focused product.
MF
Michael Fink
Co-founder · Growth & OperationsBusiness builder focused on a clear consumer offer, responsible growth and turning a useful screening into a sustainable learning platform.
BS
Benjamin Schardt
Co-founder · Marketing & Demand GenerationDigital-marketing specialist focused on understandable communication, parent trust and a simple path from concern to action.
FF
Florian Felsing
Co-founder · CTOProduct-minded engineer responsible for reliable assessment flows, privacy-aware systems and a smooth experience across devices.
From uncertainty to a useful first picture
Families do not need another generic checklist. They need to see what happened in the actual tasks and what the pattern may mean.
Before
- You notice slow, effortful reading but are unsure how serious it is
- Free quizzes ask broad symptom questions without testing skills
- School feedback can arrive late or focus on one overall score
- A full assessment may be costly or take months to arrange
- You leave with concern but no clear next step
Dyslexia99
With Dyslexia99
- A guided, age-adjusted screen completed at home
- Direct tasks across words, sounds, visual speed, reading and sentences
- Observable handwriting and read-aloud patterns shown separately
- A detailed PDF with strengths, warning signs and uncertainty
- Clear guidance on when to monitor, support or seek professional assessment
Built for the questions parents actually have
The product stays simple on the surface while collecting the detail needed for a responsible interpretation.
Fast enough to finish
The core screen is designed to stay close to 10 minutes, with one practice item before each task type.
More than a symptom quiz
The child or adult performs real reading-related tasks instead of only answering whether certain problems sound familiar.
A profile, not a label
See performance by module, task quality and uncertainty. No single number is presented as a diagnosis.
Age-adjusted content
Routes for ages 7–9, 10–12, 13–17 and adults use the same structure with different words, texts and difficulty.
Useful after the result
The report explains sensible next steps and prepares a learner profile for the upcoming adaptive reading app.
Privacy without ad tech
The website uses no advertising trackers. App consent, research permission and marketing permission are kept separate.
Sample reportAlex · age 10
Several areas deserve a closer lookGood test quality
Word recognition
Speech sounds
Visual speed
Read aloud
Sentences
Methodology
What the screen measures
Dyslexia is not captured by one trick question. Dyslexia99 combines complementary tasks so that speed, accuracy and error patterns can be considered together.
- Writing sample: visible spelling and handwriting patterns such as word and sentence counts, spelling errors, phonetic plausibility, reversals, spacing, letter size, baseline, formations, corrections and omissions/additions.
- Word recognition: one practice item followed by 20 scored real-word and pseudoword decisions, selected from a planned pool of 100 items for each age route and language set.
- Phonological awareness: three scored items in each of seven areas — rhyme, syllables, first sound, last sound, phoneme deletion, sound substitution and segmentation — selected from a planned pool of 140 items per age route and language set.
- Visual speed and sequencing: a no-audio proxy task that captures search speed, order, confusable symbols and a small working-memory component.
- One-minute read-aloud: an age-appropriate text selected from a planned set of 20 per age route and language, with pace, decoding accuracy, errors, substitutions, pauses, repetitions and self-corrections analysed.
- Sentence understanding: one practice item and 10 scored statements selected from a planned pool of 100, with accuracy and response time tracked.
PDF report readyDetailed module data
Strengths Watch areas Next steps
Your report
Every result remains visible and explainable
The first release uses transparent task rules plus model-assisted interpretation. Until a validated statistical model is available, the report avoids claiming an exact clinical probability.
- Overall interpretation with a clear screening disclaimer and uncertainty level.
- Module-by-module values, response quality and the most relevant observed patterns.
- A downloadable PDF that can support a conversation with a school, tutor or qualified professional.
- A unique learner ID that can later connect the profile to the adaptive reading app without exposing the child's name in the code.
Parent Hub2 learner profiles
Alex · report ready
Mia · not started
Connect reading app
Parent area
A small parent hub, not another complicated portal
Parents need a safe place to manage profiles and reports, but the child-facing test should remain distraction-free.
- Create and manage learner profiles for ages 7 and up.
- See completed screens and download reports again.
- Review consent choices and request deletion.
- Connect the future reading app using a short-lived QR code or transfer code.
Separate consent Pseudonymous learner ID Export & deletion controls
Privacy & security
Sensitive learner data deserves a narrow purpose
The public website has no database, no forms and no advertising cookies. The screening app will collect only what is needed for the screen and the choices a parent actively makes.
- Screening consent, optional research use and marketing email permission are separate.
- A prior dyslexia status is required for context but is not used to inflate the task score.
- Identity data, assessment events and audio or image files should be stored separately with strict retention rules.
- No learner data is sold or used for unrelated advertising.
Accessibility
Low-stress, readable and mobile-first
A screening is only useful when the person can understand what to do and comfortably complete it.
- Large tap targets, strong contrast, plain language and generous spacing.
- Dyslexia-friendly reading presentation with larger type and line spacing during read-aloud.
- Audio and camera checks happen before timed tasks.
- Reduced-motion preferences are respected throughout the website.
Phone or tabletCamera + microphone
Quiet room Writing sample ready Audio check passed
Implementation
Designed for a real home, not a perfect lab
The screen gives clear setup instructions and marks results as uncertain when the conditions are not good enough.
- A quiet room, a recent writing sample, camera access and a working microphone are required.
- The timed core remains close to 10 minutes; setup and consent happen before it.
- Audio quality, interruptions, repeated instructions and device details are tracked as quality signals.
- The same task structure is used across ages; only the content and difficulty change.
DE-DEEN-USEN-GBES-ES
Bias & fairness
Language versions are built, not simply translated
Reading patterns differ by language, age and learning history. A single global score would hide those differences.
- Initial content sets: German, US English, UK English and Spanish for Spain.
- Canadian and Australian English do not need separate website languages at launch; separate norms are added only when data supports them.
- Every item stores its language, age route, version and difficulty so performance can be audited.
- Results should be checked across age, language, sex, multilingual background and device before strong claims are made.
SCREENING
≠
DIAGNOSIS
Versioned tasks and models
Research ethics
A screening estimate must stay a screening estimate
Dyslexia99 is built to support an earlier, better-informed next step — not to replace qualified diagnosis or create false certainty.
- No medical, psychological or educational diagnosis is provided.
- Handwriting features are described as observable patterns, not proof of dyslexia.
- The visual-speed module is described as RAN-inspired, not as a classic spoken RAN test.
- Model versions, task versions and quality flags are saved so results remain auditable.
Six areas. A research-backed holistic view.
Dyslexia99 combines established approaches from word recognition, phonological processing, reading fluency, sentence comprehension, and supplementary handwriting analysis into one guided screen. The specific combination and evaluation are continuously refined with real usage data.
Single-word recognition
Lexical decisions with real words and pseudowords are used in Stanford's validated ROAR-Word assessment.
Stanford ROAR-WordPhonological awareness
Sound matching and phoneme deletion are established early-reading measures and core parts of ROAR-Phoneme.
Stanford ROAR-PhonemeNaming speed and fluency
Large reviews and meta-analyses find a meaningful relationship between rapid naming and reading performance.
Araújo et al., 2015Sentence reading efficiency
Fast, accurate sentence understanding adds information beyond single-word reading alone.
Stanford ROAR-SentenceHandwriting as supporting context
Writing can reveal useful observable patterns, but handwriting alone is not specific enough to diagnose dyslexia.
Research contextResearch behind Dyslexia99
Our team publishes on machine learning, handwriting analysis, reading patterns, eye tracking, and explainable screening methods. These publications form the technological and scientific foundation upon which Dyslexia99 is developed.
Reported performance values relate to the datasets and study designs used in the respective research records.
Dyslexia — eye tracking, handwriting and linguistics
Explainable YOLO-Based Dyslexia Detection in Synthetic Handwriting Data
Computer-vision framework for automated, explainable handwriting error analysis.
View publicationA Dual-Model Handwriting Analysis Pipeline for Dyslexia Screening
Pipeline approach combining feature extraction with classical machine learning for risk scoring.
View publicationA High-Accuracy Machine-Learning Approach for Dyslexia Screening Based on Gamified Interaction Data
Evaluates Machine Learning models on gamified interaction data from children.
View publicationEnriching Dyslexia Detection with Synthetic Data
Data augmentation pipeline to improve detection robustly across small clinical datasets.
View publicationUnsupervised Eye-Tracking Dyslexia Detection Achieving 91.52% Accuracy on the ETDD70 Dataset
Unsupervised classification of eye movement patterns during reading tasks.
View publicationSynthetic Dyslexia Handwriting Dataset (YOLO-Format)
Open synthetic dataset for training and benchmarking handwriting error detection models.
View publicationWhy families no longer want to just wait and see
Parents consistently describe the same three situations.
“I can see my child struggling — but nobody can tell me how serious it is.”
Parents often describe children falling further behind while they wait for a school referral or clearer feedback.
“Months pass before a full evaluation is possible.”
Cost, insurance coverage, location and long waiting lists can delay a comprehensive professional evaluation.
“I need more than a generic online questionnaire.”
Families want to understand the pattern behind slow or inaccurate reading and what a sensible next step looks like.
Condensed themes from public parent and dyslexia communities; not customer testimonials.
Frequently asked questions
Clear answers about what the product can — and cannot — tell you.
The screen
No. It is an educational reading and dyslexia-risk screening. It cannot diagnose dyslexia or replace an assessment by a qualified professional.
The first product starts at age 7 and routes users into age groups 7–9, 10–12, 13–17 and adult. The structure stays consistent; words, texts and difficulty change.
The timed core is designed to remain close to 10 minutes. Account setup, consent, camera setup and choosing a writing sample happen before the core timer.
A recent, uncorrected text of at least five sentences can add useful context about spelling and visible writing patterns. The app reports observable features and does not diagnose from handwriting.
Tasks and accuracy
A technology check, writing sample, word and pseudoword decisions, seven areas of phonological awareness, a visual-speed and sequencing task, one minute of read-aloud, and sentence understanding.
No. It is RAN-inspired and requires no spoken response. It measures visual search, sequencing and confusable symbols, with a short order-memory component. The report labels it accurately.
The known text is compared with the recording to estimate reading pace, word accuracy, substitutions, omissions, pauses, repetitions and self-corrections. Poor audio quality can make a result unscorable.
The finished product must be validated on independent data before any strong accuracy claim is made. The first release shows task results, uncertainty and a model-assisted screening interpretation rather than a clinical probability.
Languages and age groups
German, US English, UK English and Spanish for Spain. The public website is available in English, German and Spanish.
Not as separate website languages at launch. The interface can remain English, while pronunciation, item suitability and norms should be checked separately before region-specific scoring claims are made.
No. Word frequency, pseudowords, speech sounds and reading fluency differ by language. Each language and age route needs its own item bank, audio and validation.
Yes, but results must be interpreted in light of the person's main language of instruction and reading experience. Limited exposure to the test language can look like a reading difficulty.
Results, privacy and price
One complete screening, a parent dashboard, detailed module results and a downloadable PDF report. Market currency is shown before purchase.
Age, grade/year or education level, sex with an option not to say, language information, a required choice of diagnosed, suspected or not known, and the consents needed for the screening. Family history is not required.
No. Marketing permission is separate and optional. It is not required to purchase or complete the screen.
It keeps learner profiles, reports, consent settings and deletion controls in one place. It also allows a secure connection to the future reading app.
What happens next
Use the report as a structured starting point. Depending on the pattern and the person's day-to-day difficulties, the next step may be targeted reading support, a conversation with school, or a comprehensive assessment by a qualified professional.
The PDF can support a discussion, but schools and authorities decide what evidence they accept. Dyslexia99 does not promise eligibility, accommodations or a formal diagnosis.
A secure learner profile can set the starting level and priorities for adaptive practice. The planned reading app will use AI for personalised support while keeping the learning sequence structured and reviewable.
The parent area is designed to provide access, export and deletion controls. Exact retention periods and processing partners will be shown before the screening launches.
Stop guessing. Get a clearer picture.
Check six reading-related areas, get results immediately upon completion, and download a detailed report with concrete next steps.
Screening only — not a diagnosis.