About Nomto
Nomto is an online platform for preparing for language and admission exams, with speaking and writing checked by AI against each exam’s public criteria. This page explains who runs it, how answers are scored and how far the scores can be trusted.
Last updated:
What is Nomto?
Nomto (nomto.app) is an online exam-prep hub. Each course has practice tasks in the format of the real exam, timed mock tests and feedback on every answer: a score on the exam’s scale, notes on each criterion and corrections of specific mistakes. Practice happens in the browser at app.nomto.app, on a computer or a phone, with no download.
Speaking is practised out loud: the learner records an answer, or talks with an AI examiner in the IELTS Speaking test, and gets a score and feedback on that recording. Writing is typed and scored the same way. Reading, Listening and quantitative questions are marked against an answer key.
Who runs Nomto?
Nomto is run by Alexander Saranov, registered in Russia as an individual entrepreneur: Individual Entrepreneur Alexander Sergeevich Saranov, INN 081702723412, OGRNIP 323080000018850, Moscow, Russian Federation. He is also the seller of Nomto passes; payments are processed by Robokassa (Robokassa LLC, Russia).
Nomto is an independent project. It is not affiliated with, endorsed by or connected to any exam owner. Questions about the service, the scoring or your data: saranov.alexander@gmail.com.
Which exams does Nomto prepare for?
Nomto has twelve courses. Each one opens with a free starter part.
- IELTS: Speaking with a live AI examiner, Writing Tasks 1 and 2 (Academic and General Training), Reading and Listening practice tests.
- TOEFL iBT: the 2026 format: all 12 task types, rubric scores and adaptive mock tests.
- PTE Academic: all 22 task types in the exam’s timing and a full mock test with a four-skill report.
- GMAT Focus Edition: Quantitative, Verbal and Data Insights: 1,951 original questions with step-by-step explanations, lessons, an adaptive mock test and an AI tutor.
- Digital SAT: Reading and Writing and Math: 561 original questions with step-by-step explanations, practice by skill, an adaptive mock test on the 400–1600 scale and an AI tutor.
- GRE General Test: Verbal, Quantitative and the Analyze an Issue essay: 505 original questions with step-by-step explanations, an adaptive mock test on the 130–170 scale, essay feedback and an AI tutor.
- Duolingo English Test: every question type, with speaking and writing checked by AI and mock tests on the 10–160 scale.
- CEFR Multilevel: Uzbekistan’s national English exam: all 17 parts and mock tests on the 0–75 scale.
- VSTEP: Vietnam’s national English test for levels B1–C1: all four skills, with speaking and writing checked by AI and mock tests on the 0–10 scale.
- HSK: Chinese from the first lesson: pinyin, tones, characters, conversation with an AI partner and mock papers in the new HSK format.
- Spoken English: conversation practice with an AI partner who corrects every line and suggests a better way to say it.
- SQL: SQL for data analysis: 82 lessons and 248 tasks on a real airline database, checked in a sandbox, with an AI tutor.
How does the AI score speaking and writing?
A spoken answer is first turned into text by a speech-recognition model. A large language model then rates that text against the exam’s public criteria, such as the IELTS band descriptors or the TOEFL 0–5 rubrics, and an audio model listens to the recording itself for pronunciation. Writing goes straight to the language model with the same criteria. Anything with one right answer, including Reading, Listening, GMAT and SQL, is marked by code against the key, with no AI involved.
The models are OpenAI’s (the GPT-4o family for scoring, gpt-4o-mini-transcribe for speech recognition), called through the API. If OpenAI is unavailable, the same models are reached through OpenRouter, and speech recognition switches to ElevenLabs Scribe or Google Gemini.
Code checks what code can measure exactly. It counts words and caps answers that are far too short, gives 0 to blank or copied answers, measures pauses from the sound wave for fluency (PTE, TOEFL, Duolingo, Multilevel, VSTEP) and compares a read-aloud or repeated sentence with the original word by word.
| Exam | What is checked | Scale shown |
|---|---|---|
| IELTS | Speaking: Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, Pronunciation (from the recording). Writing: Task Achievement or Task Response, Coherence and Cohesion, Lexical Resource, Grammatical Range and Accuracy. Reading and Listening: against the key. | Bands 0–9 in half bands |
| TOEFL iBT | Speaking and Writing tasks on the ETS 0–5 rubrics (for example elaboration, language use, intelligibility). Reading, Listening and Build a Sentence: against the key. | 1–6 in half bands and a comparable 0–120 total |
| PTE Academic | Speaking: content, oral fluency, pronunciation. Writing: content, form, development, linguistic range, grammar, vocabulary. Reading and Listening: against the key by the Score Guide’s rules. | 10–90 |
| Duolingo English Test | Speaking: content, coherence, grammar, vocabulary, fluency, pronunciation. Writing: content, coherence, grammar, vocabulary. Keyed tasks: against the key. | 10–160 in steps of 5, with a CEFR level |
| CEFR Multilevel | Speaking: task, coherence, vocabulary, grammar, fluency, pronunciation. Writing: task achievement, organisation, vocabulary, grammar. Reading and Listening: against the key. | 0–75 with a CEFR level |
| VSTEP | Speaking: grammar, vocabulary, pronunciation, fluency, discourse management. Writing: task fulfilment, organisation, vocabulary, grammar. Listening and Reading: against the key. | 0–10 per skill; the average gives B1, B2 or C1 |
| GMAT Focus | Every question against a verified key; the AI tutor explains but does not mark. | 60–90 per section, 205–805 total, shown with a range |
| Digital SAT | Reading and Writing and Math questions against a verified key; the AI tutor explains but does not mark. | 200–800 per section, 400–1600 total, shown with a range |
| GRE | Verbal and Quantitative against a verified key. Analytical Writing: the Analyze an Issue essay on the ETS 0–6 holistic rubric. | Verbal and Quant 130–170, Writing 0–6 in half points |
| HSK | Mock papers against the key; the picture-sentence task by a language model. Tones are judged from the pitch of the recording. | Points of the paper |
| SQL | The query runs on a real database (DuckDB) and its result is compared with the expected one. | Solved or not, with a hint |
Spoken English has no exam score: the AI partner corrects each line and explains the mistake instead.
How accurate are the scores?
Nomto scores are practice estimates, not official results. Only the exam owners can give an official score, and they use their own trained examiners, scoring engines and score equating that cannot be reproduced from outside. Nomto maps raw points onto each exam’s scale through anchor points from the owners’ published materials, and the practice room always marks the result as approximate.
We test the scorer against human ratings. For PTE speaking it was compared with expert ratings of 32 recordings from the public speechocean762 dataset: oral fluency correlated with the experts at r = 0.83 and pronunciation at r = 0.77–0.79, with the scorer slightly strict rather than generous. Pronunciation is counted 0.2 points lower to offset a small remaining lean. In two runs on the same recordings, 29 of 32 got identical scores and the rest differed by one point on the 0–5 scale (check of 22 October 2026).
Known limits:
- Speech recognition can quietly turn a mispronounced word into the right one, so pronunciation is judged from the recording by an audio model. If that check fails, the IELTS Pronunciation band is estimated from the transcript and marked as tentative.
- A language model can give a slightly different score to the same answer on another run. Treat a difference of half an IELTS band, or one point on a 0–5 rubric, as noise.
- Background noise, a poor microphone or a very short answer lower the reliability of a speaking score.
- Scores are for practice and planning. They are not a certificate and should not be presented to a university, employer or immigration authority.
IELTS is owned by the British Council, IDP: IELTS Australia and Cambridge University Press & Assessment; TOEFL and the GRE by ETS; PTE Academic by Pearson; GMAT by GMAC; the SAT by College Board; the Duolingo English Test by Duolingo; HSK by Chinese Testing International; the Multilevel exam is run by Uzbekistan’s state testing authorities and VSTEP by Vietnam’s Ministry of Education and Training. Nomto is not affiliated with or endorsed by any of them.
Where do the practice tasks come from?
All practice tasks are original, written for Nomto in the format of each exam. None is copied from an official test or a paid book. The formats, timings and criteria follow the owners’ public documents, such as the IELTS band descriptors, the PTE Score Guide, the TOEFL Technical Manual, the Duolingo Scoring Guide and GMAC’s score tables.
Tasks are drafted with the help of AI models to a detailed brief and then checked twice. An automatic validator checks the format, length, answer key and translations; for GMAT, SAT and GRE maths a Python script recalculates every answer. Then a blind check: a separate solver answers each item without seeing the key, and an item goes live only when both answers agree. Listening recordings are voiced with text-to-speech.
At the time of writing the banks hold 1,951 GMAT questions, 561 Digital SAT questions, 505 GRE questions, 1,280 TOEFL tasks, 1,120 PTE tasks, 746 Duolingo English Test tasks, 487 Multilevel tasks and 240 VSTEP tasks that passed the blind check.
Which languages is Nomto available in?
The site, the practice room and the explanations are in English, Russian and Spanish. Some courses also speak other languages: IELTS in Portuguese, Vietnamese, Turkish, Indonesian, Arabic and Uzbek; TOEFL and Spoken English in Portuguese, Vietnamese, Turkish and Indonesian; the Duolingo English Test in Portuguese and Indonesian; GMAT in Portuguese; the Digital SAT in Uzbek, Vietnamese, Indonesian and Turkish; the GRE in Portuguese; CEFR Multilevel in Uzbek; VSTEP in Vietnamese; HSK in Vietnamese; SQL in Chinese, Portuguese and Vietnamese.
The exam material itself stays in the language of the exam: English for the English tests, GMAT, SAT and GRE, Chinese for HSK. Feedback and explanations come in the learner’s language.
How much does Nomto cost?
A free starter part of every course is open to any account; one paid pass opens all courses for a fixed period, as a single payment with no subscription. Current prices are on the pricing page.
How to contact Nomto?
Email saranov.alexander@gmail.com. Seller details are on the contacts page; how personal data is handled is in the privacy policy.