How we evaluate your PTE test
Every response you submit is scored against the same things the real PTE measures, and we are open about exactly how. Reading and Listening are scored instantly from the answer key. Speaking and Writing are scored by a pipeline that listens to your actual voice and reads your actual words, then maps the result onto the official 10 to 90 scale with a full enabling-skills breakdown. This page explains each step, why it is trustworthy, and where its limits are.
Why PTE scoring needs more than a transcript
The Speaking and Writing sections of the PTE are not simple right-or-wrong answers. The exam scores how you spoke and how you wrote, not just whether your words were correct. Most free practice tools either ignore your speaking or transcribe it and guess a score from the words. Neither tells you the thing that actually decides your speaking band: your delivery. Two people can say the same correct sentence and score very differently because one was clear and well-stressed and the other was flat, mumbled or halting. We built our pipeline to measure the same signals the real exam does, from your recorded audio.
The scoring pipeline, step by step
1Take the test under real conditions
You complete a full PTE Academic or PTE Core mock, or a single-skill practice test, in an interface that mirrors the real Pearson layout, timing and tools. Reading and Listening selections are captured for instant scoring. Speaking recordings and Writing responses are captured for the speech and language engines.
2Acoustic analysis of your voice
Your recorded audio is analysed acoustically by a dedicated speech engine (Microsoft Azure Speech). It measures pronunciation accuracy of every word down to the individual phoneme, your oral fluency (pace, pausing and hesitation), and your prosody (rhythm, stress and intonation, including flat or monotone delivery). This is a measurement of how you actually sounded, not a guess from a transcript.
3Transcription of what you said
In parallel, your audio is transcribed (using Deepgram) so the engine has an accurate record of the words you produced. This feeds content scoring for tasks like Describe Image and Re-tell Lecture, where saying the right things matters as much as saying them well.
4Content, grammar and coherence scoring
A large language model assesses content, grammar, vocabulary, spelling and written discourse against the official PTE criteria, for both your spoken transcripts and your written responses. It checks whether you covered the required points, stayed inside the word count, and organised your answer clearly.
5Blend into the 10 to 90 scale
The acoustic signals and the language signals are combined and mapped onto the 10 to 90 PTE scale, producing an overall score, four communicative-skill scores, and a breakdown of the six enabling skills. Your report then shows a word-level pronunciation heatmap, your weakest sounds, and a targeted list of words to practice.
What each skill is scored on
Speaking
Every speaking task is scored on three things, exactly as the real exam does it: content (did you say the right things), oral fluency (was your delivery smooth and well-paced), and pronunciation (were your sounds clear and intelligible). The engine is not listening for a "perfect accent"; it checks that your sounds are clear, your stress and rhythm are natural, and you speak at a steady pace without long hesitations.
Writing
Writing tasks are scored on content, form (length and format), grammar, vocabulary, spelling and written discourse. Staying inside the required word count matters: going over or under is penalised even if the writing is otherwise strong.
Reading and Listening
These are largely objective. Correct-answer item types add to your score, and some types (such as Highlight Incorrect Words or Fill in the Blanks) use partial credit: you gain marks for each correct selection and can lose marks for wrong ones. Because they are objective, we score them instantly when you finish, with a full per-question breakdown.
The six enabling skills you get back
Like the real PTE, your report breaks performance down into the enabling skills that sit beneath the four communicative skills, so you can see exactly where marks are leaking:
- Pronunciation, how clearly and accurately you produce English sounds.
- Oral Fluency, how smooth, natural and well-paced your speech is.
- Grammar, control of sentence structure in speaking and writing.
- Vocabulary, range and accuracy of word choice.
- Spelling, accuracy in written responses.
- Written Discourse, how well your writing is organised and developed.
How to read your score report
- Start with the enabling skills, not just the overall number. A single weak enabling skill, such as pronunciation, quietly drags down several task scores at once. Fixing it lifts more than one skill.
- Use the pronunciation heatmap. The colour-coded, word-by-word view shows the exact sounds and words that cost you marks, which is far more useful than a single speaking number.
- Treat the overall score as a band, not a precise prediction. Read it as "around this level," then work on the specific weak spots the report flags.
- Re-test after targeted practice. The fastest way to improve is to fix the flagged sounds and word-count habits, then run another timed mock.
Limitations: what this is and is not
We want you to trust our scores, which means being honest about their limits:
- It is an AI estimate for practice. Our score is built on the same enabling skills the real PTE measures and correlates well with real PTE banding, but it is not an official Pearson result. Use it to guide preparation, not to guarantee a final number.
- Audio quality affects acoustic scoring. A noisy room or a poor microphone can lower your measured pronunciation and fluency. Record in a quiet space with a decent headset for the most representative result.
- Automated content scoring is very good but not infallible. On unusual or highly creative answers, an LLM-based content score can differ from a human rater's judgement. If you think a content score is clearly wrong, you can flag it (see our editorial standards).
- We do not replicate Pearson's exact proprietary algorithm. No third party has it. We replicate the signals it scores, acoustic delivery and language quality, which is what makes practice here useful.
Frequently asked questions
Is the PTEMock score the same as my official PTE score?
No. It is an AI estimate for practice, well-correlated with real PTE banding but not an official Pearson result.
How does PTEMock score my speaking?
Your audio is analysed acoustically for pronunciation, fluency and prosody with Azure Speech, transcribed with Deepgram, and assessed for content, grammar and coherence by a large language model, then blended into the 10 to 90 scale.
Why score my audio instead of my transcript?
Most of your speaking score comes from how you sound, not just the words. A transcript cannot measure pronunciation, fluency or intonation, so we analyse the actual recording.
How are reading and listening scored?
Instantly and deterministically from the answer key, including partial credit where the item type uses it, with a full per-question breakdown.
See your voice scored for free
Take a free PTE mock test and get a word-level pronunciation, fluency and emphasis report on your real voice, the same enabling skills the real exam measures.
Start a free mock test →