pteexamprep
Start free
Free practiceMy progressScore toolsCountries & visasExam guidesPTE score chartPTE to IELTS

Methodology

How we find and check facts, how often we re-check them, and what our practice scores can and cannot tell you.
  • Updated 10 October 2026

How facts are sourced

  • Official sources only.Fees, test formats, results timelines and score concordances come from Pearson's own pages and documents. Visa and immigration requirements come from government sites such as the Australian Department of Home Affairs, GOV.UK, Immigration New Zealand and the Government of Canada. University requirements come from the university's own pages.
  • Sources are recorded.Each guide lists the official pages it relied on in an “Official sources” box, each with a “checked” date.
  • Original wording. We paraphrase and link rather than copy.
  • If we cannot confirm it, it stays out or is clearly marked as unverified.

What the dates on a page mean

  • Published: when the page first went live.
  • Updated: when its content last materially changed.
  • Checked: the date a source was last compared with the page. A page can be checked and found still correct without changing, so the checked date can be newer than the updated date.

How often we re-check

Pages with figures that change, namely test fees, visa and country score requirements, and test dates and availability, are re-checked against their official sources about once a month. Other guides, such as task formats, are re-checked when we learn that Pearson has published a change, and on a slower cycle. A monthly cycle is a target, not a promise: a rule can change between checks, so confirm anything you are about to pay for or submit on the official page.

Corrections

When we find or are told about an error, we fix it in the page, update its date and, for a material change, say what changed. See the editorial policy and the contact page.

How practice scoring works and its limits

Our practice questions are original, and the scores shown after them are practice estimates, not official PTE scores. We do not replicate Pearson's scoring and do not claim to. Only the official test provider issues an official score. Each result page says which of the three methods below produced it.

1. Rule-based marking

Used for tasks with a checkable answer: reading and listening fill in the blanks, reorder paragraphs, the multiple-choice tasks, Select Missing Word, Highlight Correct Summary, Highlight Incorrect Words, Write From Dictation and Answer Short Question. The per-question rules come from Pearson's published PTE Academic Test Taker Score Guide (checked 10 October 2026), and the same answer always gives the same result:

  • Fill in the blanks: 1 point per correct blank (or correctly spelled word), minimum 0.
  • Reorder Paragraphs: 1 point per correct adjacent pair, minimum 0.
  • Multiple choice, multiple answers: +1 per correct option, -1 per incorrect option, minimum 0.
  • Multiple choice single answer, Select Missing Word, Highlight Correct Summary: 1 or 0.
  • Highlight Incorrect Words: +1 per correct word, -1 per wrong highlight, minimum 0.
  • Write From Dictation: 1 point per correct, correctly spelled word. Capital letters and punctuation are not checked and word order is ignored. Pearson does not publish how much extra words cost, so we do not penalise them.
  • Answer Short Question: 1 or 0 against a short list of accepted answers, so a correct answer worded differently can be marked wrong. If the answer was read from a speech transcript, it is labelled a speech estimate instead.

What we do not know: Pearson does not publish how item points are weighted into the 10-90 skill scores. That is why the percentage we show is a practice figure for the set you just did and nothing more.

2. Speech estimates from an optional transcript

Used for Read Aloud, Repeat Sentence, Describe Image, Retell Lecture, Summarize Group Discussion and Respond to a Situation. The transcript is off unless you switch it on, and there are two kinds. The on-device transcript (recommended) runs a small Whisper speech model inside your browser after you stop recording; it works in most recent browsers that support Web Workers, WebAssembly and recording, and needs a one-time model download. The browser transcriptuses the browser's own speech recognition and works in Chrome and Edge only. Either way, the speech is turned into text and we measure the text:

  • Read Aloud: the transcript is aligned with the passage word by word. Each replaced, omitted or inserted word counts as one error, as in the Score Guide.
  • Repeat Sentence:content uses Pearson's published bands (3 for all words in the correct sequence, 2 for at least half of them, 1 for under half, 0 for almost nothing). We read “almost nothing” as under 20% of the words, and the headline percentage inside bands 1 and 2 is our own blend.
  • Describe Image, Retell Lecture, Summarize Group Discussion, Respond to a Situation: each question has a fixed list of key points, and we look for their keywords in the transcript. A good idea in other words can be missed. Reading the given text back, or listing topic words without sentences, is scaled down. In the real test a human reviews the content of these tasks; a keyword match cannot do that.
  • Fluencyis a rough estimate of speaking pace and repeated or filler words, plus long pauses (over 1.5 seconds between words) when word timing is available. Only the on-device transcript supplies word timing, so with the browser transcript only pace is estimated and pauses are not measured. The thresholds are our own, not Pearson's. If no timing is known, fluency is shown as not measured and the headline is content only.
  • The headline combines content with fluency, so strong content with the weakest fluency estimate still earns 70% of the content figure and fluency cannot rescue empty content.

Pronunciation is never measured.A transcript cannot judge vowels, consonants, stress or accent, so every speaking result shows pronunciation as “Not measured”. A good transcript match does not mean your pronunciation would score well. A recogniser can also mishear a word you said correctly, and microphone, accent and background noise all affect it. The on-device model can also quietly tidy up a slip or a repeated word, so for Read Aloud and Repeat Sentence the content estimate can run generous.

Self-checked mode (no transcript)

Without a transcript, Describe Image, Retell Lecture, Summarize Group Discussion and Respond to a Situation let you tick the key points you think you covered after listening back. The result is labelled self-checked: the ticks are your own judgement, pace is not measured, and nothing is detected from your voice. For Read Aloud, Repeat Sentence and Answer Short Question a content estimate needs the transcript, so without one you get a listen-back-and-compare view and no score.

3. Heuristic writing checks

Used for Summarize Written Text, Summarize Spoken Text and Write Essay. All checks are deterministic text checks that run in your browser. Nothing is sent to a model or to our servers.

  • Form follows the Score Guide limits: Summarize Written Text is one sentence of 5-75 words; Summarize Spoken Text is 50-70 words (40-49 and 71-100 earn partial credit); Write Essay is 200-300 words (120-199 and 301-380 earn partial credit). As the guide says, a Form or Content score of 0 gives no points, and we do the same.
  • Contentis judged by keyword matches against fixed key points for the two summary tasks. For essays it is the share of the prompt's main words you use, scaled down when few different content words are used. It cannot tell whether your argument is clear or convincing.
  • Our own simple checks:vocabulary variety, linking words, sentence length and variety, repeated words and phrases, capital letters and punctuation, and essay structure (paragraphs, a conclusion). The thresholds are ours, not Pearson's.
  • Grammar and spelling are only partly checked: a dozen specific grammar slips and a list of frequent misspellings. Most errors will not be found, and a clean result is not a sign of correct English.
  • Copying: if a long run copied from the passage, transcript or prompt makes up 40% or more of an answer, the headline is capped. The cap is our own rule.

In the real test these tasks are scored by AI with human review of content. Our checks are neither.

What is not measured

  • Pronunciation is never measured, and oral fluency is only a rough pace estimate. Intonation and accent are not assessed at all.
  • The depth of your content, the quality of your argument, coherence, tone and the naturalness of your language are not assessed by our writing checks.
  • We do not convert any practice percentage to a 10-90 score, and we do not claim an accuracy figure for any estimate. The progress page groups your averages into plain labels such as “Building” or “Solid” that describe your results on this site's practice sets only.

Estimates are not your exam score

A practice estimate should not be used to predict your PTE result, to choose a test date on its own, or to show anyone as evidence of English level. Use it to notice patterns, such as which task types need work, and to build habits. For an official indication, use Pearson's own practice materials, and for visa or admission decisions rely on an official score report. Your results are saved only in your browser (see the privacy policy). See also our terms of use.

Conversion tables

Score converters and charts, such as PTE to IELTS, use the concordance published by Pearson and say which table and date they use. Pearson notes that concordances carry measurement error, and so do our converters; the score your university or visa authority names is the one that counts.