Gurully's Mock Test Accuracy for PTE & Score Gains

Overview

The Question This Report Answers

Every test-prep platform claims its mock tests are "real" and "accurate." The harder question is whether the practice actually changes a candidate's outcome — and whether the score a student sees in a Gurully mock is a trustworthy, exam-aligned signal of where they really stand.

This report answers both using the platform's own data: 80,440 students and more than one million AI-scored responses across 91 countries. The evidence falls into two parts — how Gurully produces an exam-aligned score, and what happens to candidates' scores as they practice.

Two-Part Evidence

Score alignment + outcome improvement

How Gurully produces an exam-aligned score
What happens to scores as candidates practice

Note On Scope

This analysis is built entirely from on-platform practice data. It measures how Gurully scores a mock and how scores move with practice — not a statistical comparison against official exam results. Where that external outcome data becomes available, it can be added as a third pillar of evidence.

Scoring Methodology

How Gurully Produces an Exam-Aligned Score

The accuracy of a mock starts with how it is scored. Gurully scores every response with AI models built around the official PTE scoring framework, and three properties make those scores dependable.

01

Official PTE 10–90 Scale

Every section and overall result is reported on the same band scale candidates see on the real exam — so a mock score maps directly to the target a university or visa pathway requires.

02

Trained on Real Behavior at Scale

The scoring models draw on 1,013,351 genuine responses from learners in 91 first-language and regional contexts — not a small lab sample — scoring the full range of accents, fluency levels, and answer patterns.

03

Calibrated Per Question Type

Difficulty is modelled task by task, normalised to the percentage of each task's maximum score — so the engine distinguishes a genuinely hard task from one that simply carries a low raw score.

The granularity of that calibration is visible in how differently the engine scores each task type — from the hardest to the easiest — across hundreds of thousands of responses:

Question Type	Responses	Avg % of Max
Re-tell Lecture (hardest)	28,139	50.7%
Summarize Written Text	30,278	~57%
Write Essay	13,217	57.9%
Repeat Sentence	169,056	~62%
Describe Image	80,062	~69%
Read Aloud (easiest)	104,697	~73%

Why This Matters

A score is only useful if it behaves like the real exam

Because Gurully scores on the official band scale, models real response behaviour at scale, and calibrates each task independently, the number a candidate sees in a mock is a realistic read of exam-day readiness — not a flattering estimate.

Practice Outcomes

The Evidence That Practice Changes Scores

Evidence A

Multiple attempts are worth about +5.7 points

Among the 16,860 students who completed two or more full mocks, the average jump from first attempt to best attempt is measurable and consistent:

+5.7 Avg Score Gain

55.5 First Mock (Avg)

61.1 Best Mock (Avg)

+5.7 pts Average Gain

A ~6-point lift is frequently the exact margin between a 59 and a 65 — the difference between missing and meeting a university or immigration requirement.

Evidence B

More practice more than doubles the pass rate

Grouping students by the number of full mocks completed shows pass rates rising with volume. Those who take five or more mocks reach 65+ at more than twice the rate of those who take just one.

2.1× Higher Pass Rate

Mock Taken	Students	Avg Best Score	% Reaching 65+
1	45,970	54.2	21.1%
2	4,680	59.2	33.1%
2	3,240	60.7	35.8%
3	1,890	60.6	32.8%
5 or more	7,050	62.8	44.3%

A ~6 -point lift is frequently the exact margin between a 59 and a 65 — the difference between missing and meeting a university or immigration requirement.

Evidence C

The mock reliably diagnoses the real bottleneck

Across every score band below 79, Gurully's scoring identifies Reading as the weakest section, and the act of solving Reading is what moves a candidate up a band.

144K Attempts Analysed

SCORE LEVEL	SPEAKING	WRITING	READING	LISTENING	WEAKEST
~50 (45–54)	54.1	51.7	46.4	48.9	Reading
~65 (60–69)	65.6	62.8	61.6	65.4	Reading
~79 (75–82)	76.5	74.5	77.5	77.9	Writing

Reading is the consistent anchor at 50 and 65, then — once conquered — Writing emerges as the final gap at 79. A mock producing random scores could not reproduce this structure across 144,420 attempts.

The Bottom Line

Exam-aligned by design. Demonstrably effective in practice.

+5.7 pts Score lift on retaking

2.1× Pass rate among 5+ mock takers

Stable Consistent weakness diagnosis

Accuracy and Roadmap

What Accuracy Means Here and What Comes Next

The claims in this report rest on what the data can prove directly: that Gurully scores on the official PTE band scale, that those scores are produced by models calibrated on more than a million real responses, and that scores rise measurably and repeatably as candidates practise.

The strongest possible proof of mock accuracy is a direct comparison of each candidates predicted mock score against their official PTE result. That external outcome data is not part of this dataset. Collecting verified exam results would let Gurully report a precise predicted-versus-actual accuracy figure and add a definitive third pillar to the evidence. Until then, the exam-aligned scoring methodology and the demonstrated score gains are the honest, defensible basis for the accuracy claims.

What we can Claim

Scores on the official PTE 10-90 band scale
Models calibrated on 1M+ real responses
Measurable repeatable score gains with practice

What Comes Next

Collecting verified exam results through post-exam surveys would let Gurully report a precise predicted-versus-actual accuracy figure adding a definitive third pillar to the evidence.

Methodology and Notes

Methodology and Data Notes June 2026

Findings are derived from real student activity on the Gurully PTE Academic platform: 80,440 students and 1,013,351 AI-scored responses across 91 countries.
Scores are reported on the PTE 10-90 scale. Retake and practice-volume analyses are based on the 16,860 students who completed two or more full mocks.
Question-type difficulty is normalised to the percentage of each tasks maximum possible score. Figures marked with tilde are approximate.
This report does not compare mock scores against students official PTE exam results. No such external outcome data is included in the dataset.

Data sourced from Gurully PTE Academic platform June 2026 gurully.com

Gurully's PTE Mock Test Accuracy & Proven Score Gains

The Question This Report Answers

How Gurully Produces an Exam-Aligned Score

The Evidence That Practice Changes Scores

What Accuracy Means Here and What Comes Next

Take a Free Mock Test Now!