How We Score Your TOEFL Practice

Correct answers earn one point. Your first-module accuracy chooses the second module. We combine correct answers across the selected modules, calculate round(correct ÷ total × 30), apply the easy-route cap when relevant, then look up the section band in the ETS comparison table.

Meeting the threshold selects the hard route; a lower first-module accuracy selects the easy route. This is our published practice rule. ETS uses its own calibrated adaptive scoring process, which this site does not reproduce.

Build a Sentence earns one point only when the entire arrangement is correct; there is no partial credit. Email and Academic Discussion receive an AI estimate from 0 to 5 for task completion, organization and language use, with strengths, improvements and corrections. Written responses are evaluated through Speko Router.

All applicable caps are enforced; the lowest cap wins. For this practice format, writing raw points = Build a Sentence points + Email × 2 + Academic Discussion × 2. The set contains 10 sentence tasks, 1 email task and 1 discussion task, giving 30 available points. The section estimate is round(raw ÷ 30 × 30).

The Zhiyan speech engine supplies pronunciation, fluency and language signals. The tables below show how our rules convert those signals into an estimated 0–5 task score. They are practice thresholds, not an official ETS rubric or a calibrated prediction.

Listen and Repeat

Completeness is the number of correctly matched reference words with a pronunciation score of at least 60, divided by the number of reference words, capped at 100%. Higher task levels also require an overall pronunciation threshold. Fluency is shown as feedback; it does not independently change the task level. We use the highest qualifying row.

Missed words reduce completeness. Tips can prioritize a missing second half, unclear vowels or excessive speed. Try Listen and Repeat practice.

Take an Interview

For the upper levels, all conditions in a row must be met. Duration means detected speaking time divided by the task’s response window, capped at 100%. Word count comes from the recognized transcript. Pronunciation, fluency, grammar, vocabulary and relevance are also shown separately.

The “say more” suggestion takes priority when speaking time is below 60% of the window or the transcript has fewer than 60 words. A short, fluent answer can therefore receive a low task score. We show at most three grammar corrections. Practice an interview question.

The complete speaking section has 11 tasks worth up to 5 points each. Speaking 0–30 = round(raw ÷ 55 × 30), followed by the ETS section-band lookup. A single drill receives a task score only; it is not a complete section band.

We average the four section bands and round to the nearest 0.5. Our 0–120 practice score is the sum of the four 0–30 section estimates. Because section lookups and overall rounding are separate, that sum can fall outside the ETS comparison range for the overall band.

Use the score calculator to inspect the complete ETS conversion tables, examples and linked sources. These are estimates; universities set their own requirements.

Our questions are original practice material. They have not been calibrated by ETS, and TOEFLScorer is not endorsed or approved by ETS. Performance on our items may differ from performance on the official test. The task levels are initial, transparent rules that may be calibrated when suitable evidence becomes available.

AI writing evaluations can vary between runs. Speech recognition can mishear words; background noise, microphone placement and recording devices affect the signals. Review the feedback and your recordings alongside the score. Use official test results for admissions decisions.

We plan to publish calibration findings when we can support them with appropriate research and consent. No publication date or accuracy guarantee is promised.

We use answers and recordings to evaluate practice and provide private feedback. We do not use them to train AI models; evaluation providers also operate under their applicable terms. The homepage sentence demo does not save a replay recording on our service. Full-test, section-practice and speaking-drill recordings are stored privately for up to 30 days. Browser copies remain until you clear site storage.

Answers, scores and feedback remain in practice history under the retention and deletion rules in our privacy policy. Share links expose score summaries, never detailed answers or recordings.

2026-10-17 — Published this guide and exposed the existing scoring thresholds as shared constants. Scoring behavior is unchanged.