PTE Speaking Tips 2026: Seven Tasks, Seven Mistakes to Avoid (Updated for the New Format)

Why the Speaking section decides your PTE score

Part 1 of PTE Academic (Speaking & Writing) is the longest section at 76 to 84 minutes, and speaking tasks dominate it. Of the nine task types in Part 1, seven involve your microphone. Three of those seven also feed into your Listening score (Repeat Sentence, Retell Lecture and Summarize Group Discussion), so a weak speaking performance drags down two skills at once.

After the August 2025 update, PTE uses hybrid scoring: AI still grades every response, but human reviewers now check the content of four speaking tasks (Describe Image, Retell Lecture, Respond to a Situation and Summarize Group Discussion). That means memorised templates are detected and penalised. The only path to a high speaking score in 2026 is genuine fluency, clear pronunciation and relevant content.

The good news: speaking is also the most improvable skill. Unlike reading comprehension, which builds slowly, fluency and pronunciation respond quickly to targeted daily practice. The tips below are organised task by task so you can focus on exactly where you are losing marks.

PTE speaking tasks list: Pearson's official order and what task 7 is

Pearson's Speaking & Writing test-format page lists the speaking question types in the order below.

So if someone asks about “speaking task 7” in PTE, on Pearson's list that is Respond to a Situation. This guide covers the two new tasks the other way round (Respond to a Situation as Task 6 and Summarize Group Discussion as Task 7), so check which list a teacher or video is using.

The Personal Introduction is not scored. Pearson lists a Personal Introduction before Read Aloud and says it does not contribute to your score: it is there to get you familiar with the test technology. You get 25 seconds to read the prompt and prepare, then 30 seconds to record, and you can record it only once. Treat it as a microphone rehearsal and speak at the volume and pace you plan to use for the whole section.

How PTE speaking is scored in 2026

Before diving into task-specific tips, you need to understand the three traits the AI evaluates on every speaking response:

Critical rule: if Content = 0 (you said nothing relevant, or the response was flagged as a memorised template), the entire item scores zero. Oral Fluency and Pronunciation are not even evaluated. This is the single most important scoring rule in PTE speaking.

Pronunciation and Oral Fluency are always AI-only. No human ever listens to judge your accent or rhythm. The AI compares your speech patterns against a model trained on thousands of native and high-proficiency speakers across accents (British, American, Australian, Indian and others). You do not need a “native” accent. You need consistency, clarity and natural stress patterns.

• The engine was trained on many first languages. The Score Guide (page 13) says the data used to train Pearson's scoring engines came from a field test of more than 10,000 test takers from 158 countries who spoke 126 different first languages, including Hindi, Gujarati, Marathi, Tamil, Telugu, Urdu, Mandarin, Cantonese, Vietnamese, Korean, Thai and Spanish. The spoken portion is scored with Pearson's Versant speech technology.

What the official band descriptors reward, and which mistakes you can afford

Pearson publishes the band descriptors behind Oral Fluency and Pronunciation (Score Guide, pages 45 to 46) and a Content scale for every speaking task except Answer Short Question, which is scored on Vocabulary instead. They are the nearest thing to an official answer to “what mistakes are OK if I want 88?”. Pearson does not publish how item bands convert into your 10 to 90 Speaking score, so nobody can give you an exact mistake allowance. What the descriptors do show is where each band draws its line. In our own words:

So which mistakes can you afford at 88? This is our reading of the descriptors, not a published Pearson rule. An occasional single hesitation or restart, a few small sound slips that do not blur a word, and an accent that is clearly your own all sit inside the upper bands. What the descriptors place lower are habits that repeat: several pauses and restarts in one answer, consonants dropped again and again, and wrong stress on common words. At 88 you are two points from the top of the scale, so the most expensive mistake is not a small slip. It is an item where Content scores 0 and nothing else is counted: a silent answer (including one cut off by the 3-second rule before you start), or one recited from a template.

Task 1: Read Aloud (6-7 questions)

What happens: a passage of up to 60 words appears on screen. You get 30 to 40 seconds to prepare, then you read it aloud. This task scores Speaking only (it no longer contributes to Reading after the August 2025 update).

Why it matters: with 6 to 7 items, Read Aloud is the highest-volume speaking task after Repeat Sentence. It is also the most straightforward to improve because you can see every word before you speak.

The mistake: reading word by word

Most students read each word separately, with equal stress on every syllable. The AI hears this as robotic, choppy delivery and penalises Oral Fluency. English is a stress-timed language: content words (nouns, verbs, adjectives) receive stress while function words (the, of, a, to, in) are reduced and linked to the next word.

How to fix it

• Use your prep time to chunk. Mentally divide the passage into 3 to 5 meaningful phrases. Do not try to memorise it. Just identify where natural pauses fall (usually at commas, conjunctions and clause boundaries).
• Mark stress words. In your head, pick 2 to 3 key content words per chunk and plan to stress them slightly louder and longer.
• Start confidently. The first 2 seconds set the AI's baseline for your fluency. Take a breath, then begin at a steady, moderate pace.
• Do not self-correct. If you mispronounce a word, keep going. A single mispronunciation costs less than stopping, repeating and breaking your flow. The AI penalises hesitations and restarts more than isolated errors.
• Falling intonation at the end. Statements end with a downward pitch. Many non-native speakers raise their pitch at the end of sentences (making it sound like a question). Practise landing your final word with a clear, confident drop in tone.

Read Aloud scoring breakdown

Practise Read Aloud with AI scoring to build muscle memory before test day.

Task 2: Repeat Sentence (10-12 questions)

What happens: you hear a sentence of 3 to 9 seconds (roughly 8 to 16 words) and must repeat it immediately. There is no prep time and no replay. This task scores Listening + Speaking, so your performance directly affects two communicative skills.

Why it matters: at 10 to 12 items, Repeat Sentence has the highest question count of any speaking task and a massive dual-skill impact. Many students lose more points here than on any other task.

The mistake: trying to memorise every word

Sentences at the 9-second end contain 14 to 16 words. Your short-term memory cannot reliably hold that many individual words. Students who try to memorise word-for-word freeze halfway through, produce a long silence, then rush the second half. The AI reads this as broken fluency.

How to fix it

• Listen for meaning, not words. Your brain retains meaning in chunks far better than it retains a word list. While listening, understand the sentence (who did what, where, why) rather than trying to record each syllable.
• Shadow the sentence internally. As you hear each phrase, silently mouth it. This engages your motor memory alongside your auditory memory, doubling retention.
• Start immediately. Begin speaking within 1 second of the audio ending. The longer you wait, the more words you lose. Speed matters more than perfection.
• Get the beginning and end right. If the middle blurs, the AI still awards partial marks for matching the start and the close. Do not stop mid-sentence. Fill any gap with a plausible connecting phrase and keep your rhythm going.
• Match the speaker's intonation. The sentence is spoken by a native speaker with natural stress and rhythm. If you mirror their pattern (stress the same words, pause at the same spots), your Oral Fluency score rises even if a word or two is wrong.
• Never stay silent. A blank response scores zero across all traits. Even a partial attempt (4 to 5 correct words out of 14) earns something. Silence earns nothing.

Repeat Sentence scoring breakdown

Practise Repeat Sentence with instant feedback to train your auditory memory every day.

Task 3: Describe Image (5-6 questions)

What happens: a chart, graph, map, table or diagram appears on screen. You get 25 seconds to study it, then 40 seconds to describe it aloud. This task scores Speaking and receives hybrid scoring (AI + human content review).

Why it matters: with the August 2025 update, Describe Image increased from 3-4 items to 5-6 items, and the content scoring scale widened from 0-5 to 0-6. Combined, this makes Describe Image one of the highest-weighted speaking tasks in the 2026 format. That ranking rests on its item count and Content scale, because Pearson publishes no item weights.

The mistake: describing every single data point

Students panic at the volume of information in a graph and try to read out every number, every label, every year. They run out of breath at 20 seconds, then either stop or start rambling. The AI (and now the human reviewer) wants structure and key trends, not a data dump.

How to fix it

• Use your 25 seconds wisely. In the first 10 seconds, identify the type (bar chart, line graph, pie chart, map, process diagram), the title, the axes/labels and the time period. In the remaining 15 seconds, identify the 2 to 3 most obvious trends, extremes or comparisons.
• Follow a structure. Open with what the image shows (type + topic + time period). Then describe the main trend. Then give 1 to 2 specific data points as evidence. Close with a summary statement. This is not a template because the specific content changes every time. It is a framework.
• Use comparison language. Phrases like “increased significantly,” “was roughly double,” “the highest value was X while the lowest was Y” demonstrate analytical vocabulary without needing exact numbers.
• Round numbers. Saying “approximately 45 percent” is faster and sounds more natural than “forty-four point seven percent.” The AI does not penalise rounding. It rewards fluency.
• Fill the 40 seconds. Aim for 35 to 40 seconds of speech. If you finish your main points at 25 seconds, add a closing observation (“Overall, the data suggests that...”). Short responses leave marks on the table.

Describe Image scoring breakdown (2026 scale)

Practise Describe Image with real chart types and check our Describe Image framework guide for the opening structure.

Task 4: Retell Lecture (2-3 questions)

What happens: you listen to a lecture of up to 90 seconds (usually accompanied by an image or slide) and then get 10 seconds of prep time and 40 seconds to retell what you heard. This task scores Listening + Speaking and receives hybrid scoring.

Why it matters: Retell Lecture tests two skills simultaneously and has a widened content scale (0-6, up from 0-5), making each item worth more than before. It also has the longest audio stimulus of any speaking task except Summarize Group Discussion.

The mistake: trying to take notes on everything

Students scribble furiously on the erasable whiteboard, writing full sentences. They end up with messy, unreadable notes and spend the 10-second prep time trying to decipher their own handwriting instead of organising their response.

How to fix it

• Use keyword notes, not sentences. Write 5 to 8 single keywords or short phrases. Focus on the topic, the speaker's main argument, key examples and the conclusion. One word per point is enough to trigger your memory.
• Listen for signposting language. Lecturers use phrases like “the main point is,” “for example,” “however,” and “in conclusion.” These signals tell you what is important enough to retell.
• Glance at the image. The accompanying image often summarises the lecture topic. Use it as a visual anchor: mention what it shows in your opening (“The lecture, accompanied by a diagram of X, discussed...”).
• Structure your retelling. Open with the topic and the speaker's main claim. Give 2 to 3 supporting points. Close with the conclusion or implication. This mirrors academic discourse and scores well on Content.
• Paraphrase, do not recite. You are retelling, not repeating. Use your own words. Saying “The speaker explained that renewable energy adoption is accelerating” is better than trying to recall the exact phrase from the lecture. Paraphrasing demonstrates comprehension and avoids template detection.
• Use all 40 seconds. The wider 0-6 content scale means there are more marks available for adding detail. If you finish at 25 seconds, add a concluding sentence or mention an example you initially skipped.

Retell Lecture scoring breakdown (2026 scale)

See a Retell Lecture sample answer.

Practise Retell Lecture with real audio clips to build your note-taking speed.

Task 5: Answer Short Question (5-6 questions)

What happens: you hear a short question (3 to 9 seconds) and must answer in one or a few words within 10 seconds. This task scores Listening only (it no longer contributes to Speaking after the August 2025 update). It is marked as either correct or incorrect, with no partial scoring.

Why it matters: while each item is worth less individually (binary scoring), there are 5 to 6 of them and they flow quickly. Getting them right is easy marks, and getting them wrong usually means you did not understand the question.

The mistake: over-explaining

The question asks “What do we call a doctor who specialises in children?” and the student answers “Well, I think it is called a paediatrician, which is a doctor who treats children and adolescents.” The AI only needs one or a few words. Extra words waste time, risk errors and can confuse the speech recogniser.

How to fix it

• Answer in 1 to 3 words. “Paediatrician.” That is all you need. The AI matches your answer against a list of acceptable responses. One clean word beats a rambling sentence.
• Respond immediately. Do not repeat the question back. Do not say “I think the answer is...” Just say the answer.
• Build general knowledge vocabulary. These questions test real-world knowledge: science, geography, everyday vocabulary. If you consistently get them wrong, spend 10 minutes a day reading English news headlines and noting unfamiliar terms.
• If you do not know, say your best guess. Silence scores zero. Even a wrong guess has no penalty (no negative marking). Speak.

Task 6: Respond to a Situation (2-3 questions) [New in 2025]

What happens: you read a short real-world scenario (up to 60 words) describing a situation: a complaint, a request, a workplace problem, a social scenario. You get 10 seconds to prepare, then 40 seconds to record a spoken response. This task scores Speaking and receives hybrid scoring.

Why it matters: this is one of two brand-new tasks introduced in August 2025. It tests pragmatic English: your ability to choose the right tone, register and content for a specific real-world situation. It cannot be templated because every scenario is different. This is exactly the kind of task that separates genuine speakers from memorised-answer students.

The mistake: ignoring tone and register

Students treat it like Describe Image: state the facts and move on. But Respond to a Situation is about appropriateness. If the scenario describes a colleague who missed a deadline, a formal lecture-style response is wrong. The situation requires empathy, directness and a solution-oriented tone. The human reviewer specifically checks whether your register matches the context.

How to fix it

• Read the scenario twice in 10 seconds. First pass: who are you, who are you speaking to, what happened? Second pass: what tone is appropriate? (Formal to a manager, friendly to a colleague, empathetic to a customer.)
• Open by addressing the situation directly. “I understand you have been waiting longer than expected for your delivery” is stronger than “I am going to talk about a situation.” Show the human reviewer that you grasped the context.
• Offer a clear action or opinion. The scenario usually requires you to do something: apologise, suggest a solution, make a request, give advice. State it explicitly. Vague responses score low on Content.
• Match the tone. Use polite hedging for formal situations (“I would suggest,” “perhaps we could”). Use warmer, more direct language for informal ones (“I think the best thing would be,” “let me help you with that”). Avoid overly casual slang in any scenario.
• Use the full 40 seconds. After your main response, add a closing: a summary of next steps, a positive note, or a check-in (“Would that work for you?”). This demonstrates pragmatic completeness.

Respond to a Situation scoring

See a Respond to a Situation sample answer and our Respond to a Situation framework.

Task 7: Summarize Group Discussion (2-3 questions) [New in 2025]

What happens: you listen to a discussion between multiple speakers (up to 3 minutes, typically 3 speakers). You get 10 seconds of prep time and then 2 minutes to speak your summary. This task scores Listening + Speaking and receives hybrid scoring.

Why it matters: this is the other new task from August 2025, and it has the longest speaking time of any task (2 minutes, compared to 40 seconds for most others). It tests whether you can follow a multi-voice discussion, identify different viewpoints and synthesise them coherently. The 2-minute window means you need sustained fluency, not just a 30-second burst.

The mistake: summarising only one speaker

The discussion has 3 speakers with different viewpoints. Students lock onto the first or most confident speaker and summarise only their perspective. The human reviewer (and the AI) check whether you captured multiple viewpoints and the overall direction of the discussion. Missing a speaker means missing marks.

How to fix it

• Note-take by speaker. Draw 3 short columns on your whiteboard (Speaker 1 / Speaker 2 / Speaker 3). Jot 2 to 3 keywords under each as they speak. This forces you to track all perspectives.
• Listen for agreement and disagreement. Discussions have structure: one person states a view, another agrees or disagrees, a third adds a new angle. Identifying these relationships (“while the first speaker argued X, the second disagreed, suggesting Y”) shows higher-level comprehension.
• Open with the topic. “The discussion was about [topic], and the speakers had [similar / different / mixed] views on [specific aspect].” This frames your summary and immediately signals to the reviewer that you understood the discussion.
• Cover each speaker. Even if briefly, mention each perspective: “One speaker emphasised..., another argued..., while the third pointed out...” Three viewpoints, even briefly stated, score higher than one viewpoint in elaborate detail.
• Close with the outcome. Did the speakers reach a consensus? Did they disagree? What was the overall takeaway? A closing sentence like “Overall, the group acknowledged that X, but disagreed on Y” wraps your summary neatly.
• Pace yourself for 2 minutes. Two minutes is a long time to speak continuously. Plan a structure: 20 seconds for the introduction, 30 seconds per speaker (90 seconds total), 10 seconds for the conclusion. If you finish early, expand on the most interesting viewpoint.

See a Summarize Group Discussion sample answer and our group discussion framework.

Pearson's own PTE speaking tips and official preparation resources

If you want the source rather than a coaching opinion, these are the speaking tips Pearson itself publishes on its test-format page, in its speaking tips article and in its Test Taker Handbook (July 2025), summarised task by task:

Pearson's official preparation resources. The Test Taker Handbook lists free Guided Practice Tests (they do not give a score, but show sample answers for Speaking and Writing), four free Test Tips documents, four free video courses split by skill and a free study plan, all reachable through the Smart Prep portal in myPTE. Because the Guided Practice Tests do not score you, pair them with a timed, scored PTE speaking mock test.

How to improve fluency in PTE speaking

Oral Fluency is scored on every speaking task (except Answer Short Question). It is the single trait most students underperform on, and it responds fastest to practice. Here are the techniques that produce results.

• Speak in phrases, not words. English fluency is about connected speech. Link words together: “sort of” becomes “sor-dov,” “want to” becomes “wanna” in natural speech. You do not have to use contractions, but linking function words to content words creates the rhythm the AI expects.
• Eliminate filler words. “Um,” “uh,” “like,” “you know” are penalised as hesitation markers. If you need a beat to think, use a short, silent pause (under 1 second). A brief silence is better than a filler.
• Do not self-correct. If you say “the graph shows a decrease... I mean an increase,” the AI counts the restart as a fluency break. Say “the graph shows a decrease” and move on. One wrong data point costs less than a broken rhythm.
• Practise with a timer. Record yourself speaking for exactly 40 seconds on a random topic every day. Play it back and count your pauses. Our speaking practice module times you exactly like the real test.
• Shadow native speakers. Pick a podcast or YouTube video. Play a sentence, pause, and repeat it immediately, copying the speaker's rhythm and stress. Do this for 10 minutes a day. After two weeks, your natural speaking rhythm will shift.

Remember: more than 3 seconds of silence in any speaking task triggers the auto-stop mechanism. Your microphone cuts off and the response ends. Aim for continuous speech with only natural, brief pauses between phrases.

How to improve pronunciation in PTE

PTE's pronunciation scoring does not penalise accents. Indian, Chinese, Vietnamese, Korean, Arabic, Spanish and every other accent can reach the top Pronunciation band (5 out of 5) and score 90 in Speaking. What the AI evaluates is intelligibility: can the words be clearly understood?

• Focus on word stress, not individual sounds. English word stress is predictable by suffix: words ending in “-tion” stress the syllable before it (edu-CA-tion, not ED-u-ca-tion). Words ending in “-ic” stress the syllable before it (aca-DEM-ic). Words ending in “-ity” stress the third syllable from the end (uni-VER-si-ty). Learning these rules fixes dozens of pronunciation errors at once.
• Master the th sound. Speakers of many languages substitute “t” or “d” for “th.” The words “the,” “that,” “this,” “there,” “through” and “think” appear in almost every PTE passage. Practise placing your tongue between your teeth. Even an approximation is better than a hard “d” or “t.”
• Distinguish long and short vowels. “Ship” vs “sheep,” “bit” vs “beat,” “full” vs “fool.” Many languages do not make this distinction. In English, vowel length changes meaning, and the AI picks up on it.
• End your words. In many languages, final consonants are dropped or softened. In English, “world,” “helped,” “asked” all need their final consonant clusters pronounced. Dropping them reduces intelligibility.
• Record and compare. Use our Read Aloud practice to record yourself, then compare your recording with the model answer. Focus on the words where your stress pattern differs from the model, not on matching the accent.

How to stop dropping word endings and final consonants

If you drop word endings (“ask” for “asked”, “cost” for “costs”), fix that before you polish anything else. The Score Guide shows why. Its pronunciation descriptors (page 45) track consonants directly: at band 4 only a few consonants or consonant groups may be distorted and every word stays easy to understand; at band 3 a few consonants are regularly distorted, omitted or mispronounced in certain contexts; at band 2 some consonants are regularly omitted and consonant groups may be simplified. The top two Oral Fluency bands also allow no (or no significant) “phonological simplifications”. In Read Aloud and Repeat Sentence, where Content counts the words you say, a swallowed ending may also be heard as a different word; that last point is our inference, because Pearson does not publish how its recogniser treats a dropped ending.

The endings to drill

A six-step drill for final consonants

1. Record ten Read Aloud sentences and underline every -ed, -s and final consonant group before you listen back.
2. Listen for the endings only and mark each one you cannot hear.
3. Back-chain the hard words: say the last sounds first, then add the rest (“-skt”, then “asked”; “-sts”, then “costs”).
4. Link the ending into the next word instead of stopping hard: “asked about” flows as one unit. That linking is part of natural connected speech, which the top pronunciation band allows.
5. Do not add a vowel. “Ask-a” or “i-school” adds a syllable, and the Score Guide's lower pronunciation bands mention syllables added or missed.
6. Re-record the same ten sentences the next day and compare, then move to Repeat Sentence practice, where you cannot see the words.

Template detection: why memorised answers no longer work

Before August 2025, students could memorise a Describe Image template (“The given image is a [type] which shows [topic]. According to the [image], we can see that...”) and use it for every question. This inflated scores because the AI rewarded the fluency of a practised script.

That era is over. Pearson's updated scoring model includes template detection that identifies memorised, formulaic responses. Community reports suggest template-flagged answers can lose a significant portion of their content score, and the human reviewers (who now check 4 speaking tasks) are specifically trained to identify and penalise canned responses.

The official rule is stricter than those community reports, and it predates the August 2025 update. Pearson's previous Score Guide (Version 21, November 2024) already put significant pre-prepared or memorised material at a Content score of 0 on Describe Image and Retell Lecture, with a human review of those two tasks' Content. That guide also stated the rule for every response on page 8, and the current guide states it the same way: pre-prepared or memorised material counts as an irrelevant response, Content scores 0, and the response earns no score points at all (page 8). The August 2025 update extended the human review of Content from those two question types to seven (page 15). Pearson's task pages put it in six words: “You must use your own words.”

What still works: a framework (a structured approach where the specific content changes every time) is fine. Saying “This bar chart shows [specific topic] from [specific years]” is a framework because the bracketed parts change. Saying “The given image is a very informative and interesting representation” word-for-word every time is a template. The difference is whether your actual content responds to the specific question.

For more detail on how scoring changed, read our complete guide to PTE 2025 scoring changes.

Complete PTE speaking task overview (2026 format)

Here is every speaking task at a glance, with the current question counts, timing and scoring method. Use this as a quick reference during your preparation.

Total speaking items per test: approximately 32 to 40 questions. That is about half the entire test's question count (65-75 questions total), which is why your speaking preparation carries outsized weight.

Those totals are simple arithmetic over the Score Guide's per-type ranges. Because a test must stay within 65 to 75 questions in total, no test uses the top of every range at once across all 22 question types.

A 30-minute daily speaking practice plan

Consistency beats intensity. Thirty focused minutes every day produces faster improvement than a 3-hour weekend session. Here is a plan built around the tasks that carry the most marks.

After each session, review your recordings. Listen for pauses longer than 1 second, filler words, and words where your stress was off. Fix one specific thing per day, not everything at once.

Speaking score targets: what you need for 65, 79 and 90

Your target score determines your strategy. Here is what each level requires in practical speaking terms.

If you are targeting 79+, your biggest gains will come from Repeat Sentence (high item count, dual-skill) and Describe Image (high item count, widened content scale). If you are targeting 90, Summarize Group Discussion is the differentiator because most candidates cannot sustain 2 minutes of fluent, structured speech.

Take a free mock test to see where you stand, then focus your practice on the tasks where you lose the most marks.

What 1,561 of our mock tests show about PTE Speaking scores

We looked at every fully answered PTE Academic mock test taken on PTE Mocks between 6 July and 27 September 2026: 1,561 tests by 909 people, counting only tests where every skill scored above the 10-point floor. Speaking stood out in every per-skill cut we made:

How to read this. These are our AI-scored mock estimates, not Pearson results. Mock takers are self-selected practisers, many early in their preparation, so absolute levels run low; read the relative pattern, not the numbers as a forecast. The pattern is consistent: Speaking is the skill most often holding people below their target, a single mock is a noisy reading (40% of second mocks on a different paper came in lower overall), and the gains show up across several attempts rather than in one. For more of our score data, see our PTE mock test statistics page.

Seven common speaking mistakes (one per task)

Here is every major mistake condensed into one reference table. Print it, screenshot it, or bookmark this page and review it before every practice session.

Test day speaking checklist

On the day of your PTE, these small details make a measurable difference to your speaking score.

• Microphone check: during the equipment check at the start, speak at the volume you plan to use throughout the test. The system calibrates to your check volume. If you whisper during the check and then speak normally, the levels will be off.
• Distance from the mic: keep the microphone about 2 to 3 centimetres from the corner of your mouth (not directly in front). This avoids popping sounds on “p” and “b” while keeping your voice clear.
• Consistent volume: do not start loud and fade out, or vice versa. The AI evaluates your speech stream holistically. Inconsistent volume can be misread as hesitation or loss of confidence.
• Watch the timer. Every speaking task has a visible countdown. Glance at it once when you are about halfway through your response. If you have 20 seconds left and you are running out of content, slow down slightly and add a closing sentence rather than stopping early.
• Do not cough or clear your throat into the mic. If you need to cough, turn your head away from the microphone. The AI may interpret loud non-speech sounds as interference.
• Use the 3-second rule. More than 3 seconds of silence auto-stops your recording. If you feel stuck, say anything relevant. Even “Additionally...” followed by your thought buys you time without triggering the cut-off.

For a full walkthrough of what happens at the test centre, see our PTE exam day tips guide.

Why mock tests are essential for speaking improvement

Speaking is the one PTE skill you cannot improve by reading about it. You have to speak, record, listen, adjust and repeat. Mock tests give you the exact test environment: a timer, a microphone, and AI scoring that mirrors the real exam's evaluation.

• They expose time-management gaps. Most students have never spoken for exactly 40 seconds on command. Mock tests train you to fill the window without rushing or stopping short.
• They identify your weakest task. Your first mock test score report breaks down performance by task type. If Repeat Sentence is dragging your score down, you know to double your practice time on it.
• They build exam stamina. The Speaking & Writing section lasts 76 to 84 minutes. Sustained concentration for that long is a skill in itself. Only full-length mocks train it.
• They reduce test-day anxiety. When the real test feels familiar, you perform better. Students who complete 3 or more full mock tests before their exam date consistently report feeling more confident and less rushed.

Keep in mind: mock tests do not have human content review or template detection, so template-heavy answers may score higher on mocks than on the real exam. Focus on genuine, structured responses during mock practice and your real test score will match or exceed your mock score.

Quick reference: PTE speaking tasks at a glance

The seven speaking tasks and the scoring rules that decide your marks, condensed into one table.