PTE Speaking

Understanding the PTE Speaking Section

The PTE Academic Speaking section is unique because it's entirely computer-based with AI or hybrid AI and Human scoring. This means the computer analyzes your recordings for specific features, including pronunciation, oral fluency, and content accuracy.

Section Format

Question Type

Count

Time

Scoring Impact

Read Aloud

6-7

30-40 sec prep, 30-40 sec record

Speaking only

Repeat Sentence

10-12

3-9 sec per item

Speaking + Listening

Describe Image

3-4

25 sec prep, 40 sec record

Speaking only

Re-tell Lecture

1-2

10 sec prep, 40 sec record

Speaking + Listening

Answer Short Question

5-6

3 sec answer

Listening only

Summarize Group Discussion (SGD)

1-2

~60–90 sec audio, 10 sec prep, ~2 min record

Speaking + Listening

Respond to a Situation (RTS)

1-2

~30 sec audio + on-screen text, 20 sec prep, 40 sec record

Speaking only

Scoring Criteria

You are evaluated on three main areas:

1. Content: Accuracy and relevance of what you say
2. Oral Fluency: Smooth, natural delivery without hesitations
3. Pronunciation: Clear production of sounds, stress, and intonation.

2025 Update: Hybrid AI + Human Scoring

Since 7 August 2025, PTE Academic uses a hybrid scoring model. AI still scores every response, but for several speaking tasks, a human content rater also reviews the response. AI continues to score pronunciation and oral fluency on its own (accent and appearance are never judged by humans) and these mechanical traits remain machine-only.

Speaking tasks now subject to human content review:

• Describe Image
• Re-tell Lecture
• Summarize Group Discussion (SGD)
• Respond to a Situation (RTS)

What human reviewers check: Content relevance, originality, and whether the response actually engages with the prompt. They do not judge pronunciation, fluency, or accent - those remain AI-only.

What this means for your preparation:

• Memorised templates are now actively risky. Pearson's scoring documentation states that no credit is given for unoriginal responses. Pre-memorised openers and closers that could fit any prompt, such as, "This image shows…", "The speaker discusses…", "In conclusion, the lecture was very informative" etc., are increasingly flagged as unoriginal and can result in zero on the Content trait.
• Use structures, not scripts. A logical opening → body → closing flow with linking words is fine. Pasted-in canned phrases that don't engage with the actual content are not.
• Specificity wins. Lead with the actual topic, name specific data points, attribute views to the speaker, end with a real implication. The hybrid scoring system is designed to reward genuine engagement with the prompt.

The same principle applies to written tasks (Summarize Written Text, Write Essay) - see the Writing guide for equivalent advice on those.Question Type Strategies

Read Aloud

Read a text (up to 60 words) aloud. This task now contributes to your Speaking score only, but it's still high-value because of its frequency (6–7 per test) and influence on overall fluency scoring.

Strategy:

1. Preparation time (30-40 sec):

• Identify difficult words
• Note punctuation for pauses
• Mark stress patterns mentally

1. Recording time (30-40 sec):

• Start when the timer begins
• Maintain a steady pace
• Use punctuation as pause guides
• Don't stop for mistakes - keep going

Pronunciation Tips:

• Word stress: Content words louder/longer than function words
• Sentence stress: Emphasize key information
• Intonation: Rise for questions, fall for statements
• Linking: Connect words naturally (e.g., "pick_up")

Common Problem Words: | Word | Correct | Common Error | |------|---------|--------------| | Determine | de-TER-mine | DE-ter-mine | | Development | de-VEL-op-ment | de-ve-LOP-ment | | Particularly | par-TIC-u-lar-ly | par-ti-cu-LAR-ly | | Significant | sig-NIF-i-cant | SIG-ni-fi-cant |

Repeat Sentence

Listen to a sentence and repeat it exactly.

Strategy:

1. Focus completely - no distractions
2. Catch the beginning - first 3-4 words are crucial
3. Remember the end - the last few words are often remembered well
4. Fill in the middle - use grammar knowledge to complete
5. Speak immediately - don't wait, start talking

Memory Techniques:

• Visualize the scene described
• Group words into meaningful chunks
• Use rhythm and intonation as memory aids
• Don't try to understand, just repeat sounds

If You Miss Words:

• Say what you remember fluently
• Don't pause to think
• Partial credit is available
• Never stay silent

Describe Image

Describe a graph, chart, map, image, or process diagram in 40 seconds. You get 25 seconds of preparation.

Strategy:

1. Identify the image type and main subject in the first 5 seconds of preparation
2. Pick out 2–3 specific data points or features worth describing
3. Note the most striking trend, comparison, or relationship
4. Keep speaking for the full 40 seconds - silence hurts the score
5. Maintain a steady pace; small mistakes are fine if you keep flowing

A note on templates (2025 update): Since August 2025, Describe Image is reviewed by both AI and a human content rater. Generic openers like "This image shows…" followed by interchangeable filler add no information and are increasingly flagged as unoriginal. Pearson's scoring documentation states that no credit is given for unoriginal responses. Use a structure, not a script.

Recommended structure:

• Opening (5–7 sec): State what the image is and the main thing it conveys. Don't just label the format - say what it's actually about.
• Body (25–30 sec): Describe 2–3 specific features with real numbers, names, or comparisons. Use linking words (while, whereas, in contrast, similarly, peaking at, dropping to) to show relationships rather than listing.
• Closing (5 sec): End with the overall trend, takeaway, or what the data implies - not a generic "In conclusion" tag.

Example (bar chart of energy use by sector):

"This bar chart breaks down energy consumption across four sectors in 2023, and what stands out is how dominant industry is - at roughly 38%, it uses more than transport and residential combined. Transport sits around 25%, residential at 22%, with commercial trailing at about 15%. The pattern suggests that decarbonisation efforts focused on industry would have the largest single impact on overall energy use."

Notice this opens with substance, gives concrete numbers, links them with comparison words, and ends with an actual implication.

Quick guides by image type:

• Line graph: Lead with the overall direction (rising, falling, fluctuating). Mention the start point, peak/trough, and end point with values.
• Bar chart: Lead with the highest and lowest values. Compare 2–3 categories with concrete numbers.
• Pie chart: Lead with the largest segment and what it represents. Group small segments rather than listing every one.
• Map: Lead with the location and main features. Use directional language (north, central, along the coast) to describe distribution.
• Process diagram: Lead with what the process produces or achieves. Walk through stages using sequence words (begins with, then, after which, finally).
• Table: Lead with what the table compares. Pick the highest, lowest, and one notable contrast. Don't try to read every cell.

Re-tell Lecture

Listen to a lecture (60–90 seconds), then summarise it in 40 seconds. You get 10 seconds of preparation.

Strategy:

• Take notes during the lecture using abbreviations and symbols
• Capture the topic, 2–3 main points, and any concrete examples or numbers
• Use the 10-second prep to organise your notes into a speaking order
• Keep speaking for the full 40 seconds.

A note on templates (2025 update): Re-tell Lecture is now reviewed by both AI and a human content rater. Pre-memorised openers and closers that could fit any lecture are flagged as unoriginal. Use a structure, not a script.

Recommended structure:

• Opening (5 sec): State the actual topic and the speaker's main argument. Skip generic frames like "The speaker discusses…" - start with the content.
• Body (30 sec): Connect the speaker's main points with linking words (because, however, this led to, the speaker argued). Include any specific examples, names, or figures you caught.
• Closing (5 sec): End with the speaker's conclusion or implication, naturally - not with a tacked-on summary tag.

Example (lecture on bee colony decline):

"The lecturer was looking at why bee colonies have collapsed over the past two decades, and she identified pesticides, particularly neonicotinoids, as the main driver, though habitat loss and the varroa mite also play roles. She pointed to studies showing that even sub-lethal pesticide exposure damages bees' navigation, which prevents them returning to the hive. The takeaway was that policy responses need to combine pesticide regulation with habitat restoration rather than treat them separately."

Notice this opens with the actual topic, attributes views to the speaker, includes a specific term and finding, and ends with the speaker's conclusion.

Note-taking tips:

• Capital letters or symbols for the topic (e.g., BEES↓ for bee decline)
• Arrows and signs for relationships (→ causes, ≠ contrasts, + adds)
• Skip articles and connectors when writing; you'll add them when speaking
• Don't try to capture full sentences - focus on content words only
• 
  

Answer Short Question

Listen to a question and give a one-word or short phrase answer.

Strategy:

1. Listen for the question word (what, where, when, who)
2. Identify the topic/category
3. Answer immediately and briefly
4. One word is usually sufficient

Common Categories:

• Academic subjects
• Professions
• Places/Locations
• Time periods
• Scientific terms
• General knowledge

Examples:

• "What do we call a doctor who operates on patients?" → "Surgeon"
• "What is the study of stars and planets called?" → "Astronomy"
• "What device measures temperature?" → "Thermometer".

Summarize Group Discussion (SGD)

Listen to a discussion between (typically) three speakers on an academic topic, then summarise the key points and viewpoints. You get 10 seconds of preparation and approximately 2 minutes to speak. Scores Speaking + Listening.

Strategy:

1. Take notes during the audio: Track each speaker separately (S1, S2, S3) and capture their main point or stance
2. Listen for agreement and disagreement: Note where speakers align and where they push back
3. Identify the overall topic in the first 10–15 seconds of the audio
4. Capture concrete details: Examples, numbers, or specific terms used by any speaker
5. Use the 10-second prep to organise your notes into a speaking order, not to write new content.

A note on templates (2025 update): SGD is now reviewed by both AI and a human content rater, and Pearson has explicitly stated that templated responses are penalised on this task. Generic openers like "The speakers were discussing…" followed by interchangeable filler will be flagged as unoriginal. Originality and engagement with the actual discussion matter more here than on almost any other Speaking task.

Recommended structure:

• Opening (10–15 sec): State the actual topic and the overall framing of the discussion. Indicate that multiple perspectives were involved.
• Body (~80–90 sec): Walk through what each speaker contributed, attributing views correctly. Use linking words to show agreement, contrast, and progression (one speaker argued, another countered, a third added, while).
• Closing (10–15 sec): Note where the speakers converged or what the overall takeaway was. If they didn't reach consensus, say so - that's a legitimate observation.

Example (discussion on managing university group projects):

"The three students were debating how to handle group projects more effectively at university. The first student raised the recurring problem of unequal contribution and described past experiences where one or two people carried the workload. The second speaker pushed back slightly, suggesting that clear role allocation and shared deadlines from the start usually prevent that, and recommended using project management apps to keep everyone visible. The third speaker agreed but added that regular check-ins matter more than tools alone, because problems need to be raised early rather than at the deadline. They generally converged on the idea that communication and planning, rather than any single fix, were the key - though the first speaker remained sceptical that this would always work with disengaged team members."

Notice this opens with the actual debate, attributes specific views to specific speakers, captures genuine disagreement, and ends with the real outcome of the discussion rather than a generic summary tag.

Note-taking system:

• Top of page: T = topic
• Down the left: S1, S2, S3 with their main point next to each
• Use arrows for agreement (S2 → S1) and contrast (S2 ≠ S1)
• Capture distinctive vocabulary or examples you can use directly.

Common mistakes to avoid:

• Summarising the topic generally without attributing views to individual speakers
• Using a memorised opener that could fit any discussion
• Running out of content before 2 minutes - pace yourself; this is a long task
• Confusing speakers or attributing views incorrectly. It’s better to say "one speaker argued" if unsure of the order.

Respond to a Situation (RTS)

Listen to and read a description of a real-life scenario (academic, social, or professional), then give an appropriate spoken response - typically a voicemail, apology, request, or decline. You have 20 seconds of preparation and 40 seconds to respond. Scores Speaking only.

Strategy:

1. Read and listen to the prompt simultaneously: You have access to both, use both
2. Identify the situation type (apologising, declining, requesting, explaining, leaving a message)
3. Identify your role and audience: Are you speaking to a friend, a professor, a colleague, a stranger? The tone changes
4. Plan a clear opening, middle, and end during the 20 seconds of prep
5. Speak naturally: This task rewards conversational, real-world English, not formal academic register.

A note on templates (2025 update): RTS was specifically designed to test spontaneous, real-life communication, and Pearson explicitly discourages templated responses on this task. Memorised phrases like "I am writing to inform you that…" will sound unnatural and out of register. Speak the way you would actually speak in the situation.

Tone matters: This is the only Speaking task where social register is part of the assessment. A voicemail to a friend should sound informal; a message to a professor should be polite and slightly formal; an apology should sound genuinely apologetic, not robotic.

Recommended structure (adapt to the situation):

• Opening (5–8 sec): Identify yourself if relevant and state why you're speaking. "Hi Sarah, it's [name], I'm calling about…"
• Body (~25–30 sec): Address the situation directly. Give context, explanation, or reasoning. Be specific.
• Closing (5–8 sec): End naturally - a thank you, a next step, an offer to follow up, an apology repeated, or a sign-off appropriate to the medium.

Examples by scenario type:

Voicemail apologising for missing a meeting: "Hi James, this is [name]. I'm so sorry I missed our 10am meeting this morning. My train was cancelled and by the time I found another route I knew I wasn't going to make it. I should have called you sooner instead of trying to get there. I'm free tomorrow morning or Thursday afternoon if either of those work to reschedule. Really sorry again - just let me know what suits you."

Politely declining an invitation: "Hey Anna, thanks so much for inviting me to the dinner on Saturday. I really appreciate you thinking of me, but I'm not going to be able to make it. I've already committed to helping my sister move that weekend. I hope it goes really well, and let's catch up properly soon. Maybe coffee next week?"

Requesting an extension from a professor: "Hello Professor Chen, this is [name] from your Wednesday seminar. I'm calling to ask if it might be possible to have a short extension on the essay due on Friday. I've been unwell for most of this week and I'm worried I won't be able to do the topic justice in the time remaining. I'd be very grateful for even a couple of extra days. Please let me know what you think. Thank you."

Notice each example uses register appropriate to the audience, sounds like something a real person would actually say, and follows a natural opening → body → closing flow without sounding scripted.

Common mistakes to avoid:

• Using overly formal or academic language when the situation calls for casual speech
• Speaking in a flat or robotic tone - the assessment rewards natural delivery
• Forgetting to address the actual situation in the prompt (e.g., apologising without explaining what you're apologising for)
• Memorising opening phrases that don't fit the specific scenario
• Running short - fill the 40 seconds with relevant detail rather than ending at 25 seconds.

Quick situation checklist (adapt your response):

Situation type

Key elements to include

Apology

Acknowledge what happened, take responsibility, offer to make it right

Decline

Thank them, decline clearly, give a brief reason, suggest an alternative if appropriate

Request

State what you need, give context, be specific about timing, thank them

Explanation

Address the issue directly, give relevant detail, propose a next step

Voicemail

Identify yourself, state your purpose, give any necessary detail, sign off

Fluency and Pronunciation Optimization

Oral Fluency Tips

The AI measures:

• Smooth delivery: No unnecessary pauses
• Natural pace: Not too fast or slow
• Rhythm: Appropriate stress patterns

To Improve Fluency:

1. Practice speaking without pauses
2. If you make a mistake, continue smoothly
3. Use linking words naturally
4. Maintain consistent speed

Pronunciation Focus Areas

1. Individual Sounds:

• /θ/ (think) vs /t/
• /ð/ (this) vs /d/
• Long vs short vowels

1. Word Stress:

• Nouns: usually the first syllable
• Verbs: often second syllable
• Multi-syllable: varies

1. Sentence Intonation:

• Statements: falling tone
• Questions: rising tone
• Lists: rise, rise, fall

Practice Recommendations

1. Daily Read Aloud: Practice with academic texts
2. Shadowing: Repeat after native speakers
3. Record and Review: Listen to your recordings critically
4. Pronunciation Apps: Use technology or mock tests for feedback
5. Timed Practice: Always practice with actual time limits.

What to Expect in the Test Interface (2025 Update)

Pearson updated the on-screen interface in 2025 with a few features worth knowing about before test day:

• Live audio waveform during speaking tasks. When you speak, the screen shows a real-time visualisation of your voice volume. This is genuinely useful - it confirms your microphone is picking you up clearly. If the waveform is flat or barely moving, you're either too quiet or too far from the mic. Speak at a normal conversational volume and check the waveform looks active when you start.
• Double microphone check at the start of the test. Before the Speaking section begins, you'll be asked to test the microphone twice rather than once. This is to reduce technical issues mid-test. Speak clearly and at normal volume during both checks.
• Beep before Repeat Sentence. A short beep now plays after the audio finishes and before the recording starts. This is a clearer signal of when to begin speaking. Don't speak before the beep, wait for it, then start immediately.
• Circular progress indicator during recording shows how much time remains for your response. Use it as a pacing guide, especially for the longer tasks (Describe Image, Re-tell Lecture, SGD).

None of these change what you do, only what you'll see while doing it. 

Common Challenges and Solutions

Challenge

Solution

Unnatural pauses

Practice reading aloud daily, focus on flow

Pronunciation errors

Work with phonetic charts, shadow native speakers

Running out of content

Practice describing varied image types using a flexible structure; build a vocabulary of comparison and trend language so you can describe whatever appears 

Memory for Repeat Sentence

Practice chunking, focus on beginning/end

Nervous delivery

Practice under test conditions regularly