Interactive Speaking & Listening Updates (DET 2025)

Quick Navigation:

• 2025 DET Changes →
• Interactive Speaking →
• Interactive Listening →
• Scoring & Subscores →
• Test Comparisons →
• Technical Requirements →
• Conclusion →
• Practice Questions →

2025 changes to the Duolingo English Test

Why Duolingo changed the test

Language proficiency testing has long faced a basic trade-off between authenticity and efficiency. Direct assessments with live human interaction, such as face to face interviews, offer high authenticity but are hard to scale, costly, and can show scoring inconsistencies.¹ Semi direct, machine administered tests offer scalability, consistency, and efficiency, but they often lack true interactivity. They become a series of monologues rather than a genuine conversation.¹ The Duolingo English Test (DET) updates, effective July 1, 2025, are a planned, tech based effort to address this long standing tension. The stated goal is to greatly improve the test's authenticity, interactivity, and overall efficiency. The aim is to make the test mirror real world academic communication and feel less "robotic".²

This change starts with a clear shift in assessment philosophy. They removed tasks that test discrete skills in isolation, such as "Read Aloud" for pronunciation. At the same time, they added tasks that require a dynamic mix of skills, like "Interactive Speaking" and the revamped "Interactive Listening". The new components are explicitly designed to simulate "real time conversations" and "elicit real world interaction skills".¹ This matches the linguistic concept of "Interactional Competence," the ability to manage communication in various social contexts.³ This is not merely a format update. It redefines what Duolingo sees as the most crucial aspect of English proficiency for academic success. The test now prioritizes the ability to manage a conversation, developing topics, responding appropriately, and summarizing key information, as a more powerful predictor of success than performance on isolated linguistic skills. This has profound implications for how candidates must prepare for the test and how admitting institutions should interpret its results.

Learn More: For a comprehensive overview of DET preparation strategies and detailed test mechanics, explore our Complete DET Preparation Guide.

Summary of key updates

The July 2025 updates add a set of changes that balance new interactive tasks with the test's short duration. The test still runs in under one hour. The adaptive portion still has 52 questions. This keeps the overall test length and the amount of data collected stable.⁴

The key changes can be categorized as follows:

1. Additions: The most significant changes introduce two new components. The first is Interactive Speaking, a new task that simulates a conversation with an AI character.⁵ The second adds a Listen and Complete activity to the start of the existing Interactive Listening section. Test-takers now listen to a scenario instead of reading it.²
2. Removals: To make room for the new, more complex tasks, the test retired two question types: Read Aloud and Listen, Then Speak.² These items mostly tested basic pronunciation and simple recall. Removing them supports the test's move toward measuring more integrated, interactive communication skills.
3. Procedural Shifts: The test interface was updated to mirror the Duolingo learning app. This should make navigation easier.⁵ The pretest security and onboarding steps were also strengthened. The process now focuses on ID checks and requires a second camera, such as a phone or tablet, to monitor the testing area. This step improves test security.⁴

Effect of removing minimum time limits

Duolingo removed the minimum time requirements for several key speaking and writing tasks to make the test easier to use and simplify the flow of the exam.² This change applies to:

1. Speak About the Photo
2. Read, Then Speak
3. Interactive Writing
4. The Writing Sample
5. The Speaking Sample

Before, test takers who finished their responses early had to wait for the timer to run out before moving on. Now they can submit answers as soon as they are done, so they do not wait unnecessarily.⁶ Duolingo says internal research shows that shorter, more concise answers still provide a valid and reliable measure of a candidate's language proficiency.² This change reflects the main principle of the DET's design, which is to make the assessment more efficient without reducing psychometric rigor. For the test taker, it lets them set a more personal, smoother pace through the latter half of the exam.

Practice Writing:

Since the minimum time limits were removed, practice your Interactive Writing Questions to improve your writing speed and quality.

Interactive Speaking section explained

The new Interactive Speaking section is the main feature of the 2025 DET update. It tests spontaneous, conversational English in a way that feels natural and can be offered to many test takers. It directly addresses the limitations of previous semi direct speaking tasks.

Interactive Speaking task mechanics, turn by turn

The task is a simulated conversation with an animated AI character, either Bea or Oscar, in a series of question and answer turns.¹ The mechanics are meant to elicit rapid, unrehearsed speech.

1. Number of Questions: While initial documentation from Duolingo suggested a range of 6 to 8 questions ¹, after launch independent testing platforms found a consistent, fixed format. The task now has two distinct sets of questions, each with three sub questions, for a total of six questions per test.⁴ This fixed structure makes test administrations more consistent.
2. Timing: For each of the six questions, test takers have a strict 35 second limit to record their response.⁷ There is no dedicated preparation time between hearing the question and the start of the recording timer. That forces spontaneous, immediate speech production.¹ Some early practice versions showed a brief 6 second preparation window. A 15 second buffer is provided at the very beginning of the entire section. But the turn by turn interaction is immediate.⁸
3. Audio Prompts: The audio for each question can be played only once.⁴ That raises the importance of active listening and comprehension, since there is no chance to review the prompt.
4. Topic Shifts: The conversation covers two different topics. A clear topic shift usually happens midway through the task, typically after the third question. The test taker must change focus and vocabulary accordingly.⁹

How the adaptive engine selects follow up questions

The main technical advance of the Interactive Speaking task is its adaptive engine. The system picks follow-up questions in real time, based on what the test-taker said and the linguistic quality of their prior response.¹ This creates a more personal, responsive testing experience than a fixed, pre-scripted interview.

You should understand how this AI-driven adaptation works. Duolingo has said the system avoids the risks of "open-ended generation by using pre-written content".¹ So the AI is not a generative conversational partner like consumer chatbots. It works as a complex selection algorithm. The AI analyzes features of a spoken response, such as lexical diversity, grammatical complexity, and fluency, and uses that analysis to choose the most psychometrically appropriate

1. next prompt from a large, chosen, and pre-approved question bank.
2. For example, a response showing high proficiency might send the algorithm toward more abstract or complex follow-up questions, while a hesitant or grammatically simple reply might trigger a question meant to elicit a more basic answer.
3. This "Constrained AI" model keeps the test standardized and fair, while still letting it be interactive and adaptive.
4. For the test-taker, this means the path to a high score is not to try to have a "natural" chat, but to consistently produce speech that shows the specific linguistic markers of high proficiency that the AI is trained to detect and reward.

AI scoring of fluency, coherence, and content

A suite of in-house AI models automatically transcribes, analyzes, and scores every response in the Interactive Speaking section.¹ They are trained to evaluate performance across six distinct sub-skills that form the basis of communicative competence ¹:

1. Fluency: The smoothness, speed, and natural rhythm of speech.
2. Grammar: The accuracy and complexity of grammatical structures.
3. Pronunciation: The clarity and intelligibility of speech.
4. Vocabulary: The range, sophistication, and appropriateness of lexical choices.
5. Coherence: The logical organization and connection of ideas.
6. Task Completion: The relevance and directness of the response to the question asked.

Duolingo's research shows these AI-based scores are very reliable, with internal consistency estimates above 0.90, matching the most trusted human-rated speaking tests used today.¹ Also, the scores correlate strongly with scores from human interviews like the IELTS Speaking test. In studies that compared AI and expert human raters on task completion, agreement levels were nearly identical, showing the AI can correctly judge a response's content.¹

Strategies to improve your score

This section moves fast and has high stakes. You need procedural awareness, strategic thinking, and focused preparation.

1. Procedural Discipline: The recording timer begins as soon as the question page loads, but the microphone does not activate until the test taker clicks the "RECORD NOW" button. Click this button right away to avoid losing valuable response time.⁴
2. Strategic Timing: With a limit of 35 seconds, aim for about 30 seconds of continuous speech. This gives you time to make a clear point and add supporting details, while leaving a 5 second buffer to finish a final sentence without being cut off.⁹
3. Spontaneous Content Generation: You have no preparation time, so you cannot plan in detail. Practice answering with the first clear, relevant idea that comes to mind, since hesitation wastes precious seconds.⁹
4. Focused Listening: Audio prompts often start with a transition, for example "Interesting, let me ask you another question...". The core question usually appears in the second half of the clip. Train yourself to ignore the preamble and focus on key question words (what, why, how, describe, etc.). Visually, the audio waveform on screen can help, because the main question often lines up with a big spike.⁹
5. Efficient Response Structure: To be clear and score well within the time limit, keep responses direct and well structured. Answer Directly: Start immediately with a direct answer to the question. Avoid time wasting introductory phrases like "That's a very interesting question".⁹
6. Explain Why: Right after your direct answer, give a reason or explanation. Use linking phrases like "The main reason is that..." or "This is because...".⁹
7. Provide an Example (if time permits): If you can, add a brief, specific example to back up your reason.

Preparation with flexible stories: Memorizing full scripts backfires and will be penalized. A better approach is to prepare 3 to 4 versatile "go to stories" or core ideas tied to common themes, for example a personal achievement, a learning experience, or a work related challenge. Practice these stories as flexible frameworks, not rigid scripts, and change them quickly to fit a variety of specific questions.¹⁰

Practice Now:

Try this interactive speaking exercise to practice the conversational skills you'll need for the DET:

• Practice Interactive Speaking →

Common mistakes and how to avoid them

Test-takers should watch for several common mistakes that can lower their score in this section:

1. Robotic or Rehearsed Delivery: AI scoring models detect unnatural speech patterns. Responses that sound memorized or recited will score lower. Practice speaking with natural intonation, pace, and variation.¹⁰
2. Irrelevant or Off-Topic Responses: Forcing a pre-prepared story into an answer that does not fit is a major error. The response must address the specific question directly and stay on point. The test checks your ability to adjust, not just to recite.¹⁰
3. Ineffective Time Management: Rushing and giving a response that is too short, under 20-25 seconds, or speaking too slowly and getting cut off can hurt your score. Practice with a 35-second timer to develop an internal clock for consistent performance.¹⁰

Guide to the Interactive Listening section

The Interactive Listening section has been updated to give a broader, more realistic measure of academic listening and interaction skills. The task appears twice on the test, and now runs through several stages, starting with a listening comprehension check and then moving on to the conversational part.¹¹

Stage 1: The audio scenario and "Listen and Complete" task

The biggest change in this section is replacing the initial written scenario with an audio one.¹¹ That raises the cognitive load, because test takers must process information by ear instead of by sight. Reading lets you refer back easily. Listening forces active processing, holding details in short term memory, parsing spoken language, and then using that information. The change better reflects real academic situations, such as understanding a professor's instructions before a class discussion. It thus gives a stronger test of integrated listening skills under time pressure.

The mechanics of this first stage are as follows.

1. Format: The section starts with an audio recording that sets up a hypothetical academic scenario, for example a conversation with a professor or a peer. After listening, the test taker sees 3 to 4 fill in the blank questions that check key details from the scenario. This first part is called "Listen and Complete".²
2. Audio Replay: Test takers can replay the scenario audio an unlimited number of times. The audio player stays available throughout the "Listen and Complete" part and the later "Listen and Respond" stage. The timer keeps running during replays. That forces a choice between confirming understanding and saving time.¹¹
3. Scoring Nuances: Scoring focuses on meaning rather than exact form. Minor spelling, punctuation, or grammar mistakes do not lose points if the intended meaning is clear.¹¹ Test takers do not have to transcribe answers word for word from the audio. Paraphrasing is allowed and may be needed to fit the answer box.¹¹ Keep answers short and direct. They usually require fewer than 10 words.¹¹

Stage 2: The "Listen and Respond" conversation

After the initial comprehension check, the task moves to the "Listen and Respond" stage, which is the main part of the simulated conversation. It stays structurally similar to the version of the test before the update.¹² It has about 5 to 6 conversational turns. In each turn, the test-taker hears a prompt from one of the AI characters. Then they pick the best response from a list of multiple-choice options to move the conversation forward logically.⁵ The main rule in this stage is that the audio for each individual turn can be played

only once, requiring focused, real time listening comprehension.¹¹

Stage 3: The 75 second "Summarize the Conversation" task

The final stage of the Interactive Listening section asks the test-taker to sum up the whole interaction. The task's format is the same as in the previous version of the test.¹² The candidate is given

75 seconds to write a concise summary of the conversation they just participated in.⁵

A strong, high-scoring summary must capture the dialogue's main points. One good strategy is to structure the summary around three key questions:

1. Who was involved in the conversation?
2. What was the primary purpose or topic of the conversation?
3. What were the key outcomes, decisions, or pieces of information exchanged? ¹¹

Using a simple three-sentence template that covers these points helps you produce a complete, clear summary within the short time limit.¹¹

Timing and pacing strategies

Test-takers must know the timing for the Interactive Listening section. There are two such sections on the full test.¹¹

1. One shared timer of 6 minutes and 30 seconds covers the combined "Listen and Complete" and "Listen and Respond" stages.¹²
2. The "Summarize the Conversation" task has its own separate 75 second timer.⁵

This shared timing for the first two stages means you must manage time carefully. Aim to spend no more than 2.5 minutes on the "Listen and Complete" questions. That leaves at least 4 minutes for the more mentally demanding "Listen and Respond" portion. In that part, you cannot replay audio prompts, and you must consider response options carefully.¹¹

Practice Now:

Try this interactive listening exercise to practice the conversation comprehension skills you'll need for the DET:

• Practice Interactive Listening →

Scoring, weighting and subscores

Understanding how the new interactive tasks are scored and how they affect the overall score profile helps you prepare for the test. The DET's scoring system shows how language skills work together, and the new tasks are heavily weighted toward subscores that measure these combined abilities.

Interactive tasks and the four integrated subscores

The DET gives an overall score from 10 to 160, based on performance on questions that test the four language skills: Reading, Writing, Listening, and Speaking.¹³ Besides the overall score, the test reports four integrated subscores, calculated by averaging pairs of these basic skills ¹³:

1. Literacy: Reading + Writing
2. Comprehension: Reading + Listening
3. Conversation: Speaking + Listening
4. Production: Writing + Speaking

The new interactive tasks are multifaceted, so they add to several of these subscores at the same time ¹⁴:

1. Interactive Speaking adds to the Comprehension subscore, since it requires listening to a prompt, and to the Conversation subscore, since it involves listening and speaking.
2. Interactive Listening is more complex, and it adds to the Comprehension subscore, because test takers listen to the scenario and prompts, to the Conversation subscore, because they take part in a dialogue, and to the Production subscore, because they write the final summary.

This integrated scoring matches Duolingo's belief that testing combined skills shows real-life language use better than testing skills separately.¹³

AI assessment criteria explained

AI grading engines score all open-response items, including the spoken responses in Interactive Speaking and the written summary in Interactive Listening.¹⁵ They judge responses on several language measures. They check content relevance, discourse coherence, lexical diversity (vocabulary), grammatical accuracy and complexity, fluency, and pronunciation.¹⁰

AI powered scoring also drives the test's computer adaptive design. For many question types, the system checks a test taker's performance in real time and changes the difficulty of later questions. A correct or high quality response usually leads to a harder question. An incorrect or low quality response usually leads to an easier one.¹⁵ This adaptive system tailors the test to each individual's ability level. That gives a more precise and efficient measure of their proficiency.

Interpreting your score profile

The introduction of these highly interactive tasks will likely affect the Conversation and Production subscores, which many test-takers find to be the most challenging areas.¹⁴ A close look at a score report can give useful diagnostic information. For example, a relatively low Conversation subscore may point to weakness in understanding spoken language and responding spontaneously. Those skills are measured directly and strongly by the new Interactive Speaking and Listening formats.

Because the precise weighting and mechanics of the scoring algorithm remain proprietary, test-takers cannot "game" the system by focusing on a single question type. The new tasks can affect multiple subscores from a single performance. Preparation therefore needs to cover several skills at once. The subscores are the best diagnostic tool available. Instead of only practicing individual question types, candidates should use their subscore profile to find and address weaknesses in the combined skills the test measures. For example, a student seeking to raise a low "Production" score should do exercises that combine writing and speaking, since those mirror the test's demands.

Score Analysis:

Learn how to interpret your DET scores and understand the registration process in our detailed From Registration to Results Guide.

Comparing the DET with other major tests

The 2025 DET updates do more than refine an existing product. They mark a major change in assessment philosophy from other major English proficiency tests. Duolingo now fully embraces an interactive model powered by AI. The company is acting as a digital first leader. It challenges the long standing methods used by its competitors.

Comparing speaking test formats

A comparative analysis of the speaking sections of the world's leading English tests shows a clear split in approach. The DET now favors a model that uses many data points from short, fast, AI-mediated interactions. In contrast, traditional tests like IELTS and Cambridge still value sustained, nuanced speech evaluated by a human expert as a whole. TOEFL iBT sits between these approaches. It uses a highly structured, academic, semi-direct (recorded) format.

This split reflects a deeper teaching debate. Should language ability be measured by handling a series of quick, digital exchanges, or by developing and keeping complex ideas in a longer, traditional conversation? The DET model fits the fast-paced communication common in the 21st century. Traditional models match the demands of a university seminar. The long-term success of the DET's new format will depend on which model institutions ultimately find to be a more reliable predictor of academic success.

The following table gives a detailed comparison of these distinct models.

Table 5.1: Comparative Framework of Major English Proficiency Speaking Tests. Data compiled from sources.¹

Comparing listening task formats

The DET's revamped Interactive Listening section also stands out from its competitors. While traditional tests like IELTS and TOEFL mainly test passive listening by having candidates listen to lectures or conversations and then answer multiple choice or short answer questions, the DET's format is meant to test interactional listening. By having the test taker make choices that steer a conversation and then summarize the interaction, the DET checks not just whether a candidate understood the content, but also whether they can use that understanding to take part in a goal oriented dialogue.

Implications for test takers and institutions

These differences in assessment philosophy and method have clear practical effects. For test-takers, the choice of exam may now hinge more on their specific communication strengths. A candidate who does well at quick, spontaneous speaking but has trouble building a long, structured argument might prefer the DET. A candidate with the opposite profile might do better on IELTS or TOEFL.

For the thousands of institutions that accept these scores, the DET's change raises questions about the comparability of results and the validity of different assessment models. They must now consider whether an AI assessed score that shows proficiency in rapid, adaptive conversation is equivalent to a human assessed score that shows proficiency in sustained academic discourse. The answer will likely depend on the specific linguistic demands of their academic programs.

Technical requirements and test security

Alongside the changes to question types, the 2025 DET update introduces stricter, more complex security measures and technical requirements. These measures respond directly to the security challenges of high-stakes, remote, on-demand testing. They are meant to increase the validity and trustworthiness of the results.

ID verification and secondary camera setup

The test's onboarding process was redesigned to put security and identity checks first.⁴ Test takers must now go through a multi step setup before the adaptive test starts:

1. ID Verification First: After initial system checks, the first graded step is ID verification. Test takers use a secondary device, a phone or tablet, to scan a QR code on their computer screen and enter a verification code.⁴ They then select their nationality and ID type and take a photo of their government issued ID.⁴
2. Mandatory Secondary Camera: A new requirement is a secondary camera for the whole test. After ID verification, the rear camera on the secondary device turns on. The device must sit so it clearly shows the test taker, their computer screen, and their keyboard for the exam.⁴
3. Automated Environment Validation: To check positioning, the system uses an improved device detection feature. A white box appears on the computer screen, and the system uses it to identify and confirm the camera's field of view. The test cannot proceed until the secondary camera is placed to capture that area, preventing test takers from starting in an insecure environment.⁴

This stronger security protocol is part of a technological "arms race" against academic dishonesty. These measures help keep the test fair and keep accepting institutions' trust, but they come at a cost. They add technical complexity to the setup. Some users report significant lag and trouble connecting cameras.⁴ A test taker's success now depends not only on English ability, but also on access to reliable technology and on getting through a complex technical process under pressure. This could create new barriers for candidates with older devices, shaky internet connections, or lower tech skills.

Updated system and environment requirements

To take the DET after July 2025, candidates must make sure their setup meets the following requirements:

1. Computer: A laptop or desktop running Windows 10 or macOS 10.15 or later is required. The test does not support Chromebooks, iPads, or other tablets. The computer must have a working front facing camera, microphone, and speakers.
2. Secondary Device: You must use a phone or tablet with a working camera for the secondary camera setup.
3. Internet Connection: You need a stable internet connection. Duolingo recommends at least 2 Mbps download and 1 Mbps upload. Some third party preparation platforms recommend a stronger 5 Mbps connection for safety.
4. Testing Environment: Take the test in a quiet, private, well lit room. No other people may be present, and the test taker's face and ears must be visible at all times.

Technical Setup:

Get detailed instructions for setting up your testing environment in our DET Technical Setup Guide.

Conclusion and future outlook

Summary of findings and DET position

The Duolingo English Test updates of July 2025 are more than an incremental refinement. They mark a major, decisive step toward a new approach to language assessment. Duolingo retired tasks that measure isolated skills and added AI driven adaptive components that simulate real time conversation. The company has committed to a model that focuses on measuring integrated, interactional competence. The introduction of Interactive Speaking, in particular, with its dynamic, AI selected follow up questions, is a first of its kind innovation in large scale, high stakes testing. While the overall difficulty level of the exam is reported to remain consistent, the nature of the challenge for test takers has fundamentally shifted. Success now depends less on rote memorization or structured argumentation, and more on the ability to listen actively, think quickly, and communicate spontaneously and coherently in a series of rapid exchanges.

Recommendations for test takers and educators

These changes require a shift in how test-takers and the educators who support them prepare.

1. For Prospective Test-Takers: Preparation must go beyond practicing fixed question types. Focus on building conversational fluency and active listening. Practice timed, spontaneous speaking on a wide range of topics, and do it regularly. Because the security setup is more complex, candidates must know the required hardware and software well before test day. Technical problems could invalidate their results.
2. For Language Educators: Course plans that prepare students for the DET should include more interactive, timed, and spontaneous communication activities. These activities should mirror the demands of the new tasks. Use the DET's detailed subscore report to find specific weaknesses. Instead of focusing on overall scores, teachers should direct instruction to improve specific integrated skills, such as the "Conversation" and "Production" subscores, which are strongly influenced by the new interactive components.

Ready to Practice?

Start building your skills with our comprehensive Practice Question Library featuring Interactive Speaking, Listening, and Writing exercises designed for the 2025 DET format.

Future outlook for AI in assessment

The 2025 DET updates are a key moment in the ongoing changes in educational technology. The "Constrained AI" model used in Interactive Speaking limits the AI to a curated set of prompts instead of letting it generate language freely. This is a clever and necessary compromise. It balances interactivity with the psychometric demands of standardization and fairness. This model is likely a stepping stone. As generative AI matures and becomes more reliable, future versions of the DET and other language tests could shift toward more fully generative conversational assessments. For now, the July 2025 Duolingo English Test signals the direction of language assessment: one that is more digital, more personalized, more interactive, and increasingly powered by artificial intelligence.

Works cited