TOEFL Take an Interview: can I use a template?

What does the Take an Interview task ask you to do?

In Take an Interview, you hold a simulated conversation with a prerecorded interviewer and answer four questions aloud. Page 19 of the Test Overview gives situations such as applying for a scholarship or taking part in a research study as examples, and gives 45 seconds to answer each question. The earlier questions ask about facts and personal experience, and the later ones ask you to state and support an opinion on a broader issue (same page).

The same page lists four things this task measures.

• Whether you answer the question with appropriate, well-organized elaboration
• Whether you keep an even, conversational speaking pace
• Whether you are easy to understand and use rhythm and intonation to convey meaning
• Whether you use a range of vocabulary and grammatical structures effectively and accurately

Page 2 of ETS's "TOEFL iBT Test: 2026 Update Test Blueprint and Specifications Document" (hereafter the Specifications Document, no publication date, checked 2026-10-30) describes the ability Speaking aims to measure as answering interview-style questions by speaking "spontaneously and meaningfully". Page 7 counts Take an Interview as 4 items, and page 18 gives a maximum of 5 points per item. Table 2 on page 30 of ETS's "TOEFL iBT Technical Manual" (hereafter the Technical Manual, ETS Research Report RR-25-12, October 2025, checked 2026-10-30) gives the same: 4 items, 0 to 5 points each. Table 1 on page 21 of the Test Overview gives the Speaking raw score (the score before it is converted to a band) as 0 to 55, and 20 of those points belong to this task (4 items times 5 points, our calculation).

Every score in this article is on the 0 to 5 scale for a single question. That is a different scale from the Speaking 1 to 6 band on the score report; for how to read the bands, see "How do TOEFL 1 to 6 scores compare with 0 to 120 scores?". To look at both Speaking tasks at once, read "How do I prepare for TOEFL Speaking: Listen and Repeat and Take an Interview?", and if you missed a sentence in the earlier task, read "What do you do if you miss the first TOEFL Listen and Repeat sentence?".

Which row of the scoring criteria does an answer built on a memorized template reach?

In our view, a memorized answer that never actually answers the question meets the wording of the 2-point row. The 2-point row on page 28 of the Test Overview leads with "a response that attempts to address the question but is not meaningfully or intelligibly supported", and gives as an example an answer that is only loosely connected to the interviewer's question and has little relevant elaboration.

The question and answers below are examples the Unlock Prep editorial team wrote for illustration. We did not copy any ETS item. The score rows are our judgment from holding each answer against the wording of the scoring criteria on pages 27 to 28 of the Test Overview, and ETS has not scored them. Because the answers are written out, we cannot judge the pace, pronunciation, rhythm or intonation lines, so we compared only the content line and the vocabulary and grammar lines.

Our question (a later, opinion question): Some universities require all first-year students to live on campus. Do you think this is a good policy? Why or why not?

Answer A: a memorized template pasted in as is (our view: close to the 2-point row)

• What was said: That's a great question, and there are many different perspectives on this issue. Personally, I think it has both advantages and disadvantages. On the one hand, it can be helpful for many people. On the other hand, it can also cause some problems. For example, when I was in high school, I learned that every experience teaches us something. In conclusion, this is a very interesting topic.
• What decided it: The answer never says whether requiring first-year students to live on campus is a good policy. It also never fills in what the "advantages", "problems" or "experience" are. Because the sentences are memorized, they can be spoken without breaks.
• Matching wording: the 2-point row's "attempts to address the question but is not meaningfully or intelligibly supported" and "only loosely connected to the interviewer's question, with little relevant elaboration". It does not reach "addresses the question", the lead sentence of the 5-, 4- and 3-point rows. However, the 2-point row's intelligibility, grammar and vocabulary lines do not fit this answer, so this judgment rests on the lead sentence and the content line.

Answer B: a short answer attached to a memorized opener (our view: close to the 3-point row)

• What was said: That's a great question, and there are many different perspectives on this issue. Personally, I think it is a good policy for first-year students to live on campus. Because they can make many friends. So in conclusion, I think it is a good policy.
• What decided it: The answer states agreement, but the reason is a single line and is not extended with an example or an experience. The first and last sentences would fit any question, so that much of the 45 seconds went to something other than elaborating the answer (our view).
• Matching wording: the 3-point row's "addresses the question but lacks elaboration or clarity" and "mostly on topic but may have relatively limited elaboration".

Answer C: only the framework is reused, and the words fit this question (our view: the content, vocabulary and grammar lines reach the 4-point row or higher)

• What was said: I think it's a good policy, at least for the first semester. Living in a dorm puts new students close to the library and to each other, so it's easier to settle into a routine. When I moved into a school dormitory in high school, having classmates down the hall made it much easier to ask questions and join study groups. That said, I'd let students with family responsibilities apply for an exception.
• Framework: a one-sentence answer, then a reason, then one experience of your own, then a one-sentence qualification. Each slot is filled with words from this question (dorm, library, first semester), and connectives such as so and That said link the reason and the qualification.
• Matching wording: the 4-point row's "on topic and elaborated, but may lack effective connectives between sentences" and the 5-point row's "on topic and fully elaborated". Whether it reaches the 5-point row depends on the pace, pronunciation and intonation lines, which the written text alone cannot decide.

Answer A also follows an order: opener, two sides, example, conclusion. In our view, what separated the three answers was not the order of what was said but whether each slot in that order was filled with content for this question.

Where do templates help, and where do they not?

In our view, a framework helps you set the order of your answer and its connectives, but it cannot supply content and words that fit the question. Positions in industry guides split three ways: for (set a structure and practice it from memory), framework only (fixed structure, flexible wording), and optional (keep templates as separate material or treat them only as a support). This is based on four English-language prep guides we read on 2026-10-29, and the one among them that covered Speaking was on the "for" side. Unlock Prep's position at the top of this article is closest to "framework only".

There are four places in ETS documents that bear on this position.

• Where a framework helps: The 4-point row describes answers that are on topic and elaborated but may lack effective connectives between sentences (Test Overview, page 27). Table 6 on page 38 of the Technical Manual also has an Organization column among its examples of features the scoring engine looks at for this task, listing discourse coherence and the use of discourse connectives (words such as so and however that link sentences).
• Where memorized sentences do not help, 1: Page 16 of the Specifications Document groups the evidence statements for the abilities Speaking aims to measure into three CEFR (Common European Framework of Reference for Languages) bands. The basic (A1 to A2) band includes "answers simple questions with short, formulaic responses", and the proficient (C1 to C2) band includes "answers complex questions fluently and spontaneously". These are level descriptions, not scoring criteria, but in our view memorized sentences that would fit any question look closer to the basic band.
• Where memorized sentences do not help, 2: Page 20 of the Test Overview says of the 5-point sample answers that pauses and fillers such as um and uh are natural and do not get in the way of meaning. What the 3-point row describes is an answer broken up by frequent or long pauses, with frequent fillers (page 28). We find little reason in these scoring comments to fill time with a memorized opener just to avoid a few pauses (our view).
• Where memorized sentences do not help, 3: Page 14 of ETS's "TOEFL iBT Information Bulletin" (hereafter the Information Bulletin, effective July 2026 to June 2027, no publication date, checked 2026-10-30) says that if there is evidence that a speaking or writing response contains text substantially similar to responses from other TOEFL tests, ETS may, in its judgment, cancel the score. The same paragraph also covers responses that repeat words or ideas from the internet or other sources without credit, on the grounds that such responses do not show the independent speaking ability the test is meant to measure. In our view, saying a widely shared template sentence word for word can overlap with other test takers' responses.

In our view, there are three ways to use a framework.

1. Keep separate frameworks for experience questions and opinion questions. As page 19 of the Test Overview describes, the earlier questions ask about facts and experience, and the later ones ask for an opinion and support.
2. Memorize only the slot names of the framework (answer, reason, example, qualification), not the sentences that go in the slots.
3. Open with a sentence that answers the question. If you use an opener at all, keep it to one short phrase.

Does the AI scoring engine detect memorized answers?

The ETS documents we checked do not say whether the engine separately detects memorized answers. What we confirmed goes only as far as who scores and with what; score cancellation for responses substantially similar to other test takers' responses is covered separately from scoring, on page 14 of the Information Bulletin (see "Where memorized sentences do not help, 3" above). The Writing tasks Email and Discussion, and Speaking, use both the AI scoring engine and human raters, and responses the AI is not confident about are scored by humans (Technical Manual, pages 35 and 39).

Page 37 of the Technical Manual says the Speaking scoring engine analyzes features that represent fluency (such as words per minute), intelligibility (such as pronunciation accuracy), grammatical accuracy and coherence (such as discourse transition cues), and is trained on responses scored by humans. Table 6 on page 38 gives examples of the features the engine uses for this task in four columns.

• Fluency: speaking rate, length of continuous speech without breaks, number of pauses, number of hesitations
• Intelligibility: pronunciation accuracy, naturalness of speech rhythm and prosody (such as syllable stress)
• Vocabulary and grammar: vocabulary diversity, use of less common words, grammaticality, grammatical accuracy
• Organization: discourse coherence, use of discourse connectives

Table 6 lists these features as "examples", so it cannot be read to mean that the engine ignores anything not on the list. Table 7 on page 39 of the same document reports, for a sample of 1,521 Speaking responses each scored by two human raters, a correlation (Pearson r) of 0.89 between the average human score and the AI score, and 0.96 between the human scores for the same response.

The engine is trained on human scores, and responses it is not confident about are scored by humans. So in the ETS documents we checked, we found no grounds to expect that an answer that does not answer the question by human scoring criteria would score high only with the engine.

What has ETS not disclosed?

In the Test Overview, the Specifications Document, the Technical Manual, the Information Bulletin and the ets.org Speaking page, we did not find the following, or found only a method and no table (checked 2026-10-30).

• Whether there is separate preparation time after you hear a question and before you answer. Page 19 of the Test Overview gives only the 45 seconds of response time.
• How an answer that does not fill the full 45 seconds affects the score. The scoring criteria divide the rows not by seconds or length but by elaboration and clarity.
• Whether you can hear a question again.
• Whether human raters or the engine separately flag memorized answers or template sentences, or reflect them in the score.
• What standard and method are used to find the "substantially similar" responses on page 14 of the Information Bulletin.
• The table that converts the raw score total (0 to 55) for the four questions and the seven Listen and Repeat sentences into the Speaking 1 to 6 band. Page 35 of the Technical Manual describes the method: the conversion uses weighted equipercentile linking, which adjusts for small differences in difficulty between test forms. The conversion table itself is not in the ETS documents we checked.

We found no numbers that differ between ETS documents. The Test Overview, the Specifications Document and the Technical Manual agree on 4 items, a maximum of 5 points per item and a Speaking raw score of 0 to 55. We did not cross-check numbers in company or blog articles. Page 3 of the Specifications Document still carries a note that small revisions may be made before launch.