How to Transcribe Interviews for Research, Journalism, and Hiring

An interview is only as useful as its record. The moment the conversation ends, the exact words start to fade — and with them, the phrasing that tells you what the person really meant. A transcript preserves the words, the pauses, and the order in which things were said, so you can quote accurately, code themes honestly, and check your memory against the record instead of trusting it.

This guide shows you how to transcribe interviews well: how to record for clean results, how to handle the details that matter most in interview work (exact wording, speaker turns, sensitive material), and why a transcript is worth the effort for researchers, journalists, and hiring teams. You can transcribe the recording with AudioToText and download the result as text, subtitles, or timestamped JSON, instead of typing it out by hand.

Why Interviews Need Transcripts

An interview is a data source. For a researcher, the transcript is the evidence the analysis rests on. For a journalist, it is the source of every quote and the defence against a complaint that a quote was invented. For a hiring team, it is the record that lets a second reviewer judge the same answers the first interviewer heard. In all three cases, the transcript has to be faithful — not a summary, and not a cleaned-up version of what the person "probably meant".

Memory is a poor substitute. Studies of eyewitness memory show how quickly detail is lost and how easily it is reshaped by what we expect to hear. The same happens with interviews: a week later, you remember the gist and lose the wording, and the wording is often the finding. A participant who says "I guess the system sort of works" is telling you something different from one who says "The system works." A summary would flatten both to "the system works". A transcript keeps the difference.

There is also a practical reason: a transcript is searchable and shareable. A research team can code the same text, a newsroom editor can check a quote in seconds, and a hiring panel can compare candidates answer by answer. None of that is possible with audio alone, because audio cannot be skimmed, searched, or pasted into a report.

Why Exact Wording Matters

In interview work, the words are the data. A researcher coding themes needs to know whether a participant said "frustrating" or "annoying" — those are different themes. A journalist needs the quote verbatim, including the hesitation that shows the speaker was unsure. A hiring manager needs to know whether the candidate said "I led the project" or "I was on the project" — those are different claims.

This is why interview transcripts are usually kept close to verbatim: the false starts, the "ums", and the half-finished sentences stay in, at least in the working transcript. You can clean the text later for publication, but the working copy should show what was actually said. If you edit as you transcribe, you are analysing before you have the data, and you will lose the material that does not fit your first impression.

The U.S. Library of Congress, which holds large oral history collections, treats the transcript as the durable record of an interview and the audio as its companion — the text is what researchers search, cite, and quote. See the Library of Congress for how oral histories are preserved and used.

How to Record an Interview for Transcription

A good interview transcript starts with the recording. Three things matter most: consent, microphone placement, and the room.

Step 1: Get consent, on the record

Before you ask the first question, say that you are recording and why, and ask for agreement — on the recording itself. "I'm recording this so I can quote you accurately. Is that okay?" That sentence protects you and the interviewee. It also sets the tone: the person knows their words will be kept, so they speak more carefully, which is usually a good thing. For research and journalism, written consent is often required as well; check the rules that apply to your work before the interview, not after it. Our guide to transcription privacy explains consent and data handling in more detail.

Step 2: Put the microphone close to the person

The interviewee's voice is the one that matters most, and it is usually the quietest in the room — they are thinking, speaking softly, or turning away. A small lapel microphone, or a phone placed on the table between you, beats a laptop microphone across the room every time. If you interview by phone or video call, record the call in the app rather than holding a second device up to a speaker. Test the setup with thirty seconds of talk before the real questions start, and listen back. A test takes a minute; a ruined recording costs the whole interview.

Step 3: Choose a quiet room

Background noise is the enemy of accuracy: café music, a humming air conditioner, traffic through an open window. A quiet office or a parked car is better than a busy café. If you must meet in public, sit away from the speakers and the espresso machine, and keep the microphone close. Tell the interviewee at the start that you may ask them to repeat a line if a noise covers it — people accept that gladly when they know why.

Step 4: Say names and spell them

Start the recording with an introduction: the date, the place, and the full names of everyone present, spelled out. "Today is May 6, 2026. I'm talking with Ana Kowalski, K-O-W-A-L-S-K-I, in Chicago." That line fixes the two most common interview transcription errors — names and dates — before they happen. It also gives the transcript a clean header for filing.

Step 5: Let people finish, then ask the follow-up

Overlapping speech is the hardest thing to transcribe. Interviewers cause most of it, by jumping in with the next question while the answer is still coming. Wait for the pause. Count to two in your head. Then ask. The pause also helps the interviewee: silence is the best follow-up question there is, because people fill it with the thing they were holding back.

How to Transcribe an Interview Step by Step

Step 1: Transcribe everything, in order

Work through the recording from start to finish, in order. Do not skip the small talk at the start — it often contains the context that explains an answer later. Keep the speaker turns in order, and mark who is speaking. A simple format works: Interviewer: and Ana: on their own lines, with a timestamp every few minutes so you can find the passage in the audio again.

Step 2: Keep the words, mark the rest

Keep the words verbatim. Mark the non-words in brackets so a reader can tell them apart from speech: [laughs], [pause], [inaudible 12:34], [crosstalk]. A pause of more than a few seconds is worth marking, because a long pause before an answer means something — hesitation, thought, or discomfort — and a reader of the transcript should be able to see it.

Step 3: Check names, numbers, and quotes

After the first pass, read the transcript through and check every name, every number, and every passage you plan to quote. Use the timestamps to jump to the audio and listen to those exact moments. A name spelled wrong in a published quote is embarrassing; a number quoted wrong is worse. This check takes ten minutes and is the difference between a working transcript and a publishable one.

Step 4: Clean a copy for readers, keep the original

For publication or sharing, you usually want a cleaned copy: remove the "ums", tidy the false starts, and fix obvious slips — but only with the interviewee's agreement, and only where the meaning is unchanged. Keep the verbatim original as the master copy. If anyone questions a quote, the original is your evidence. Never clean the only copy.

How Different Interview Types Change the Work

Research interviews. The transcript is the data, so verbatim matters most. Code themes from the transcript, not from memory. Keep a consistent format across interviews in the same study so coding is comparable, and store transcripts with the same naming scheme (date, participant code, topic). If your study has ethics approval, follow its rules on storage, anonymisation, and deletion — and check whether your institution requires transcripts to be stored on approved systems. The National Institutes of Health publish guidance on handling interview and qualitative data; see NIH for research standards that apply to federally funded work.

Journalism interviews. Speed matters, but accuracy matters more. Transcribe the same day, while the context is fresh, and flag any quote you plan to use. Send sensitive quotes back to the source for checking when your outlet's policy requires it. Keep the recording and the transcript until well after publication — they are your record if a quote is challenged.

Hiring interviews. The transcript is a fairness tool: it lets a second reviewer judge the same answers, and it protects the company if a decision is challenged. Be careful with consistency — transcribe the same way for every candidate, and do not clean one candidate's transcript more than another's. Tell candidates they are being recorded and transcribed, and keep the transcripts only as long as your hiring policy requires.

Oral history and life-story interviews. These are long, meandering, and full of names, places, and dates from decades ago. Accuracy on names matters enormously, because the transcript becomes a historical record. Ask the interviewee to spell names, and check place names against a map afterwards. The Library of Congress oral history guidance is a good model for format and archiving.

What a Good Interview Transcript Looks Like

A transcript that follows this shape can be read, searched, coded, and quoted without further work. It is also honest: a reader can see where the record is uncertain, instead of being misled by a confident-looking guess.

Common Mistakes to Avoid

Guessing at unclear words. If you cannot hear a word, mark it [inaudible] with the timestamp and move on. Guessing turns an honest gap into a false quote. If the passage matters, ask the interviewee to repeat it — during the interview, or in a follow-up.

Cleaning the transcript too early. Removing "ums" and false starts feels like tidying, but it is editing, and editing before analysis throws away data. Clean a copy for readers; keep the verbatim master.

Transcribing from memory. Do not write the transcript from your notes of the interview. Transcribe from the recording, with the audio playing. Notes are a guide to what to listen for, not a substitute for listening.

Losing the recording. Keep the audio until the transcript is checked and the work is done. A transcript is a claim about what was said; the recording is the proof. Store both in the same place, with the same name, so they cannot be separated.

Forgetting the consent. If you did not get consent to record, you may not be able to use the transcript at all — and in some places, recording without consent is unlawful. Consent first, questions second, every time.

How to Work Faster Without Losing Accuracy

Transcribing by hand takes roughly four hours per hour of audio. Automatic transcription cuts that to minutes, and the time you save goes into checking rather than typing. A good workflow is: transcribe automatically with AudioToText, read the transcript while listening at higher speed, fix the names and numbers, and mark the unclear passages. That gives you a checked transcript in about an hour per hour of audio — and the checking is the part that matters.

For long research projects, keep a shared list of names, places, and terms that recur across interviews, with their correct spellings. Feed that list to whoever checks the transcripts. The tenth interview will be checked twice as fast as the first, because the names are already known.

For journalism, transcribe the same day and highlight the quotes you plan to use before you write the story. For hiring, transcribe every candidate the same way and store the transcripts in the hiring file, not in a personal inbox. See pricing to find the plan that fits your interview volume — the Free plan covers 60 minutes a month, and Starter covers 10 hours.

Frequently Asked Questions

How long does it take to transcribe an interview?

By hand, about four hours per hour of audio. With automatic transcription, the transcript itself takes minutes, and checking it takes about an hour per hour of audio. The checking is where the accuracy comes from — do not skip it for anything you plan to quote or publish.

Should I include "um" and false starts?

In the working transcript, yes — they are part of what was said, and they can carry meaning. In a published quote, you can clean them out, with the interviewee's agreement, as long as the meaning is unchanged. Keep both versions.

What if the interviewee asks to go off the record?

Stop the recording, or mark the passage clearly and do not transcribe it. "Off the record" means different things to different people, so agree on what it means before you continue: not to be quoted, not to be used at all, or usable without attribution. Write the agreement down.

Can I transcribe a group interview?

Yes, but it is harder: overlapping speech and similar voices confuse both software and humans. Ask each person to introduce themselves, keep the group small, and remind people to speak one at a time. Expect to spend more time checking a group transcript than a one-on-one.

How do I handle sensitive material?

Store transcripts of sensitive interviews securely, limit who can read them, and anonymise names and identifying details before sharing. Delete the recording and the transcript when your retention rule says to — and tell the interviewee how long you will keep them.

Key Takeaways

• Record with consent, on the record, with the microphone close to the interviewee in a quiet room.
• Transcribe verbatim and in order, marking pauses, laughter, and inaudible passages instead of guessing.
• Check every name, number, and planned quote against the audio using the timestamps.
• Keep a verbatim master copy; clean a separate copy for readers, and never clean the only copy.
• Store the recording with the transcript until the work is done — the recording is the proof behind the text.

!Interview transcription workflow

Image description: The image should show an interview scene as a simple line drawing: two people facing each other across a table, a small microphone between them, and a sound wave running from the microphone to a document page on the right. The document shows speaker labels and timestamps. A caption reads "Record with consent, transcribe verbatim, check every quote."

References

• Library of Congress — oral history collections and guidance: https://www.loc.gov/
• National Institutes of Health — research standards and data handling: https://www.nih.gov/
• U.S. Department of Education — resources on research ethics: https://www.ed.gov/