Interview Preparation & Transcription with AI
Interviews are the backbone of great journalism and compelling content. AI can help you prepare smarter questions, transcribe conversations instantly, and extract the most powerful quotes -- without losing the human connection that makes interviews work.
What You'll Learn
- How to use AI to research interview subjects and prepare targeted questions
- Tools and techniques for AI-powered interview transcription
- How to extract key quotes and themes from transcripts
- Best practices for maintaining authenticity when AI assists your interview workflow
AI-Powered Interview Preparation
Walking into an interview well-prepared is the difference between surface-level quotes and genuine revelations. AI helps you prepare in a fraction of the time.
Researching Your Subject
Before any interview, use this prompt:
I'm interviewing [person name], who is [their role/relevance]. The interview
is for a [type of piece] about [topic].
Give me:
1. Key points from their recent public statements, articles, or interviews
2. Their known positions on [relevant issues]
3. Any controversies or criticisms they've faced
4. What they've been asked in recent interviews (so I can avoid repetition)
5. Gaps in their public statements -- things they haven't addressed
Generating Interview Questions
Once you have background, generate questions that go beyond the basics:
Based on the research above, generate 15 interview questions organized as:
- 3 warm-up questions (easy, establishes rapport)
- 5 core questions (directly about the topic, specific and probing)
- 4 follow-up questions (anticipated based on likely responses)
- 3 challenging questions (addresses contradictions, controversies, or gaps)
Make each question specific enough that it can't be answered with a
generic talking point. Reference specific statements, data, or events.
Pro tip: The best AI-generated questions reference specific things the subject has said or done. "Tell me about your approach to leadership" is generic. "In your 2024 report, you said X but your company did Y -- how do you reconcile that?" is the kind of question that produces real answers.
Preparing for Different Interview Styles
For investigative interviews:
I'm conducting an accountability interview with [subject] about [issue].
They're likely to deflect or give vague answers. For each of my core
questions, give me 2-3 follow-up questions that pin down specifics
if they try to avoid the question.
For profile/feature interviews:
I'm writing a profile of [subject]. I want to reveal who they really are,
not just their public persona. Generate questions that:
- Explore formative experiences and turning points
- Reveal their decision-making process
- Uncover surprising aspects of their character
- Get specific stories rather than abstract answers
Transcription: From Recording to Text
AI transcription has become remarkably accurate and saves hours of manual work.
Recommended Tools
Otter.ai -- Best for real-time transcription during live interviews. Features include speaker identification, searchable transcripts, and AI-generated summaries. Free tier allows 300 minutes per month.
Descript -- Best for post-interview editing. Upload your recording, get a transcript, and edit the audio by editing the text. Removes filler words automatically.
OpenAI Whisper -- Most accurate for difficult audio (background noise, accents, technical terminology). Free and open-source, available through various apps and web interfaces.
Transcription Best Practices
-
Record with quality -- AI transcription is only as good as your audio. Use a dedicated recording app, not your phone's built-in recorder. Place the recording device close to the speaker.
-
Identify speakers -- Most tools auto-detect speakers, but review and correct speaker labels before using quotes.
-
Don't delete the recording -- Keep the original audio even after transcription. If a quote is disputed, the recording is your evidence.
-
Review for accuracy -- AI transcription is typically 95-98% accurate. That 2-5% error rate can include names, technical terms, and numbers -- exactly the parts that matter most. Always review.
Extracting Value from Transcripts
A 45-minute interview produces thousands of words. AI helps you find the gems.
Finding the Best Quotes
Here is a transcript of my interview with [subject] about [topic].
Extract:
1. The 5 most quotable moments (strong language, clear opinions, emotional)
2. Any surprising or unexpected statements
3. Key data points or specific claims they made
4. Moments where they contradicted themselves or common narratives
5. The single most compelling exchange in the interview
For each, include the timestamp and exact quote.
Generating an Interview Summary
Summarize this interview transcript as a journalist's working document:
1. Main themes (3-5 key themes with supporting quotes)
2. Key revelations (anything new or newsworthy)
3. Unanswered questions (what they avoided or didn't fully address)
4. Potential follow-up needed (additional reporting to verify claims)
5. Suggested lede approaches (based on the strongest material)
Transcript:
[paste transcript]
Comparing Multiple Interviews
When you've interviewed several people for the same story:
I have transcripts from 3 interviews for a story about [topic]:
Interview 1: [person, role]
Interview 2: [person, role]
Interview 3: [person, role]
Compare their perspectives:
1. Where do they agree?
2. Where do they disagree?
3. What unique information does each provide?
4. What contradictions exist between their accounts?
Maintaining Authenticity
AI assists the process but should never replace the human connection. Keep these principles in mind:
- Be present during the interview -- Don't rely on AI transcription so much that you stop actively listening. The best follow-up questions come from genuine curiosity, not a script.
- Verify AI-extracted quotes -- Always check extracted quotes against the original recording. AI can occasionally merge nearby sentences or miss important context.
- Use quotes in context -- When AI pulls a "quotable moment," make sure the surrounding context supports how you plan to use it.
- Respect off-the-record requests -- AI transcription captures everything. Be rigorous about honoring any off-the-record agreements.
Key Takeaways
- Use AI to research interview subjects and generate specific, probing questions that go beyond generic talking points
- AI transcription tools (Otter.ai, Descript, Whisper) save hours but require human review for accuracy
- Prompt AI to extract the best quotes, key themes, and contradictions from long transcripts
- Always verify AI-extracted quotes against the original recording
- Stay present during interviews -- AI handles the mechanics, you handle the human connection

