The problem with direct quotes in modern journalism
A direct quote is the highest-stakes element in a journalistic piece: words inside quotation marks imply that those are exactly what the source said, without interpretation or paraphrase. Yet when journalists work under time pressure, manual transcription introduces subtle errors that can alter meaning — a changed word, an omitted negative, a lost temporal nuance.
AI transcription tools don't eliminate the journalist's editorial responsibility, but they drastically reduce the error rate in initial text capture. The result is a more reliable working foundation from the very start.
The five-step extraction process
Step 1: Transcription with diarization from the start
The first step is getting a transcription with speaker diarization right from the beginning. Uploading audio to a platform like CallsIQ automatically generates text where each speaking turn is labeled with the speaker's name or number. This eliminates the most common source of confusion when reviewing long interviews: not knowing who a specific statement belongs to.
Step 2: Locate candidate quotes with semantic search
Not every statement in an interview deserves a direct quote. Semantic search lets you quickly locate moments when the source spoke about a specific topic: search for "municipal budget" and the tool takes you directly to the relevant fragments, complete with timestamps. This saves listening through the whole recording when you're looking for a particular statement.
Step 3: Verification against the original audio
Before including any quote in an article, click the timestamp and listen to those 15–20 seconds of audio. Confirm the transcription matches word for word, especially for proper names, figures, and technical terms where AI may make errors. This verification takes less than a minute per quote and is the guarantee of fidelity.
Step 4: Contextualize the quote before extracting it
A quote without context can be misleading. Before copying it to your article, re-read the two or three exchanges preceding it in the transcript to make sure you understand exactly the context in which it was said. The audio never lies, but decontextualized quotes can distort the meaning of a statement.
Step 5: Document the source for the archive
Save the complete transcript linked to the original recording. In case of a dispute about what was said in an interview, the timestamped transcript plus the audio constitute robust documentation that protects both the outlet and the journalist.
Golden rule: if you have any doubt about a word in the automatic transcription, listen to the audio. Never publish a direct quote without verifying it against the original sound source.
Common mistakes when working with direct quotes
- Cutting too much: a shortened quote can change the meaning. If you need to trim, use ellipses in brackets and ensure it doesn't alter the core message.
- Correcting spontaneous speech: filler words, self-corrections, and colloquial speech are part of the quote. They're only removed if the outlet has a clear editorial policy on this.
- Trusting the transcript without listening: even the best AI models make errors with proper nouns, acronyms, and technical terms. Auditory verification is always necessary.
- Mixing paraphrase with direct quote: inside quotation marks, there can be no word the source didn't literally say.
How AI speeds up workflow without compromising accuracy
The key is understanding what AI does and what remains the journalist's responsibility. AI captures text with high fidelity, identifies speakers, generates timestamps, and enables instant searches. The journalist verifies, contextualizes, and decides what deserves to be quoted and how. This division of labor is what makes automatic transcription a tool for journalistic quality — not a shortcut that sacrifices rigor.
With CallsIQ, each extracted quote is linked to the exact second of audio, creating a verification trail that speeds both the internal editing process and any subsequent checks by editors or fact-checkers.