Full Verbatim vs Intelligent Verbatim vs Edited: Choosing the Right Level (With Examples)

Most people ordering transcription for the first time do not specify a style. They send the audio, the transcript comes back, and only then do they discover that the thing they needed the hesitation before an answer, the exact wording of a denial, the fact that two people were talking over each other has been quietly tidied away.

It cannot be put back. The transcript is a new document, not a lossy copy you can re-expand. Whatever was removed is gone unless someone re-listens to the original.

So the choice of style is worth two minutes of thought before you upload, not two weeks of regret afterwards. Here is what the three main levels actually mean, shown on the same piece of audio.

The three levels at a glance

The line people find hardest is the one between full and intelligent verbatim. The line between intelligent verbatim and edited is easier: edited transcripts change what the sentence is, not just what gets left out.

One recording, three transcripts

Here is a fragment of a business interview. First, everything that was actually on the tape.

What the audio contains

So I mean, the thing is, um, we’d been using them, what, three years? Three, maybe four. And it was fine, it was, you know, it did the job. But then sorry, can I just — [pause] — yeah, sorry. But then last, last March they changed the, the pricing structure and nobody, nobody told us. Nobody told us. And I found out because I, because Sarah in finance flagged it. [laughs] Typical, right?

Full verbatim

R2: So — I mean, the thing is, um, we’d been using them, what, three years? Three, maybe four. And it was fine, it was, you know, it did the job. But then — sorry, can I just — [pause 4 sec] — yeah, sorry. But then last, last March they changed the, the pricing structure and nobody, nobody told us. Nobody told us. And I found out because I, because Sarah in finance flagged it. [laughs] Typical, right?

Nothing is removed. The pause is measured. The interruption where the speaker breaks off to take a drink is preserved because, in some contexts, the fact that someone paused for four seconds at that exact point in the sentence is the most interesting thing on the page.

Intelligent verbatim

R2: So the thing is, we’d been using them three years — three, maybe four. And it was fine, it did the job. But then last March they changed the pricing structure and nobody told us. Nobody told us. And I found out because Sarah in finance flagged it. [laughs] Typical, right?

Note what survived. The repeated “Nobody told us” is still there, because that repetition is emphasis — the speaker is making a point, not tripping over their words. The [laughs] is still there, because it changes how “Typical, right?” should be read.

This is the part that separates a skilled transcriptionist from a find-and-replace operation. Intelligent verbatim is a series of judgement calls about which disfluencies carry meaning. Strip them mechanically and you flatten the speaker.

Edited

R2: We had used the supplier for three or four years without issue. In March, they changed their pricing structure without notifying us, and our finance team identified the change.

Clean, professional, quotable in a report. Also: the exasperation is gone, Sarah has become “our finance team”, and you can no longer tell that the speaker was unsure whether it was three years or four. For a board paper, that is a fair trade. For a research transcript, it is the destruction of your data.

When the style choice changes the meaning

The example above costs you tone. In some settings, the wrong style costs you the substance.

Consider a recorded interview during an investigation:

What was said:

I didn’t — I never touched it. I mean, I might have moved it, but I never — no. No, I didn’t touch it.

In intelligent verbatim:

I never touched it. I might have moved it, but no, I didn’t touch it.

Both are accurate in the narrow sense. Every word in the second version was genuinely spoken. But the first shows a person changing their account twice mid-sentence and then reverting, and the second shows a person giving a slightly qualified denial. A solicitor reading the second transcript would not know to go back to the audio.

This is why police, court and deposition work is full verbatim as a matter of course, and why the same applies to recorded statements used in insurance claims. The transcript may end up being read by someone who will never hear the recording. It has to carry everything.

Which level do you need?

Choose full verbatim when

The transcript may be evidence. Legal proceedings, arbitration, tribunal hearings, police interviews, disciplinary investigations, insurance statements. Anywhere the precise form of words could be argued over.

You are analysing how something was said. Conversation analysis, discourse analysis, clinical linguistics, speech and language therapy, any methodology where pauses and repairs are the data rather than noise around it.

The speaker’s state matters. Psychological assessment, safeguarding interviews, some medical contexts.

Be aware of the trade: full verbatim takes longer to produce, is harder to read, and costs more per audio minute. That is not a supplier upcharge, it is genuinely more work — a minute of overlapping speech with four participants can take fifteen minutes to render properly.

Choose intelligent verbatim when

You need the speaker’s voice but not their every stumble. This is the default for most interview transcription, and it is almost always right for focus groups and market research, where you want quotable material that still sounds like a real person.

You are coding qualitative data. Most academic and research transcription sits here. Thematic analysis rarely needs every “um”, but it does need the participant’s own phrasing intact — the moment you paraphrase, you are coding your words rather than theirs.

One caution for researchers: check your methodology and your ethics application before defaulting to intelligent verbatim. If you have told a committee you are using a particular analytic approach, the transcription convention is part of the method, and some approaches require full verbatim.

You are publishing the words. Podcasts, documentary work, media and anything destined for captions generally reads better in intelligent verbatim, with a further pass for caption timing and line length.

Choose edited when

The output is a document, not a record. Board minutes, meeting notes, corporate reports, conference write-ups, internal summaries.

A clinician dictated it for the record. Medical dictation is almost always edited and formatted to a house template — nobody wants “um” in a discharge letter. The skill here is different: correct terminology, correct drug names, correct structure.

The speaker will be quoted in print. Speeches, press material, ghostwritten content.

What to specify when you order

Style is one of five things worth stating up front. Sending these with your first file saves a revision cycle:

1. Style — full verbatim, intelligent verbatim, or edited
2. Speaker labels — names if you can supply them, or Interviewer/Respondent, or Speaker 1/2/3
3. Timestamps — none, every change of speaker, or fixed intervals (every 2 or 5 minutes is typical)
4. Inaudible convention — how you want unclear audio marked, and whether you want a timecode against it so you can check the original
5. Terminology — a short list of names, places, products, drugs or case references that appear in the audio

That fifth one has more effect on accuracy than anything else you can do. A transcriptionist who knows the surname is “Featherstonehaugh” will get it right; one who does not will spell it as heard.

Common mistakes

Ordering full verbatim “to be safe”. If nobody is going to read the disfluencies, you have paid more for a harder-to-read document. Safe is matching the style to the use.

Mixing styles across one project. If half your interviews are full verbatim and half are intelligent, your dataset is not internally comparable. Decide once, at the start, and put it in writing.

Assuming AI tools produce intelligent verbatim. Automated output is neither full verbatim nor intelligent verbatim — it is unreliable at both. It drops fillers inconsistently, invents plausible words where the audio is unclear, and has no concept of which repetition was meaningful. That inconsistency is the real problem: you cannot tell from the transcript which parts you can trust.

Leaving it to the supplier to guess. A good supplier will ask. If they do not ask, that itself tells you something.

Frequently asked questions

What is the difference between full verbatim and intelligent verbatim? Full verbatim captures every sound a speaker makes, including fillers, false starts, stammers, repetitions and non-verbal sounds. Intelligent verbatim removes the ones that carry no meaning while leaving the speaker’s own words and phrasing intact. Full verbatim is a record of the audio; intelligent verbatim is a readable version of what was said.

Is intelligent verbatim the same as a summary? No. Intelligent verbatim keeps every substantive word the speaker used. A summary or edited transcript restructures sentences and may condense. If someone offers you a “summary transcript”, establish which they mean.

Which style should I use for qualitative research interviews? Intelligent verbatim for most thematic and content analysis; full verbatim if your method examines interaction, pausing or repair, or if your ethics submission specified it. When in doubt, full verbatim is the recoverable choice — you can always simplify later, but you cannot restore what was removed.

Does full verbatim cost more? Usually, because it takes considerably longer to produce, particularly with multiple speakers or poor audio. Ask your supplier how they price it rather than assuming a flat uplift.

Can I change the style after delivery? Going from full verbatim to intelligent verbatim is straightforward editing. Going the other way means transcribing again from the audio, and you will pay for it twice. Choose carefully at the start.

Not sure which you need?

Send us five minutes of your audio and tell us what the transcript is for. We will return the same passage in two styles so you can see the difference on your own material before committing to a full project.

Our verbatim transcription service is entirely human — which is the only way these judgement calls get made consistently. Pricing starts at £0.95 per audio minute, and you can send the file without a sales call.