Verbatim Transcription: Verbatim vs. Clean Verbatim (With Examples)

A verbatim transcript records speech exactly as spoken, and the transcript format you choose — full verbatim, clean verbatim or edited — decides how much of that you keep. Below: what a verbatim transcript includes, how clean verbatim differs, one passage written three ways, when you need word-for-word text, what AI transcription keeps, and formatting rules to copy.

What is a verbatim transcript?

A verbatim transcript is a word-for-word written record of a recording that keeps every word and sound as spoken, including fillers, repetitions, false starts and pauses. False starts are sentences the speaker abandons and restarts, as in "We should— let's wait." The style also goes by full verbatim, true verbatim or strict verbatim transcription.

It is typed from audio recordings of interviews, hearings or calls to capture the speech as heard: who hesitated, who laughed, where a speaker trailed off. A grammar slip stays on the page exactly as it was said.

What is a verbatim statement?

A verbatim statement is a person's own words, recorded or written down exactly as said, with no rewording. The term comes up with witness testimony and other legal documents: under Federal Rule of Civil Procedure 26(b)(3)(C), a recording or its transcription counts as a person's statement when it "recites substantially verbatim the person's oral statement."

What does a verbatim transcript include?

A verbatim transcript includes everything audible: words, filler words (um, uh, like), repetitions, stutters, false starts, interruptions, and non-verbal cues such as laughter or long pauses. Each element has a standard mark:

Fillers, stutters, repetitions and false starts are forms of speech disfluency; with non-verbal cues, they show the hesitations that words alone miss. The timestamp in an [inaudible] tag lets a reviewer jump to that spot in the audio file.

What is clean verbatim (non-verbatim) transcription?

Clean verbatim transcription removes fillers, stutters and false starts but keeps the speaker's own words, order and meaning — nothing is paraphrased. It is also called intelligent verbatim or non-verbatim transcription, and it trades detail for readability.

Clean verbatim keeps:

• the speaker's vocabulary, slang and word order;
• grammar mistakes, as spoken;
• sentences cut off by another speaker.

Repetitions and most non-verbal tags, such as [laughs], go along with the fillers.

Edited transcription goes further: it fixes grammar, rewrites clumsy sentences and finishes unfinished ones, so its quotes are no longer word for word. The labels blur between providers; TAUS even says edited transcription is "also frequently called 'clean verbatim.'" Before you order, ask for a sample paragraph and check whether the grammar changed.

Phonetic transcription, which records how words sound, is a different discipline.

Verbatim transcription example: one passage, three styles

Here is the same 28-second exchange from a scripted project meeting, transcribed in full verbatim, clean verbatim and edited form. The passage is a scripted test recording: a manager and an engineer, voiced with synthetic voices, discuss a delayed launch.

Full verbatim

Style: full verbatim
Speakers: Manager, Engineer

Manager:  What happened with the launch date?
Engineer: Um, so the— we, uh, we had the build ready on, like,
          Tuesday, but testing found a— QA found a bug in the, um,
          login flow. And I-I didn't want to ship that, because,
          you know, the login is the thing clients actually use.
          So we moved it a week. I mean, it's not great,
          but—
Manager:  Is the new date firm?
Engineer: Yeah. Yeah, the twelfth is firm. Unless, uh— [pause]
          no. It's firm.

Clean verbatim

Style: clean verbatim
Speakers: Manager, Engineer

Manager:  What happened with the launch date?
Engineer: We had the build ready on Tuesday, but QA found a bug in
          the login flow. And I didn't want to ship that, because
          the login is the thing clients actually use. So we moved
          it a week. It's not great, but—
Manager:  Is the new date firm?
Engineer: Yeah, the twelfth is firm. No. It's firm.

Edited transcript

Style: edited
Speakers: Manager, Engineer

Manager:  What happened with the launch date?
Engineer: The build was ready on Tuesday, but QA found a bug in the
          login flow. I decided not to ship it, because clients
          rely on the login. We moved the launch by one week. It
          isn't ideal.
Manager:  Is the new date firm?
Engineer: Yes, the 12th is firm.

What each step does to the engineer's lines:

The first step only deletes; the second rewrites. "I decided not to ship it" is a paraphrase that shifts the meaning: the engineer only said "I didn't want to ship that," a preference rather than a decision. Clean verbatim loses something too: without "Unless" and the pause, "No. It's firm." hides the doubt behind it, so keeping that hesitation is the transcriber's call.

When do you need a verbatim transcript?

Choose a verbatim transcript when how something was said matters as much as what was said — legal testimony, qualitative research, journalism quotes and compliance reviews; choose clean verbatim for meetings, notes and content.

In a legal context, the record comes from a court reporter or another person the court designates to produce it. Federal law requires each session of a US district court to be "recorded verbatim by shorthand, mechanical means, electronic sound recording, or any other method" (28 U.S.C. § 753(b)), and the reporter certifies the transcript. For depositions, Rule 30 of the Federal Rules of Civil Procedure allows "audio, audiovisual, or stenographic means," and the officer certifies that the deposition "accurately records the witness's testimony."

Qualitative research names the same choice: naturalism, where "every utterance is transcribed in as much detail as possible," versus denaturalism, which removes stutters, pauses and nonverbals (Oliver, Serovich and Mason, 2005). Conversation analysis goes further than full verbatim, with notation based on Gail Jefferson's system: timed pauses such as (2.6), (.) for a just noticeable pause and square brackets for overlap (Loughborough University tutorial). In an interview, a laugh or a long pause before an answer carries emotion the words alone lose.

Verbatim notes vs. meeting minutes

Verbatim notes record what each speaker said word for word; meeting minutes record what the group did. As a note-taking method, verbatim notes mean writing down everything a speaker says as you hear it. The Austin Community College study-skills guide warns that such notes are "usually incomplete," because a speaker can talk faster than you can write, and that they "require attention and reorganization to be useful."

Minutes skip the wording on purpose. In the words of the FAQ on the official Robert's Rules of Order site, "Minutes are a record of what was done at a meeting, not a record of what was said," so minutes are not verbatim by definition. At a business meeting, write down decisions, owners and dates. If someone needs the exact words, record the call with everyone's consent and keep the transcript next to the minutes.

Can AI produce a verbatim transcript?

AI transcription tends toward clean verbatim: in our test on a scripted clip voiced with synthetic voices, Meetlens produced a clean-verbatim style transcript that dropped most fillers, keeping 4 of 10 filler words and 2 of 3 false starts, with no pause marks or dashes. Deepgram's filler words setting is off by default, and without it "uh" and "um" are "stripped out of the transcript to improve readability." AssemblyAI's documentation says the same: "By default, filler words are removed from the transcript before it's returned."

Accuracy scores skip the same material: before counting errors, the Whisper study from OpenAI researchers strips bracketed text and filler words such as "uh" and "um" from both the model's output and the reference. Transcript accuracy measured that way says nothing about whether fillers or [laughs] survive. This AI smoothing helps readability, and it's why an AI draft needs a human pass before it can stand as a verbatim transcript.

The draft still saves the typing. Meetlens, an AI transcriber that runs as a Chrome extension, records browser meetings in Google Meet, Zoom Web, Microsoft Teams Web or Webex without a bot and transcribes them automatically, with speaker labels and MM:SS timestamps. Its transcription page calls the result "an AI-generated working transcript, not a certified or professional record."

To make it verbatim, replay the recording, correct the text and rename speakers in the meeting workspace, and type in the tags from the style guide below. Check short replies: a reply of three words or fewer between two turns of the same speaker can merge into that speaker's turn. Pro and Business export the result as Markdown, TXT or PDF.

Verbatim transcription rules (style guide)

A verbatim transcription style guide keeps the spelling of fillers consistent, marks unclear audio with a timestamp, puts anything that is not speech in brackets, and never corrects the speaker's grammar. Copy these rules into your project notes:

Verbatim transcription style guide

1. Header: name the style ("Style: full verbatim") and list every
   speaker label.
2. Speakers: use the same label at every turn, e.g. "Manager:".
3. Fillers: one spelling each (um, uh, mm-hm), never "umm" in one
   place and "um" in another.
4. Unclear audio: [inaudible 00:03:12], only after replaying the spot.
5. Non-speech: square brackets, lowercase: [laughs], [coughs],
   [pause], [crosstalk].
6. Breaks: a dash where a speaker breaks off or is cut off ("but—")
   and again where they resume; hyphens for a stutter ("I-I-I").
7. Grammar: transcribe as spoken; never fix a word, tense or sentence.
8. Guesses: put a doubtful word in brackets with a question mark,
   e.g. [login?].

Rules 4, 6 and 7 follow the federal format rules cited above: "inaudible" only "when it is impossible to transcribe the record," a dash "at the point of interruption, and again at the point the speaker resumes speaking," and grammar as spoken. Those rules put pauses and gestures in parentheses but allow brackets; pick one and keep it. Consistency matters most when multiple people transcribe one project: a search for "um" misses every "umm."

Pick the style first, then the transcriber.