---
title: "The Art of the Transcript: Scoping and Proofreading Best Practices"
description: "How scopists, proofreaders, and reporters work a US deposition or court transcript: the passes, the research, the hard cases, and who certifies the record."
url: "https://depostack.com/blog/art-of-the-transcript/"
published: "2026-10-13"
updated: "2026-10-15"
author: "Andrew Mayes"
topic: "Transcripts"
---

# The Art of the Transcript: Scoping and Proofreading Best Practices

How scopists, proofreaders, and reporters work a US deposition or court transcript: the passes, the research, the hard cases, and who certifies the record.

Published October 13, 2026 · Updated October 15, 2026 · By Andrew Mayes · Topic: Transcripts

Canonical URL: https://depostack.com/blog/art-of-the-transcript/

## Key takeaways

- Fix the machine, never the witness: correct what the writer or the software got wrong, keep what was said, and flag what you cannot confirm.
- Scopists, proofreaders, and reporters do different work, and only the reporter certifies the transcript.
- Scope against the notes and the audio, finish research before the read, then proofread cold for sense and continuity.
- Format is local: know which statute, court policy, or manual governs the job before you edit.
- NCRA's 2026 guidance lets AI support the reporter's work but never generate the record.

Every certified transcript ends with a signature. Before it, the words pass through several hands: the reporter who wrote them, often a scopist who edits them against the steno notes and the audio, sometimes a proofreader who reads the pages cold, and the reporter again, who reads, decides each open question, and signs. The tools in that chain have changed many times. The signature has not.

This guide is for the people in that chain: court reporters, scopists, and proofreaders working on US deposition and court transcripts. It gathers what the rules, the profession's guidance, practitioners, and the research say about doing the job well, and where the evidence stops. Where a checklist or an order of work is our own synthesis rather than a published standard, we label it that way.

To practise the method on invented pages and synthetic audio, with the rule and its source behind every answer, try the free companion: [The Art of the Transcript training](https://depostack.com/training/art-of-the-transcript/). It is practice, not a credential.

## The one rule: fix the machine, never the witness

Every editing decision in a transcript comes back to one distinction. Some errors belong to the machine or the writer: a misstroke, an untranslate, a wrong homophone the dictionary produced, a word the audio shows was dropped. Those are yours to fix. Other "errors" belong to the witness: the grammar, the false start, the wrong date said with confidence. Those are the testimony, and they stay.

Federal court policy says it in two sentences: "The transcript must provide an accurate record of words spoken in the course of proceedings. All grammatical errors, changes of thought, contractions, misstatements, and poorly constructed sentences must be transcribed as spoken."[1] That is § 520.40.10(b)(1) of the Guide to Judiciary Policy, Volume 6. Texas's Uniform Format Manual uses nearly the same words, and it too says speech "must be transcribed as spoken."[2] Michigan's manual for reporters and recorders adds that "a court reporter or recorder should not, even on the order of a judge, strike from the record that which actually took place," because "Any deviation in a transcript or certified record stultifies appellate review."[3] The federal Guide keeps even what a court strikes: "the material ordered stricken, as well as the order to strike, must all appear in the transcript."[1]

The rule has a reason. Under Federal Rule of Civil Procedure 30(f)(1), "The officer must certify in writing that the witness was duly sworn and that the deposition accurately records the witness's testimony."[4] The certificate vouches for what the witness said, not for what the witness meant. An editor who improves the testimony makes the certificate less true.

> **Figure.** The one rule on a page: correct the machine's error, keep what was said, flag what you cannot confirm. The reporter decides. Drawing: A transcript page with three marks: a correction inserted above one line, two lines bracketed and kept as spoken, a flag beside a third line, and the reporter's signature and seal at the foot.

Three verbs carry the whole method:

- **Correct** what the machine or the writer got wrong, when the notes or the audio confirm it.
- **Never correct** what the witness, the attorneys, or the court actually said.
- **Flag** anything you cannot confirm, and leave the decision with the reporter.

Our table below sorts the everyday cases. It is DepoStack's synthesis, not an industry standard, drawn from the federal and Texas editing-of-speech rules, from Rule 30(e), which gives the witness, not the editor, the way to change an answer,[4] and from two NCRA advisory opinions. One says the reporter should not make a change to accurately reported remarks "even if requested by the court," and that parties may change a transcript "by agreement of all parties in writing or by ruling of the court."[5] The other says "A reporter should not include conversations unless they were actually heard and reported by the reporter."[6]

| Correct it, when the notes or audio confirm | Never correct it | Flag it for the reporter |
| --- | --- | --- |
| A mistranslation, untranslate, or misstroke | The witness's grammar, slang, or dialect | A name or term you cannot verify |
| A wrong homophone the writer produced | A misstatement, or a wrong date or name, as spoken | A number that looks wrong but may be what was said |
| A word-boundary or affix error | False starts and changes of thought, subject to the local filler rule | A passage you cannot make out |
| A dropped word the audio confirms | An attorney's own words, without written agreement or a court ruling | Speech on the audio or video that is not in the notes |
| A speaker ID or byline error the notes or audio confirm | A document misread aloud: the spoken version stays | An unmarked switch into English in interpreted testimony |
| Format, pagination, and parenthetical form | Anything the court ordered stricken: it stays, with the order | Anything that would change meaning if you guessed |

## Three roles, one signature

NCRA defines the scopist by one skill: "A scopist is a professional transcript editor for court reporters. However, unlike an editor or a proofreader, a scopist has the ability to compare a court reporter's shorthand to the finished transcript."[7] The federal Guide's compound, "scopist/notereader," ties the role to the older skill of reading another writer's shorthand.[1]

| Role | Works from | The job |
| --- | --- | --- |
| Court reporter | The proceeding, the notes, and any audio | Writes, edits, decides every open question, and certifies |
| Scopist | The CAT file: steno, translation, and synced audio where provided | Fixes the machine's errors, researches, and flags what cannot be confirmed |
| Proofreader | The transcript text, with audio spot-checks where provided | Reads for sense, continuity, and consistency |

That summary is ours. What the rules say is narrower and firmer: the certificate belongs to one person. California's Code of Civil Procedure requires that "The deposition officer shall certify on the transcript of the deposition, or in a writing accompanying an audio or video record of deposition testimony, as described in Section 2025.530, that the deponent was duly sworn and that the transcript or recording is a true record of the testimony given."[8] NCRA Advisory Opinion 34 says "A reporter should not certify that a transcript is a true and accurate record of the proceeding if the transcript has not been personally produced by that reporter (or under that reporter's direction and control) or verified by that reporter."[9] NCRA's 2026 guidance on AI opens from the same point: "The stenographic reporter or captioner is the sole creator, guardian, and certifier of the official record."[10]

No rule we found gives a scopist or a proofreader a certifying role. Texas states the other half of that most clearly. Its Code of Ethics provides that "A certified shorthand reporter or court reporting firm who employs or supervises an unlicensed person to assist in the preparation of a transcript or reporter's record is responsible for any acts of the unlicensed person relating to court reporting," and that the reporter "must take reasonable remedial action" when someone under their direct supervision breaks the Code.[11] A firm "must not change the record of a certified shorthand reporter," though the reporter may let the firm correct identified errors on the title, appearance, or index page.[11] Our reading: the work can be delegated, the responsibility cannot. Once the reporter signs, the scopist's edits are the reporter's edits.

Practitioners add that the roles work best apart. The authors of the Journal of Court Reporting's baker's dozen of proofreading tips put it this way: "Scoping and proofreading are not the same function. For the best results, scoping should be done by someone else and at a different time than when you proofread."[12] Kathy McHugh, RPR, CRR, agrees: "I personally think you shouldn't scope and proof your work. It's best to have two eyes on the transcript."[13]

## The hand-off to the scopist

Scoping starts before the job reaches the scopist, with what the reporter captured. "You have to speak up because if you didn't hear something, your scopist is not going to hear it," says Margary Rogers, RPR, CRI, an official reporter in Washington, D.C.[14] Linda Evenson, a retired scopist and founder of Internet Scoping School, is blunter: "A scopist's job isn't to clean up a mess." She also warns reporters not to "set the perfectionism bar so high that the scopist can't reach it."[14]

New reporters may do better to wait. Ninette Butler, RMR, CRR, CRC, advises new professionals to start by doing their own scoping, "at least for the first year or so," "to develop a dictionary and clean up your writing."[14] The warning is old. In 1981 the National Center for State Courts found that "Scopers used too early in reporter training may lead to continual delay because the reporters may not be forced to clean up their shorthand."[15] Debi Smolinske adds a point about the relationship itself: working with a scopist is "not an inalienable right," and scopists "are businesspeople."[14]

What should travel with the job? No published standard lists it. Our checklist, drawn from the scoping practice Linda M. Lee describes and from Michigan's manual:

- The CAT file and the notes, and synced audio where the reporter provides it.
- The case's names and terms, with their source. Michigan tells electronic-recording operators that "When possible, the operator should ask the attorneys to supply a list of names and terms prior to the court session," and, for a person who is not available, to "check with the keeper of the exhibits."[3] The passage is written for recording operators, but the habit carries over to steno.
- The exhibits, or at least the exhibit list. Where the case's numbering is still being settled, our free [exhibit number planner](https://depostack.com/tools/exhibit-numbers/) lays the numbers out by deponent and builds a log to print.
- The reporter's preference sheet, including the verbatim choice. Lee calls strict versus intelligent verbatim "a hot, but real, topic that affects how the quality of a scopist's work is perceived."[16]
- The deadline, the delivery format, and any confidentiality terms (see the section on responsibility below).

Questions come back the same way. Lee's practice is to "Leave hidden notes for the reporter to check (which conscientious ones who care about their work product are happy to receive)."[16] A hidden note is the flag in its working form: visible to the reporter, never printed in the transcript, and never a guess presented as fact.

## Five passes, in order

The order of work below is DepoStack's synthesis, not a published standard. No source lists these five passes together, but each one rests on cited practice: scoping against audio, finishing research first, reading cold, sweeping afterward, and the reporter's certificate.

1. **Scope against the notes and the audio.** Work line by line in the CAT file; this pass catches the machine's errors.
2. **Finish the research before the read.** Names, terms, numbers, and exhibits, settled or flagged.
3. **Proofread cold, from the text.** Read for sense and continuity, with audio spot-checks only.
4. **Sweep mechanically.** Speaker IDs and bylines, special terms, doubled words, and consistency.
5. **The reporter's read.** The reporter decides each flag, reads the final, and signs.

Each pass has its own job. A later pass is not a second try at an earlier one.

> **Figure.** Our five passes on one transcript: scope with the audio, research first, proofread cold, sweep, then the reporter reads, decides, and signs. Drawing: Five transcript pages on one centre line: the first beside an audio waveform, then a word list, a page with a pencil, a page with its speaker lines ticked, and a page signed and sealed.

### Pass 1: scope against the notes and the audio

Lee, a full-audio scopist, describes the pass as line-by-line work in the software, and her instruction on audio is plain: "Replay audio passages however many times is needed for accuracy's sake (even for immediate copy, the tightest of time frames)."[16] If you are new to the software side, our note on [what CAT software does](https://depostack.com/blog/what-is-cat-software/) covers how steno becomes English and where untranslates come from.

Numbers belong to this pass. Lee's rule: "Pay close attention to all dates, numbers, Bates stamps, and monetary values, making corrections as needed (it is not a proofreader's responsibility to double-check these for accuracy)."[16] The Eastern District of Washington's August 2026 newsletter to counsel explains why: "Numbers run together, digits get dropped, and a single transcription error in a financial figure can create a significant record problem." Its example pairs "Exhibits 110, 11, 12, 13, 14" with "Exhibits 110, 111, 112, 113, 114."[17] Automated checks do not cover this. Stenograph's help page for its Check It feature says "Numbers are ignored by Check It and there will be no suggestions generated for numbers."[18] Lee runs spellchecks "inside of the software (which is usually subpar) and with other applications."[16]

One jurisdiction draws a hard line around audio. California's Court Reporters Board writes that "The stenographic/voice notes are the official record," that in a complaint "the Board looks to the transcript and the stenographic/voice notes, not an audio file," and "In other words, do NOT rely upon backup audio media for transcript production."[19] For depositions recorded both ways, California makes the stenographic transcript "the official record of that testimony."[20] Full-audio scoping elsewhere uses audio as a checking aid, not as the record. Know which side of that line your job is on.

### Pass 2: finish the research before the read

Lee's standard for research is to "Perform online research of even the most mundane words since we can't assume anything," and she adds that "even spellings given to the scopist by the reporter sometimes contain errors."[16] The baker's dozen tips explain why research comes first: "All research should be completed prior to commencing proofreading. You will lose the flow necessary for contextual reading if you're stopping every half page to double-check a spelling or perform an online search for a term." The same tips advise choosing "a reputable primary dictionary to follow (Merriam-Webster, American Heritage, Oxford)" and to "be wary of accepting the first word that pops up in a Google search that seems to fit your phonetic."[12]

Our recommended order for checking a word, which no source states in this form but which follows from Lee's caution and Michigan's advice:

1. **The record itself.** A spelling given on the record, or the same name later in the job.
2. **The exhibits and case documents.**
3. **The case word list**, from the reporter or counsel.
4. **Specialist references** for the field: medical, technical, or legal.
5. **The open web, last**, and only when the result ties back to the record.

When the trail runs out, the word gets a hidden note, not a guess.

### Pass 3: proofread cold, from the text

The proofread is a reading pass. Lee describes her own proofreading hat this way: "I am strictly in reading mode, editing anything that was missed by the scopist as s/he performed all of the above, which ideally should be minimal or else I'm scoping behind the scopist."[16] If the proofreader is fixing translation errors, the scoping was not done.

Audio has a smaller role here. The baker's dozen tips: "While spot-checking the audio can be helpful, listening to continuous audio is not recommended. It is difficult to read for context, pay attention to punctuation, and listen to audio at the same time."[12] Tara Hudson, RPR, CRR, uses audio to settle a puzzle: "If a question or answer is a puzzle, I will go back and listen to the audio. It helps me understand not just what was said but also how it was said, and then I can punctuate accordingly."[13]

What to read for, in the Council members' own words:

- **Continuity.** "Proofreading is not just about correct punctuation and spelling, it's about continuity." (Hudson)[13]
- **Clarity over polish.** "To me it is more important that the attorney reading it be able to understand the answer rather than have it be grammatically perfect." (Aimee Suhie, RPR)[13]
- **The text made strange.** Hudson will "Always proofread in a different font than I see when I'm doing realtime or scoping."[13]
- **Stumbles.** Shellene L. Iverson, RPR, reads aloud: "I find that if I stumble while reading, which is less likely when I read to myself, chances are it's an area that needs another look."[21]

Some habits fade. One Council member said: "I was taught in school to read backwards to check for misspellings but don't do it anymore since most are caught in spellcheck."[21]

The slips practitioners name are mostly real words: "they're/there/their," "it's/its," "as/at," "it/is," missing "a" and "the," "formal/former, contact/contract," and "I did/did I,"[12] with "elicit versus illicit and affect versus effect" from Lee.[16] Spellcheck passes every one of them.

### Pass 4: sweep mechanically

After the read, a different kind of attention. "After you've finished your initial proofreading, go back and double-check bylines and speaker identifications as well as consistency with any special terms you've become aware of during the job. It's easy to read right past such errors when you're focused on reading for context."[12] The same tips keep a final spellcheck: "Spell-check is very good at catching double words such as 'the the' and 'that that,' which are easily missed while reading." Word's grammar check helps with missing prepositions and "form" for "from," though it "should never be taken as gospel." And "If you've used more than one scopist to get the job done, pay special attention to consistency in capitalization, hyphenation, and other potential differing styles among scopists."[12]

### Pass 5: the reporter's read

The last pass belongs to the person who signs. The reporter answers each hidden note, accepts or rejects each suggestion, and reads the final, because Opinion 34 asks that the reporter produced, directed, or verified what they certify.[9] The scopist and the proofreader make that read faster. They do not make it optional.

## Why separate passes work

The proofreading research studied students and general readers, not professional scopists, so read it as an explanation, not a measurement of our work. It still lines up with what practitioners do.

**Real-word errors are harder to catch.** In two eye-tracking experiments with 48 readers each, people proofreading for real but wrong words (trial for trail) caught fewer errors, 81 percent, than people proofreading for misspellings that made non-words, 89 percent. The authors concluded that "checking for real words that were inappropriate in the sentence context was more difficult than checking for spelling errors that produce nonwords."[22] The two rates come from different experiments with different participants. A 2022 study found the same pattern and one more: "Overall, word errors (i.e., wrong words) were detected less often than non-word errors (i.e., misspellings), and function word errors were detected less often than content word errors."[23] Function words are the "a," "the," "as," and "at" on the practitioners' list.

**Time on the text matters.** In the same study, "Students who made more text fixations and spent more time proofreading detected more errors."[23]

**Your own work may not be the problem you think.** The well-known finding that people catch fewer errors in their own writing comes from a 1993 study in which, as the 2022 authors report it, students proofreading their own essays "detected 20% fewer errors than those who proofread a familiar essay written by someone else," with "small samples (e.g., n = 10 per condition in Experiment 1)." The 2022 replication did not find a significant effect: "Experiment 1 failed to reveal a self-generation effect."[23] The case for a separate proofreader rests better on fresh attention and a different job than on the idea that writers cannot see their own errors.

**Reading aloud helps; odd fonts do not.** A 2022 study in the Journal of Applied Research in Memory and Cognition found that reading aloud improved proofreading, while a "disfluent" font, Sans Forgetica, did not.[24]

**Screen or paper matters less than you might think.** A 2022 Cambridge review of 26 experiments found "no conclusive evidence that proofreading on screen will, in general, lead to worse performance than proofreading on paper," though only seven of the experiments were published after 2000, and the review cautions that its findings may not generalize to speech transcripts.[25]

Our reading of all this: steno mistranslations and speech-recognition errors usually come out as real words, which puts them in the harder category. The practitioners' habits (a separate scoper and proofreader, a pause between passes, a new font, reading aloud, and sweeps after the sense read) each break the expectation-driven reading that lets those errors through.

## Know your jurisdiction

Format is local, and it is not optional. The federal uniform format dates to 1944: "The Judicial Conference first adopted the uniform transcript format in 1944 to assure that each party is treated equally throughout the country." The current Guide is firm about it: "No court, judge, supervisor, court reporter, or transcriber may authorize a deviation from the requirements provided by the Judicial Conference," and "The per-page transcript rates are based on strict adherence to the prescribed format."[1]

| Jurisdiction | Where the format lives | A few of its rules |
| --- | --- | --- |
| Federal courts | Guide to Judiciary Policy, Vol. 6, Ch. 5 | Speech as spoken; fillers normally left out, but kept "whenever their exclusion could change a statement's meaning"[1] |
| California | 16 CCR § 2473 for licensees | At least 25 lines a page; "Q" and "A" within the first 8 spaces; searchable electronic format "if requested"[26] |
| Texas | Uniform Format Manual | "Uh-huh" for yes and "Huh-uh" for no; dashes at an interruption; the Gregg Reference and Merriam-Webster as references[2] |
| Michigan | SCAO Manual for Court Reporters and Recorders | Semicolons in a complicated series; a comma or a period after yes or no[3] |

California's minimum standards apply to its licensees, and court rules govern official transcripts there. The regulation, as amended effective January 1, 2025, also says "Failure to comply with these minimum standards, as noted above, constitutes grounds for disciplinary action."[26] Texas warns that "failure to comply with mandatory provisions of this manual by either official or freelance court reporters may be grounds for discipline through the Court Reporters Certification Board."[27]

Punctuation is the transcript's only tone of voice, and it can change the testimony. Michigan's manual gives this example for a complicated series: "I know Sharon; her mother, Rose; and Beatrice."[3] Our reading, which the manual does not spell out: with the semicolons, the witness knows three people, and Rose is Sharon's mother. Swap them for commas and a reader could count four. The same manual sets a rule for yes and no: "Always use a comma after the response yes or no. If the rest of the response adds something new to the information, use a period after the yes or no." Its examples are "A No, I did not see him running from the scene." and "A Yes. We went back the next day, too, to see what happened."[3]

Even one manual is not always consistent. Michigan punctuates one tag question "You saw the man enter the door; isn't that right?" and another "That is right, isn't it?"[3] Where the binding rule is silent or split, follow the reporter's preference sheet and apply it the same way across the job. Our order of authority, for what it is worth: a statute, court rule, or court policy first; then the official manual for that court or state; then the reporter's preferences; then a general style guide.

## The hard cases at a glance

Difficult proceedings are where editors are most tempted to improve the record. The defaults below mostly say the opposite: mark the situation, keep what was said, and send uncertainty back to the reporter. NCRA's advisory opinions require members to follow them only to the extent they are consistent with their own state and local laws, rules, and regulations, so local law comes first.[5]

| Situation | The documented default | What the editor does |
| --- | --- | --- |
| Interpreted testimony | Federal: "it will be assumed that answers are made in a foreign language and interpreted unless a parenthetical '(in English)' is inserted."[1] Texas uses the same presumption.[2] | Flag an unmarked switch into English; an unmarked answer reads as interpreted |
| A gap in a recorded transcript | Federal policy for electronic sound recording calls 'indiscernible' or 'inaudible' labels "highly undesirable," to be used "only when it is impossible to transcribe the record."[1] | Return the gap to the reporter; never fill it with a likely word |
| Speech the reporter did not hear | Not included unless "actually heard and reported by the reporter"; with all parties' written agreement, it may go in marked as a video excerpt[6] | Query it; never insert it from the audio or video |
| A judge or attorney asks for a change | Even at a judge's request, accurate remarks are not changed alone; the court "should confer with all counsel"[5] | Route the request to the reporter |
| Overlapping speakers | "Overlapping voices are one of the greatest threats to an accurate record."[17] | Keep every captured word; mark interruptions the way the governing manual does |
| Remote proceedings | Say it was remote; "The term 'videoconference' or 'remote' is preferable to the term 'virtual,' which may imply simulated"; chat is not in the record "unless one of the parties reads it into the record"[28] | Keep the technical-difficulty parentheticals; do not smooth over dropouts |
| Exhibits | California's Board suggests an "Exhibit Index" and omitting a confidential exhibit's full description from the open transcript[29] | Reconcile every "marked" reference with the index |
| Confidential designations | In California, the reporter cannot mark a transcript confidential without a protective order or a stipulation[30] | Match page and line references across volumes |

Two of those deserve more words.

**"Inaudible" is not one rule.** California's statute requires the label for recordings: "Transcripts derived from electronic recordings shall include a designation of 'inaudible' or 'unintelligible' for those portions of the recording that contain no audible sound or are not discernible."[31] California's Court Reporters Board advises live reporters never to use "(inaudible)" and to interrupt instead. Its pointer reads: "Pursuant to Government Code 69957, never use the parenthetical 'inaudible' because that is only used when transcribing audio recordings. A live court reporter is required to interrupt to protect the record."[19] That is the Board's reading of the statute, and it is guidance: the pointer says the pointers are not regulations and are not a basis for discipline.[19] For a scopist, the practical point is the same either way. A gap goes back to the reporter, who heard the room.

**Remote jobs add to the record.** California's Board says the appearance page should show whether each "attendee was remote or in person," allows a "technical difficulty" parenthetical, and has the reporter ask, "Is there anyone else in the room with you?"[32]

## The finished transcript

A rough draft and a certified transcript are different products, and the rules keep them apart. NCRA's guidelines say "An uncertified transcript may not be filed with the court," "An uncertified transcript may only be distributed to ordering parties to the case," and each page should carry a header or footer stating "uncertified draft transcript only."[28] The same guidelines hedge on cleanup: "Where possible, all untranslated steno strokes and conflicts should be resolved before an uncertified draft transcript is provided to any party."[28] California's statute is firmer: a rough draft "may not be certified and may not be used, cited, or transcribed as the certified transcript of the deposition proceedings," and "may not be cited or used in any way or at any time to rebut or contradict the certified transcript of deposition proceedings as provided by the deposition officer."[8] Federal realtime must say "Realtime Unedited Transcript Only" on each page and "should not include an appearance page, an index, or a certification."[1] Our note on [the state of transcript review in 2026](https://depostack.com/blog/state-of-transcript-review-2026/) covers realtime, rough, and certified text side by side.

When the final goes out, the rough's marks must not. California's Board lists telltales to remove from a rough, among them a mention of swearing in the witness by name and a footer with the firm name, the reporter's name, or a license number.[33]

Editors never change the substance; the witness changes it through errata. (The other routes are the ones NCRA names: a written agreement of all parties, or a ruling of the court.[5]) Under Rule 30(e), on request before the deposition is completed, the deponent has 30 days after notice to review the transcript and, "if there are changes in form or substance, to sign a statement listing the changes and the reasons for making them." The officer "must note in the certificate prescribed by Rule 30(f)(1) whether a review was requested and, if so, must attach any changes."[4] California gives the deponent 30 days after notice to change "the form or the substance of the answer to a question," and an unapproved transcript takes effect "as though it had been approved, subject to any changes timely made by the deponent."[34] Texas wants changes "on a separate sheet of paper, together with a statement of the reasons for making the changes," with "No erasures or obliterations of any kind," and a witness who does not return the transcript within 20 days "may be deemed to have waived the right to make the changes."[35]

Our final production check is below. It is DepoStack's synthesis, not a published checklist: we found no NCRA or trainer checklist for the final review. Each line rests on the rules and practice cited in this guide.

1. **Title page** matches the case style, number, date, time, and place. Rule 30(b)(5)(A) requires the officer's opening statement to give "the date, time, and place of the deposition."[4]
2. **Appearances** list everyone present, and show who was remote. (Rule 30(b)(5)(A)(v); CRB Pointer No. 11.)
3. **The oath** is reflected, and the interpreter's where one was used. (Rule 30(b)(5)(A)(iv).)
4. **Indexes**: every examination and exhibit indexed to the right page, and the exhibit index matches every "marked" reference. (CRB Pointer No. 13.)
5. **Speaker IDs, bylines, and special terms** swept after the read. (Baker's dozen tips.)
6. **Numbers, dates, Bates, and money** checked against the audio and the exhibits. (Lee; the Eastern District of Washington newsletter.)
7. **Spellcheck and doubled words**, and the preference sheet applied across every volume and every scopist. (Baker's dozen tips.)
8. **The closing statement**, with any stipulations about custody. Rule 30(b)(5)(C) requires the officer to "state on the record that the deposition is complete" and set out those stipulations.[4]
9. **The certificate** states the oath and the accuracy of the record, whether review was requested, and attaches any changes. Texas adds items such as "the amount of time used by each party at the deposition."[35]
10. **A signature and errata page** where read and sign was not waived. (Rule 30(e); California section 2025.520.)
11. **No rough-draft labels or page and line references** from the rough remain. (Texas UFM § 4; CRB Pointer No. 4.)
12. **The electronic copy** meets its rules: California appeal transcripts must "Be in full text-searchable PDF (portable document format) or other searchable format approved by the court" and "Permit users to copy and paste, keeping the original formatting, but with headers, footers, line numbers, and page numbers excluded,"[36] and Texas records "must not be password-protected, encrypted, or protected by rights management."[2]
13. **Confidential volumes** line up with the open transcript, and no confidential exhibit is described in the open index. (CRB Pointers No. 5 and 13.)

Then the record leaves your hands. Federally, the officer seals the deposition in an envelope or package "bearing the title of the action and marked 'Deposition of [witness's name]'" and sends it promptly to the attorney who arranged for it.[4] A federal court transcript carries the certification "I (we) certify that the foregoing is a correct transcript from the record of proceedings in the above-entitled matter."[1] Retention runs by state: California, for one, keeps deposition notes "not less than one year from the date on which the transcript is produced," and longer when no transcript is made.[20]

## Responsibility, confidentiality, and AI

The file carries its obligations with it. NCRA's position statement on outsourced transcription says: "Whenever, during the process of creating the record, the work in progress leaves the custody and control of the court reporter him- or herself, the accountability for the security and confidentiality of its contents and compliance with all applicable laws in the jurisdiction pass with it to the scopist, proofreader, transcriptionist, production facility, or whomever."[37] California makes confidentiality a licensing standard: licensees must "Maintain confidentiality of information which is confidential as a result of rule, regulation, statute, court order, or deposition proceedings," and "Perform professional services within the scope of one's competence."[38]

On AI, 2026 brought two NCRA statements. The first is a position statement adopted by NCRA's Board on February 24, 2026.[39] Its last two sentences: "We recognize the potential of Artificial Intelligence (AI) as a supplemental tool but firmly hold the position that AI or Automatic Speech Recognition (ASR) technologies cannot replace a trained, certified, impartial human court reporter and/or captioner in the creation, preservation, and certification of the official record. Human oversight is essential to ensure accuracy, context, reliability, and ethical stewardship of the record."[37] The second is COPE guidance published on July 17, 2026.[40] Its main lines:

- "AI may be used only to support the stenographic reporter's work and not to generate the record. Permissible uses could include technical research and/or clarification of spelling and vocabulary."[10]
- "Disclosure should affirm that the stenographic reporter created, reviewed, and certified the record."[10]
- "There must be no unauthorized transcript uploading, data sharing, storage, or reuse of proceeding content."[10]
- Its policy statement allows AI "only as a limited assistive tool for the stenographic reporter's sole use in quality control."[10]

The guidance does not address scopists. Our conservative reading: a scopist working for a reporter may use AI for spelling and term research, but should not let it generate or rewrite transcript text, and should not upload the transcript anywhere without the reporter's authorization.

One appeals court has noticed. In two 2026 memorandum decisions, the Indiana Court of Appeals catalogued transcript errors and remarked that "it appears that generative artificial intelligence may have assisted with the preparation of this transcript." In *Williams v. State* (July 23, 2026), the court found "numerous typos that change the meaning of the testimony, question, or objection," names "reported incorrectly," a motion attributed to the trial court, an objection attributed to the Bailiff, and the State's closing argument attributed to the trial court.[41] In *Orr v. State* (March 6, 2026), a witness's answer appeared inside an attorney's question and the reverse, answers and questions were mislabeled, and an objection was attributed to the trial court.[42] Both opinions add the same advice: "While AI can improve efficiency and be a productive tool for many professionals, it is incumbent upon those using such systems to proofread and ensure the accuracy of the generated product." That is the wording in *Williams*; *Orr* ends "generated responses."[41][42]

Read those cases carefully. They are memorandum decisions, which in Indiana are not binding precedent and may be cited only for persuasive value. The AI remark sits in a footnote and is an observation ("it appears … may have assisted"), not a finding of fact. And in *Williams* the court said the errors "complicated but did not substantially impede our review."[41] The lesson for editors stands anyway. Indiana's rule is that "The Court Reporter shall certify the Transcript is correct,"[43] and misattributed speakers are exactly what Pass 4 exists to catch. Our reading: a transcript cleaned by a machine can read cleaner than it is, because its errors are well-formed words in the wrong place.

## Where the craft came from

Much of the method above is old, and so is the division of labor. The tools changed who touches the words and how fast. The certificate did not move.

| When | What changed |
| --- | --- |
| 1888 | According to sound historian Patrick Feaster, citing the Washington Post of April 29, 1888, the first official delivery of a Bell-Tainter graphophone went to Washington court reporter Edward Easton, and reporters began dictating their shorthand into the phonograph for typists.[44] |
| 1944 | The Judicial Conference adopted the uniform federal transcript format.[1] |
| 1963 | Stenograph dates its first Data Writer, which recorded machine notes on magnetic tape for computer transcription, to 1963.[45] |
| 1975 to 1976 | Baron Data of San Leandro, California, initiated the computer-aided transcription system it was still marketing in 1981.[15] |
| End of 1980 | Five vendors had viable systems, with "about 345 CAT computers operating in approximately 280-300 sites," and eleven state courts had court-sponsored CAT.[15] |
| 1981 | The NCSC studied six established court CAT systems: "All of them employed a scoper to assist in operating the system and editing."[15] |
| 2026 | NCRA's AI position statement and COPE guidance define AI as assistive, with the reporter as certifier.[39][40] |

The 1981 report is worth reading for how familiar it sounds. It predicted CAT would "encourage individual reporters to purchase or lease their own edit-only terminals for home or office usage," which describes today's scopist's desk. It hoped the technology would "allow sufficient time to proofread final transcripts to ensure high accuracy." And it called the scoper "a mechanism for handling volume or scheduling problems," not a prerequisite.[15] Dictation to typists, then a scoper at a terminal, then synced audio and speech recognition: each step moved mechanical work off the reporter's desk. None of them moved the signature. That is the sense in which this craft modernizes a tradition rather than replacing it.

## Doing it well for years

Scoping and proofreading are long-session work, and the evidence on pacing is thin but useful.

- **Breaks.** The baker's dozen tips warn against trying to read 400 pages in one sitting.[12] A NIOSH field study of 42 data-entry workers found that adding a five-minute break each hour reduced discomfort and eyestrain, and "These beneficial effects were obtained without reductions in data-entry performance." It studied data entry, not proofreading.[46]
- **Playback speed.** A study of lecture videos found "minimal costs incurred by increasing video speed from 1x to 1.5x, or 2x speed, but performance declined beyond 2x speed," but they measured gist comprehension, not word-for-word checking.[47] A small conference study found that a 50 percent speed-up in background noise made all four talkers harder to understand, "but the effect was much greater for the female talkers than the male talkers."[48] Slow down for the hard passages, and do not treat 2x as free.
- **Throughput.** Evenson's test for a new scopist is about viability, not pay: "if they can't get 20-25 pages an hour after they've put in a little time with someone, it's not worth their while."[14] Measure your own pages per hour by job type rather than relying on anyone's average.
- **Fresh eyes.** The defamiliarizing habits in Pass 3 (a new font, reading aloud, a pause between passes) are also how you stay accurate on page 300.

## What nobody has measured

A guide should say where the evidence stops. We looked, and here is what we did not find.

- **We found no recognized credential for steno scopists.** NCRA defines the role, but we found no certification for it.
- **We found no published accuracy standard for a finished transcript.** The only numeric bars are exam floors. NCRA's RPR requires "95 percent accuracy on each leg to pass,"[49] and California's dictation and transcription part passes at 97.5 percent.[26] Those measure raw skill under time pressure. The legal bar for the finished record is qualitative: it "accurately records the witness's testimony."[4]
- **We found no current, independent survey of scoping or proofreading rates.** That is why this guide quotes none.
- **We found no research on professional scopists.** The proofreading studies above used students and general readers.

Until those gaps close, the method is what holds the work together: correct what the machine got wrong, never change what the witness said, flag the rest, and let the person who signs decide.

## How DepoStack is approaching this

Transcript by DepoStack is Planned to help you check a rough draft against its audio. You would still decide every change. The review steps in this guide apply to any tool, ours included: suggestions wait for the reporter's call, and the reporter certifies. The software never does. Follow the plans on the [Transcript by DepoStack page](https://depostack.com/transcript/).

## Sources

Reviewed October 13, 2026.

1. [Guide to Judiciary Policy, Vol. 6: Court Reporting (Administrative Office of the U.S. Courts)](https://www.uscourts.gov/sites/default/files/guide-vol06.pdf) (uscourts.gov; Ch. 5 last revised April 25, 2024: §§ 520, 520.40.10, 520.40.20(e), 520.63.30; Ch. 3: §§ 310.20.20, 320.50.30)
2. [Uniform Format Manual for Texas Reporters' Records](https://www.txcourts.gov/media/190785/Uniform-Format-Manual-07012010.pdf) (txcourts.gov; records prepared on or after July 1, 2010: §§ 2.9, 3.9, 3.12, 3.13, 3.15, 4.2, 8.3)
3. [Michigan State Court Administrative Office, Manual for Court Reporters and Recorders](https://www.courts.michigan.gov/4ade4c/siteassets/publications/manuals/crr/crr_mnl.pdf) (courts.michigan.gov; Section 5, Transcript Format; Section 8, Reference Materials)
4. [Federal Rules of Civil Procedure, Rule 30 (Legal Information Institute)](https://www.law.cornell.edu/rules/frcp/rule_30) (law.cornell.edu; Rule 30(b)(5), 30(e), and 30(f)(1))
5. [NCRA Advisory Opinion 4, Altering the official record](https://www.ncra.org/home/the-profession/NCRA-Code-of-Professional-Ethics/COPE-Advisory-Opinions/group-page-%28advisory-opinions%29/Advisory-Opinion-4) (ncra.org; written 1987, revised 2013)
6. [NCRA Advisory Opinion 47, Inclusion into the transcript of comments picked up by videotape but not heard by reporter](https://www.ncra.org/home/the-profession/NCRA-Code-of-Professional-Ethics/COPE-Advisory-Opinions/group-page-%28advisory-opinions%29/Advisory-Opinion-47) (ncra.org; 2015)
7. [NCRA, Scopists](https://www.ncra.org/home/the-profession/Court-Reporting/Scopists) (ncra.org)
8. [California Code of Civil Procedure section 2025.540](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CCP&sectionNum=2025.540) (leginfo.legislature.ca.gov; subdivisions (a) and (b))
9. [NCRA Advisory Opinion 34, Certifying a transcript from a source other than directly from the reporter](https://www.ncra.org/home/the-profession/NCRA-Code-of-Professional-Ethics/COPE-Advisory-Opinions/group-page-%28advisory-opinions%29/Advisory-Opinion-34) (ncra.org; 1998)
10. [NCRA/COPE Guidance, Use of Artificial Intelligence by Stenographic Court Reporters and Captioners](https://www.ncra.org/docs/default-source/uploadedfiles/operations/ncra-notices/ncra-cope-guidance-july-2026.pdf?sfvrsn=f140e508_7) (ncra.org; July 17, 2026: Foundational Principle; §§ I.3, I.4, I.5, I.8; § IV)
11. [Code of Ethics for Certified Shorthand Reporters and Court Reporting Firms (Supreme Court of Texas, Misc. Docket No. 18-9060)](https://www.txcourts.gov/media/1441402/coe-court-reporters-cr-firms-2018-website.pdf) (txcourts.gov; April 12, 2018: §§ 3(c) and 13)
12. [Journal of Court Reporting (NCRA), PROOFREADING TIPS: A fresh and tasty baker's dozen](https://www.thejcr.com/2017/06/30/proofreading-tips-a-fresh-and-tasty-bakers-dozen/) (thejcr.com; Brenda Rogers-Fiscus, Deborah Smolinske, and Beverly Thomas, June 30, 2017)
13. [Journal of Court Reporting (NCRA), Proofreading tips and conundrums](https://www.thejcr.com/2017/03/07/proofreading-tips-and-conundrums/) (thejcr.com; with NCRA's Proofreading Advisory Council, March 7, 2017)
14. [Journal of Court Reporting (NCRA), Get time and a teammate with a scopist](https://www.thejcr.com/2022/03/07/get-time-and-a-teammate-with-a-scopist/) (thejcr.com; Megan Stolz Rogers, March 7, 2022)
15. [National Center for State Courts, Computer-Aided Transcription in the Courts: Executive Summary (NCJRS 75958)](https://www.ojp.gov/pdffiles1/Digitization/75958NCJRS.pdf) (ojp.gov; February 1981)
16. [Journal of Court Reporting (NCRA), What does a scopist do?](https://www.thejcr.com/2021/06/25/what-does-a-scopist-do/) (thejcr.com; Linda M. Lee, June 25, 2021)
17. [U.S. District Court for the Eastern District of Washington, Attorney Newsletter: Your Day in Court, Tips to Assist in Capturing the Verbatim Record](https://www.waed.uscourts.gov/sites/cst/waed/files/documents/Attorney-News-Letter---Verbatim-Transcript.pdf) (waed.uscourts.gov; Vol. 1, Issue 1, August 2026)
18. [Stenograph Solution Center, Check It](https://solutioncenter.helpjuice.com/checkit) (solutioncenter.helpjuice.com)
19. [Court Reporters Board of California, Best Practice Pointer No. 1, How to Interrupt Proceedings](https://www.courtreportersboard.ca.gov/formspubs/best_interrupt.pdf) (courtreportersboard.ca.gov; revised 3/2025; the Board's pointers are guidance, not regulations)
20. [California Code of Civil Procedure section 2025.510](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CCP&sectionNum=2025.510) (leginfo.legislature.ca.gov; subdivisions (e) and (g))
21. [Journal of Court Reporting (NCRA), Forwards, backwards, print side down, NCRA's PAC gets it done](https://www.thejcr.com/2024/02/26/forwards-backwards-print-side-down-ncras-pac-gets-it-done/) (thejcr.com; February 26, 2024)
22. [Schotter, Bicknell, Howard, Levy, and Rayner, Task effects reveal cognitive flexibility responding to frequency and predictability: Evidence from eye movements in reading and proofreading, Cognition 131(1)](https://doi.org/10.1016/j.cognition.2013.11.018) (doi.org; 2014)
23. [Burgoyne, Saba-Sadiya, Harris, Becker, Brascamp, and Hambrick, Revisiting the self-generation effect in proofreading, Psychological Research](https://doi.org/10.1007/s00426-022-01699-3) (doi.org; 2022)
24. [Cushing and Bodner, Reading aloud improves proofreading (but using Sans Forgetica font does not), Journal of Applied Research in Memory and Cognition 11(3)](https://doi.org/10.1037/mac0000011) (doi.org; 2022)
25. [Sylvia Vitello, What impacts success in proofreading? A literature review of proofreading on screen vs on paper (Cambridge University Press & Assessment)](https://www.cambridgeassessment.org.uk/Images/666540-what-impacts-success-in-proofreading-a-literature-review-of-proofreading-on-screen-vs-on-paper.-.pdf) (cambridgeassessment.org.uk; November 16, 2022)
26. [Court Reporters Board of California, 16 CCR §§ 2420, 2422, and 2473, Final Approved Language](https://www.courtreportersboard.ca.gov/lawsregs/ccr_2420_2422_2473_final_app_lang_09-11-24.pdf) (courtreportersboard.ca.gov; filed September 11, 2024, operative January 1, 2025)
27. [Texas Judicial Branch Certification Commission, Uniform Format Manual](https://txcourts.gov/jbcc/court-reporters-certification/statutes-rules-and-resources-for-court-reporters-firms/uniform-format-manual/) (txcourts.gov; preface)
28. [NCRA, COPE Guidelines for Professional Practice](https://www.ncra.org/home/the-profession/NCRA-Code-of-Professional-Ethics/cope---guidelines-for-professional-practice) (ncra.org; remote proceedings; Section III, uncertified draft transcripts)
29. [Court Reporters Board of California, Best Practice Pointer No. 13, Exhibit Handling for Depositions](https://www.courtreportersboard.ca.gov/formspubs/best_exhibit.pdf) (courtreportersboard.ca.gov; revised 3/2025)
30. [Court Reporters Board of California, Best Practice Pointer No. 5, Confidential Depositions](https://courtreportersboard.ca.gov/formspubs/best_conf_depositions.pdf) (courtreportersboard.ca.gov; revised 3/2025)
31. [California Government Code section 69957](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=GOV&sectionNum=69957) (leginfo.legislature.ca.gov; subdivision (a))
32. [Court Reporters Board of California, Best Practice Pointer No. 11, Videoconference/Remote Reporting](https://www.courtreportersboard.ca.gov/formspubs/best_video_conf_11.pdf) (courtreportersboard.ca.gov; revised 3/2025)
33. [Court Reporters Board of California, Best Practice Pointer No. 4, Rough Draft Transcripts](https://www.courtreportersboard.ca.gov/formspubs/best_rdtranscripts.pdf) (courtreportersboard.ca.gov; revised 3/2025)
34. [California Code of Civil Procedure section 2025.520](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CCP&sectionNum=2025.520) (leginfo.legislature.ca.gov)
35. [Texas Rules of Civil Procedure (Texas Judicial Branch compilation)](https://www.txcourts.gov/media/1463426/texas-rules-of-civil-procedure.pdf) (txcourts.gov; compilation dated October 1, 2026: Rules 203.1 and 203.2)
36. [California Rules of Court, rule 8.144](https://courts.ca.gov/cms/rules/index/eight/rule8_144) (courts.ca.gov; subdivision (d)(1); amended effective January 1, 2024)
37. [NCRA Policies and Procedures Manual, Section 11, Position Statements](https://www.ncra.org/home/about-ncra/leadership-and-governance/NCRA-Policies-and-Procedures/policies-and-procedures-manual---section-11---position-statements) (ncra.org; Ensuring the Confidentiality and Security of Outsourced Transcription; NCRA Artificial Intelligence (AI) Position Statement)
38. [16 CCR § 2475, Professional Standards of Practice (Legal Information Institute)](https://www.law.cornell.edu/regulations/california/16-CCR-2475) (law.cornell.edu; operative January 1, 2014)
39. [Journal of Court Reporting (NCRA), NCRA Board adopts new AI position statement](https://www.thejcr.com/2026/03/02/ncra-bod-adopts-new-ai-position-statement/) (thejcr.com; March 2, 2026; adopted February 24, 2026)
40. [Journal of Court Reporting (NCRA), NCRA COPE AI best practices and guidelines](https://www.thejcr.com/2026/07/17/ncra-cope-ai-best-practices-and-guidelines/) (thejcr.com; July 17, 2026)
41. [Williams v. State, No. 25A-CR-3041 (Ind. Ct. App. July 23, 2026), memorandum decision](https://public.courts.in.gov/Decisions/api/Document/Opinion?Id=wtB2Hgu_u-EGG0gxlduyw5RFUGIdMV59K7ubp5KhLmL2-8QdzriKfOXKz8o_oLWx0) (public.courts.in.gov; slip op. at 2 to 3 and n.1; not binding precedent)
42. [Orr v. State, No. 25A-CR-1786 (Ind. Ct. App. March 6, 2026), memorandum decision](https://public.courts.in.gov/Decisions/api/Document/Opinion?Id=hYNyaVCcYFXQInQU8XG2I0b6mt2Ii4r3fz-2uLOWkSMNDlh4gG4w7lf-l7ZBiuBY0) (public.courts.in.gov; slip op. at 5 n.1; not binding precedent)
43. [Indiana Rules of Appellate Procedure, Rule 28](https://rules.incourts.gov/Content/appellate/rule28/current.htm) (rules.incourts.gov; Rule 28(B))
44. [Patrick Feaster, Ever Heard Phonographic Court Reporting, Interrog? (Griffonage-Dot-Com)](https://griffonagedotcom.wordpress.com/2020/03/21/ever-heard-phonographic-court-reporting-interrog/) (griffonagedotcom.wordpress.com; March 21, 2020, citing the Washington Post of April 29, 1888)
45. [Stenograph, History of Writing Machines](https://www.stenograph.com/history-writers) (stenograph.com; vendor history)
46. [Galinsky, Swanson, Sauter, Hurrell, and Schleifer, A field study of supplementary rest breaks for data-entry operators, Ergonomics 43(5)](https://doi.org/10.1080/001401300184297) (doi.org; May 2000)
47. [Murphy, Hoover, Agadzhanyan, Kuehn, and Castel, Learning in double time: The effect of lecture video speed on immediate and delayed comprehension, Applied Cognitive Psychology 36(1)](https://doi.org/10.1002/acp.3899) (doi.org; 2022)
48. [Johnson and Ferguson, Does increasing the playback speed of men's and women's voices reduce their intelligibility by the same amount? (Acoustical Society of America lay language paper)](https://acoustics.org/does-increasing-the-playback-speed-of-mens-and-womens-voices-reduce-their-intelligibility-by-the-same-amount-eric-m-johnson-sarah-hargus-ferguson/) (acoustics.org; 171st ASA meeting, May 2016; not peer reviewed)
49. [NCRA, Registered Professional Reporter](https://www.ncra.org/certification/NCRA-Certifications/registered-professional-reporter) (ncra.org)

Court rules, statutes, and professional guidance change and vary by jurisdiction. The rules that govern your proceeding, and your certifying body's guidance, control. The tables and checklists marked as ours are DepoStack's synthesis of the sources cited, not an industry standard.

## Related Field Notes

- [The State of Scoping and Transcript Review in 2026: Drafts, Audio, AI, and Who Decides](https://depostack.com/blog/state-of-transcript-review-2026/)
- [What Is CAT Software? A Guide for New Court Reporters](https://depostack.com/blog/what-is-cat-software/)
- [Realtime Court Reporting: Equipment, Setup & Getting Started in 2026](https://depostack.com/blog/realtime-court-reporting/)