usdigitalmarketingindex.com SEO
How can a video transcript be prepared to support discovery of its spoken subject matter?
Updated November 15, 2026
By turning what is said in the video into accurate, readable text on the same page as the video: the words as spoken, corrected against the recording, with speakers identified, divided under headings by topic, and marked with time cues so a reader can jump to the moment. Speech inside a video file is not text on a page. A transcript makes the content available to people who cannot hear it, to people who would sooner read, and to search engines, which index the words a page contains. It supports discovery of what the video covers. It does not guarantee that the video itself ranks or appears in video results, which depends on other things. Its first duty is fidelity: a transcript reports what was said.
Transcript accuracy and speaker cues
Automatic speech recognition gives a draft. A person makes it a transcript.
Errors to correct in a machine draft:
| Error type | Example |
|---|---|
| Technical terms | A product name or unit rendered as a similar-sounding common word |
| Names | People, companies, and places misspelled |
| Numbers | "fifteen" heard as "fifty"; units dropped |
| Homophones | Words that sound alike |
| Missing punctuation | Run-on text with no sentences |
| Speaker confusion | Two voices merged |
| Dropped negatives | "can" for "can't", which reverses the meaning |
| Inaudible passages | Guessed words |
Numbers and negatives need the closest check, since an error there changes what the transcript claims. In instructional or safety content, check every figure and every “do” and “do not” against the audio.
Decide a style and apply it throughout.
| Choice | Options | Recommendation |
|---|---|---|
| Verbatim or clean | Every "um" and false start, or lightly cleaned | Clean verbatim: remove fillers and stumbles, change no meaning |
| Corrections of the speaker | Leave errors, or fix them | Leave the words as spoken; add a bracketed editor's note where the speaker misspoke on a fact |
| Speaker labels | Names, or roles | Names on first appearance with role; then names |
| Non-speech sound | Include or omit | Include where it carries meaning: [boiler fires], [laughter] |
| On-screen text and actions | Include or omit | Describe briefly in brackets where the speech depends on them: [points to pressure gauge] |
| Inaudible | Guess, or mark | Mark as [inaudible]; do not guess |
Do not improve the content. A transcript that adds facts the speaker did not state, tightens a hedge into a claim, or inserts search phrases is no longer a transcript. If the page needs more, put it in a separate written section.
Quotations elsewhere on the page that are attributed to the speaker must match the transcript.
Readable sectioning
A wall of text is accurate and unusable. Structure it.
- Break into paragraphs at changes of thought.
- Add headings at changes of topic. Headings are editorial additions, so word them as neutral labels of what follows.
- Add time cues at each heading, and at intervals in long sections.
- Keep speaker labels at each change of speaker.
A short sample transcript excerpt. The video and speakers are fictional.
Transcript: Checking boiler pressure (6:12)
Speakers: Dana Reed, heating engineer; Sam Ortiz, presenter
[00:00] Introduction
SAM ORTIZ: Today we're looking at the pressure gauge on a combi
boiler, and what the reading tells you. Dana, where do we start?
DANA REED: Start with the heating off and the system cold. The
gauge is on the front panel on this model. [points to gauge]
[00:48] Reading the gauge
DANA REED: The needle should sit inside the band your manual
gives for a cold system. On this one the manual says between one
and one and a half bar. Yours may differ, so check the manual
for your model.
SAM ORTIZ: And if it's below that?
DANA REED: Then the system has lost pressure, and the manual
will have a section on topping it up. If it keeps dropping after
that, there's a leak somewhere and you need an engineer.
[02:31] When to call an engineer
DANA REED: Anything involving the gas side, or opening the
casing, is not a job for the homeowner. [inaudible] ...that's the
rule.
What the sample shows: a title with duration, speakers introduced with roles, time cues at topic changes, headings that label, an action noted where the speech refers to it, and an inaudible passage marked instead of invented.
Time cues become useful when they are links that move the player to that point. Where the player supports it, make each cue jump to its moment.
Audio has the same need without the picture. How a podcast episode page should present what was said is covered in the guide to how podcast episode pages should summarize audio-only material.
Accessibility and captions
Captions and transcripts are related and different.
| Format | What it is | Serves |
|---|---|---|
| Captions | Text synchronized with the video, shown on the player | Viewers who are deaf or hard of hearing; anyone watching without sound |
| Transcript | The full text, on the page, readable without playing | People who read instead of watch; screen reader and braille users; search engines |
| Audio description | Narration of important visual content | Blind and low-vision viewers |
Provide both captions and a transcript. A caption file can be the starting point for the transcript, and the same corrections apply to both.
For accessibility, a transcript should include what a viewer would otherwise miss: who is speaking, meaningful sounds, and visual information that the speech relies on. “As you can see here” is useless in text unless the transcript says what is seen.
Accessibility requirements for video are set by law in some jurisdictions and sectors. Which standard applies to a given organization is a matter for its own legal and accessibility advisers.
Rights. A transcript reproduces what people said. Confirm before publishing:
- that the organization has the right to publish the recording and its text;
- that interviewees and guests agreed to publication;
- that quoted third-party material, such as a clip or a reading from a book, may be reproduced in text;
- that personal information spoken in passing, such as a customer’s name or address, is removed or was consented to.
Integration with page content
Where and how the transcript appears.
| Decision | Recommendation |
|---|---|
| Location | On the same page as the video, below it |
| Format | HTML text, not an image or a download only |
| Visibility | Present in the page's HTML; if collapsed behind a control, the full text should still be in the served markup |
| Length | A long transcript can sit under a clear heading, after a summary |
| Summary | A short written summary above the transcript, for readers who want the gist |
| Duplication | The transcript appears once, on the video's own page; other pages link to it |
| Links | Add links to sources or products mentioned, in a section beside the transcript; do not alter the spoken words to carry them |
| Captions file | Attached to the player |
| Structured data | The schema.org VideoObject type has a transcript property; if it is used, keep it identical to the visible text |
A page layout:
Title of the video
[Video player, with captions]
Summary: three or four sentences on what the video covers
Chapters: linked time cues
Key points: a short list, written, clearly separate from the transcript
Transcript: full text with speakers, headings, and time cues
Related: links to the written guide and to sources mentioned
Keep the editorial sections and the transcript distinct. Summary and key points are the publisher’s words. The transcript is the speakers’.
QA before publishing:
| Check | Method |
|---|---|
| Accuracy | A person reads the transcript while listening to the whole recording |
| Numbers and names | Checked individually |
| Speakers | Labels correct at every change |
| Time cues | Each cue lands on the right moment |
| Inaudibles | Marked, not guessed |
| Nothing added | Compare for words the speaker did not say |
| Rights and personal data | Reviewed |
| In the HTML | Fetch the page; the transcript text is present |
| Matches captions | Same wording |
| Updated with the video | If the video is re-edited, the transcript is redone |
On what to expect: the transcript adds the video’s spoken content to the page as text, so the page can be found for what was said. Google’s guidance on videos sets its own conditions for a video to be indexed and shown in video features, and this guide does not claim a transcript changes those outcomes. Treat the transcript as serving readers and the page.
Whether the video should have a page of its own to hold all this is a prior decision, covered in the guide to when a video should have its own watch page instead of appearing only within an article.
The rule: transcribe what was said and nothing else, correct it against the recording, label speakers, add headings and time cues, publish it as text on the video’s page, and keep the publisher’s own summary separate.
Add your comment to this answer