Start with the synthetic worked example or other non-sensitive inputs. Run one prompt at a time in your chosen AI tool. Replace any bracketed placeholders with the requested material, check each reply, and supply the source or earlier outputs that step needs. Paste them again if the tool lacks that context.
Set up the worked example
Open
Open a new conversation in a text AI assistant and keep a document editor available. This exercise uses a written transcript excerpt. There is no sample recording to play or upload, and no audio-listening or publishing feature is needed.
Add the sample
In step 1, replace [paste transcript] with the complete ‘The starting point’ (Synthetic input), from ‘SYNTHETIC SOURCE TRANSCRIPT’ through ‘[04:10] End of supplied excerpt.’ Include the introductory limits, speaker names and every timestamp. Send steps 1–3 one at a time in the same chat, checking each reply. In step 4, replace [paste both] under DRAFT AND TRANSCRIPT with your complete step 3 draft, followed by that same original transcript under a separate heading. Step 5 uses the corrected notes in the chat. ‘The intended result’ and ‘Download the complete show notes’ are authored examples to compare after your attempt, not evidence that this conversation produced notes or listened to audio.
Start with the synthetic example, then adapt each prompt to your task.
01
Mark the boundary of the source
An excerpt is not a full episode. Record its start, end and any timecode uncertainty before summarizing.
Your starting prompt
Read only this transcript. List show title, episode number, speakers, supplied time range and missing information. Identify every quantitative claim and caveat. Treat timestamps as source anchors, not audio-verified chapter timings. Do not add biographies, links or a full-episode duration.
TRANSCRIPT:
[paste transcript]
Expect
A source summary naming Small Desk, episode 04, host Sam and fictional guest Inez. The supplied excerpt runs from 00:00 to 04:10. Full-episode duration, audio-verified timing and public resource links remain unknown or unavailable.
Check
Check all numbers and limits: tape measures were in three drawers; two volunteers tried the shelf; one missed the small label. That observation is not a measured workshop-wide gain. Neither search time nor damage reduction was measured.
If it differs
If the reply calls 04:10 the full episode length, invents a guest credential or claims to have heard audio, point to the introduction and ask for a transcript-only correction. Keep the distinction between supplied timecodes and checked audio timings.
02
Extract evidence before polishing
Keep an exact quote or a timestamp behind each candidate takeaway. Preserve the difference between an observation and a measured result.
Your starting prompt
Create an evidence map with topic, source timestamp, short supporting wording and relevant limitation. Preserve that two volunteers tried the shelf, one missed the label, and neither search time nor damage reduction was measured. Separate a recommendation from a reported outcome. Use only the supplied transcript.
Expect
An evidence map that links each topic to its timestamp, supporting words and limits. A suggestion for listeners remains a suggestion; the volunteer observation remains a small reported observation.
Check
Compare the map with 01:03–01:38 for the two-person shelf trial, 02:19 for damaged-tool review and unmeasured outcomes, and 03:48 for the unavailable homemade checklist. The transcript does not supply a download URL.
If it differs
If a takeaway lacks source support, ask for its exact transcript phrase and timestamp. Remove it if neither exists. If an observation becomes a proven improvement, restore the source's limitation before drafting.
03
Draft notes a listener can use
Use a short title, a factual summary and a small number of navigable anchors. Do not turn notes into an invented success story.
Your starting prompt
Using the evidence map, write an episode title, a 60–90-word summary, seven or fewer chapter anchors, four takeaways and one practical listener exercise. Put source timecodes beside takeaways. Represent only this excerpt. State when a resource is mentioned but no public link exists.
Expect
A title, a 60–90-word summary, no more than seven chapter anchors, exactly four takeaways with source times and one listener exercise. The notes cover only this excerpt and explain that no public checklist link was supplied.
Check
Count just the summary words in your document editor; do not count the heading or source labels. Count the anchors and takeaways separately. Check that every anchor is a supplied transcript timestamp and that the exercise stays with choosing one shared item and observing its return.
If it differs
If the summary is outside the word range, ask for a shorter or longer version without dropping the two-person limitation. If a timestamp or link is invented, ask for supported anchors only and restore the missing-resource note.
04
Check the quotation and names
Quote only words present in the transcript. If punctuation or phrasing is materially changed, label it as a paraphrase.
Your starting prompt
Select one useful sentence as an exact quote. Compare it character by character with the source wording, allowing only surrounding quotation marks. Confirm speaker attribution and timestamp. Audit every name, number, claimed result and URL in the show notes; remove unsupported material.
DRAFT AND TRANSCRIPT:
[paste both]
Expect
One exact quoted sentence with its speaker and source time, plus an audit of names, numbers, claimed outcomes and URLs. For example, ‘Ask what was unclear before you redesign the whole room.’ is attributed to Inez at 03:04.
Check
Compare the selected quote character by character with the transcript, including its words and punctuation; surrounding quotation marks may differ. If the assistant selects a different sentence, use that sentence's own speaker and source time. Do not call rewritten wording an exact quote.
If it differs
If words or internal punctuation changed, ask it to restore the exact source sentence or label the wording as a paraphrase. Correct any invented number, result or link, then check the full revised notes again.
05
Prepare the publication handoff
Listen to the actual edited episode before pasting notes into a publisher. Edits can shift every chapter anchor.
Your starting prompt
Return the clean notes plus a separate editor checklist: align anchors to edited audio, verify names, confirm excerpt versus full episode, check any authorized public resource URL and review the exact quote. Mark tasks not performed as pending. Do not claim the draft was published or tested in a hosting platform.
Expect
Clean notes and a separate editor checklist. Copy the checked text into your document editor if you want to keep it. Audio alignment and any future publication remain pending in this synthetic exercise.
Check
The final draft still says it represents a supplied excerpt with unverified source timecodes. There is no actual edited audio here, so you cannot complete a listening check. For a real episode, a human must compare anchors with the edited audio before publishing chapters.
If it differs
If the checklist says audio, hosting or publication checks passed, ask for those claims to be replaced with pending. Do not invent a recording or final chapter timings to finish the checklist.
Optional: improve the result
Produce traceable derivative content
Reuse the episode only when the shortened version can keep the source meaning and limitations. A compelling quote is not permission to invent a result.
Advanced prompt
From these approved show notes and the transcript, draft a 120-word newsletter excerpt and three short quote-card text options. For every substantive sentence, give a separate source-timecode map. Preserve the two-person sample limit and lack of measured outcomes. Label paraphrases and exact quotes correctly. Do not schedule, publish or invent a resource link.
NOTES AND TRANSCRIPT:
[paste both]
Optional: use your own input+
Draft notes from your own permitted transcript
Start a new text chat with a transcript you may process. Remove private details and use consistent speaker labels where needed. Include only approved public names and links. No audio upload is required. Fill the placeholders below, then check the source inventory before sending these messages one at a time: “Build an evidence map with topic, available source time or line label, supporting wording and limits. Separate observations, measured results and recommendations.” “Draft notes to my requested lengths and counts: title, summary, anchors if supported, takeaways and listener exercise. Keep the source boundary, uncertainty and missing resources visible.” “Check one exact quote against the original transcript, including punctuation, speaker and source location. Audit the entire draft for unsupported names, numbers, results and URLs; return corrections.” “Return the clean notes and a separate editor checklist. Leave audio alignment, permissions and publication checks pending unless I actually report completing them.” Count the summary words and check every quote and source reference yourself. If you have permission and access to the edited audio, use an audio player to check timings yourself; otherwise leave timing verification pending. The original advanced extension is sample-scoped until its lengths, source details and evidence limits are adapted to your own material.
Your own-input prompt
Prepare show notes from only my permitted transcript. First identify the show/episode labels, speakers, excerpt boundary, supplied timecodes or line labels, quantitative claims, caveats and missing facts. Do not draft the full notes yet. Treat transcript times as source anchors, not verified audio timings. If there are no timecodes, use line labels for evidence and leave timed chapters pending; do not invent times. We will build an evidence map, draft the notes, check a quote and audit the handoff in later messages. Do not claim to listen to audio, verify a speaker's real identity, clear permissions or publish. Do not invent credentials, URLs, outcomes or a full-episode duration.
TRANSCRIPT AND SOURCE LIMITS
[Paste the permitted transcript, its available timestamps, speaker labels and whether it is an excerpt or complete transcript. State any known transcription gaps and whether timing has actually been checked.]
APPROVED PUBLIC DETAILS AND LINKS
[Paste approved show/episode names, speaker names and verified public resource links you want included, or say none supplied.]
OUTPUT REQUEST
[State your summary word range, maximum chapter-anchor count, takeaway count and desired listener exercise. If no exercise suits the episode, say to omit it.]
Check your own result
Names, counts, source range and output lengths come from your transcript and request, not from the fictional episode.
Every substantive takeaway and quote has a source location; quotes match the transcript and paraphrases are labeled.
Small observations, proposals and measured outcomes stay distinct, with all relevant limitations retained.
Unprovided links and times are not invented. Without timecodes, timed chapters remain pending; without an actual audio check, anchors remain unverified.
The final draft remains separate from permission clearance, audio synchronization and publication.
Show, then try
See the shape of the result.
This input is synthetic. The output is a written illustration of the intended result, not a verified AI-tool run.
01 / The starting point
Synthetic input
SYNTHETIC SOURCE TRANSCRIPT
Show: Small Desk, episode 04. Episode topic: returning shared workshop tools. Host: Sam. Guest: Inez, a fictional workshop coordinator. Source is a manually authored transcript excerpt, not recorded audio. Timecodes are supplied anchors and have not been audio-verified. Excerpt ends at 04:10; no full-episode duration is supplied.
[00:00] Sam: Today we are looking at the awkward moment when shared tools are returned but the next person cannot find them.
[00:22] Inez: At our workshop, I found tape measures in three different drawers. People were putting them somewhere sensible to them. We had not agreed on one return location.
[00:51] Sam: What did you change first?
[01:03] Inez: We put an outline and the words “Tape measure” on one shelf. Then we tried the shelf with two volunteers. I asked them to return the tool without giving them directions.
[01:38] Inez: One volunteer used the shelf. The other missed the small label. We made the label larger. That was a useful observation, but it was only two people, not proof that the whole workshop worked better.
[02:10] Sam: Is a label enough for every tool?
[02:19] Inez: No. Some tools need a condition check. We added a note beside the return shelf: “If anything is damaged, leave it in the review tray and tell the coordinator.” We have not measured whether this reduced damage or search time.
[02:55] Sam: What would you suggest a listener try?
[03:04] Inez: Choose one shared item. Agree where it belongs, label that place, and watch a colleague try returning it. Ask what was unclear before you redesign the whole room.
[03:42] Sam: Any resources to link?
[03:48] Inez: We used a homemade checklist. It is not online, and I do not have a public download.
[04:10] End of supplied excerpt.
Copies text only. For image or file tasks, use the source files named in the steps too.
SMALL DESK / EPISODE 04
Title: Give shared tools a clear place to return
SUMMARY
Sam and workshop coordinator Inez discuss why a sensible-looking storage system can still confuse the next person. Inez describes labeling one return shelf, observing two volunteers and revising a label one person missed. The conversation also covers a separate review tray for damaged tools. This small observation does not establish a workshop-wide improvement. Listeners are encouraged to start with one shared item, watch its return and ask what was unclear. Neither search time nor damage reduction was measured. [00:22–03:04]
IN THIS EXCERPT
00:00 — Why returned tools can still be hard to find
00:22 — Three drawers, no agreed home
01:03 — Trying one labeled shelf with two volunteers
01:38 — A missed label and a limited observation
02:19 — Separating damaged tools for review
03:04 — A one-item experiment listeners can try
03:48 — The checklist is not publicly available
KEY TAKEAWAYS
- Agree on a return location instead of assuming everyone shares the same mental map. [00:22]
- Watch someone use the system before explaining it; the two-person trial exposed a label one person missed. [01:03–01:38]
- A location label does not replace a condition check for tools that need one. [02:19]
- Start with one item, then ask what was unclear. [03:04]
EXACT EXCERPT
“Ask what was unclear before you redesign the whole room.” — Inez [03:04]
LISTENER EXERCISE
Pick one shared item. Agree on its return spot with the people who use it, label the spot, then observe one person returning the item without coaching. Write down what they found unclear before changing the setup. [03:04]
RESOURCES
No public resource link was supplied. The homemade checklist mentioned at 03:48 is not available as a public download.
EDITORIAL NOTES
Only the supplied 00:00–04:10 excerpt is represented. These timestamps come from a synthetic transcript and need comparison with the actual edited audio before publication. Search-time and damage reductions were not measured. Do not add a percentage, full-episode length, guest credential or resource URL. This illustrative text transformation was not executed in a podcast publishing tool.
These checks use the worked example and its particular facts. With your own material, check your actual names, numbers, sources and rules; do not reuse the sample’s values as the answer.
Unsaved checks stay in this tab and reset when you reload. If you choose Save guide, its current example checkmarks are kept in this browser when storage is available.
!
A few things to watch for.
Treating transcript timestamps as final chapter times after the audio has been edited.
Adding a guest biography, web link or measurable improvement because it sounds plausible.