Before sharing a claim from a PDF podcast, record where you heard it and where the source supports it. A timestamp gets you back to the audio; a page and passage get you back to the evidence.
This is a manual review method, not a walkthrough of PDF AI controls. It does not require an automatic transcript or source links. You need authorized access to the source PDF, audio you can replay, and somewhere to keep notes. Do not upload a private file simply to complete this check.
1. Identify the two versions
Keep the original PDF unchanged. Record its filename and revision or date. For the audio, record an available filename, generation date or session identifier, plus a distinguishing title. If you cannot reliably distinguish this audio from another version, resolve that before marking a claim checked.
Use elapsed playback time, not remaining time. Note a short start–end interval around the claim and replay the surrounding sentence. If your player does not expose timing, use a clearly labeled approximate locator and note that another reviewer may not find the same interval precisely.
2. Write what you heard, then locate the passage
Record the claim manually. Use quotation marks only when you have checked the wording; otherwise label it a paraphrase. If a transcript is available, compare it with the audio rather than treating it as a second independent source. Without a transcript, replay and take notes. If speech remains unclear, mark it unable to verify.
Open the same PDF version used for the audio. Search for a distinctive phrase if search is available, or inspect the relevant section. A failed search does not prove the claim is absent. Record the PDF viewer's page number, the printed page label if different, and a heading plus a short passage locator.
Read the surrounding conditions, not just the matching words. Does the audio describe a proposal as a completed decision, or omit who the statement applies to?
3. Keep a claim-to-source record
The following practice details are fictional and manually written. There is no underlying recording, PDF or product test. Times demonstrate how to fill in a record; they are not measured results.
| Audio version and interval | Heard claim | Source version and location | Source condition | Result and action |
|---|---|---|---|---|
| Practice audio A, 01:12–01:20 | Paraphrase: the workshop will open to everyone in November | Practice plan rev B; viewer page 4, printed page 2; “Pilot scope,” paragraph 2 | “A November pilot is proposed for invited teams, subject to venue approval.” | Mismatch: proposal, audience and approval conditions were lost. Do not share this claim as stated. |
For your own review, use one row per claim:
Audio identifier/version:
Elapsed interval (exact or approximate):
Heard wording or labeled paraphrase:
PDF filename/revision:
Viewer page / printed page label:
Heading and passage locator:
Supporting passage and relevant conditions:
Result: match / mismatch / unable to verify
Action, reviewer and review date:Use match only when the located passage supports the claim with its relevant conditions intact. Use mismatch when you can identify a contradiction or meaning-changing omission. Use unable to verify when the source, version, location or audio is uncertain. A match establishes support in that source, not that the source itself is true.
4. Act on the result and preserve the boundary
In the fictional example, a checked written note would say: “Practice plan rev B proposes a November pilot for invited teams, subject to venue approval.” It is a correction derived from the practice passage, not corrected audio.
Keep an unsupported claim out of your written summary. If the audio contains it and cannot be corrected and reviewed, pause sharing that audio and ask the source owner to resolve the discrepancy. A note in your private log does not repair the recording.
After audio is regenerated or edited, old timestamps may point elsewhere. Start a new audio-version record and replay each relevant interval. If the PDF changes, compare against the new version and review the mapping again. Do not silently carry forward a “match.”
Checking one claim does not certify the whole podcast. State which claims and versions you reviewed, and what remains unchecked.
Choose the next check
For source-text problems, use the scanned-PDF text checking guide. For changes to a document itself, use the edited-PDF review guide.
The PDF AI product page separates store-listing claims by platform; do not assume a transcript or timestamp-to-source control exists in your app. For an application problem, use the support channel in your app or store listing with its version and a non-sensitive reproduction—not a private source document.
Before sharing your next podcast, complete one claim-to-source row and resolve any mismatch or uncertainty it exposes.