Publisher Trust and Editorial Policy
Who publishes LongScribe public content?
The public transcript and citation-dossier surface is published by LongScribe, using the same publisher identity declared across the site: https://longscribe.com/#organization. That identity is the publisher reference used by LongScribe public transcript structured data. This page does not create a second publisher or add claims about outside oversight or institutional authority.
How transcripts are produced
LongScribe turns public audio or video links, and files submitted by users, into text. Ordinary raw transcription output may come from source-provided captions when a platform makes captions available, from automated speech recognition, or from LongScribe processing that combines download, transcription, timestamps and export formatting. Raw user transcripts are automated and are not reviewed by a person before the user sees them.
Public curated transcript pages and citation dossiers are a narrower publishing surface. They are selected for public source pages and may add source links, event/date metadata, summaries, key facts, topic grouping, key exchanges, editorial notes, source segments, citation anchors, coverage manifests and rights notes when those records are available. A curated public transcript or citation dossier should not be read as a guarantee that every LongScribe transcript has received the same treatment.
Editorial treatment
For curated public pages, LongScribe separates source material from LongScribe editorial summaries. A full transcript is used only when the publication record supports reproducing the source text. A citation dossier uses original LongScribe editorial prose tied back to source segments, timestamps, or claim-level provenance. Some older curated pages predate the current job-linked citation workflow and say so in their editorial note instead of pretending to have newer evidence records.
Current dossier validation requires factual blocks to cite source segments, claim indexes to match provenance records, timestamps to derive from source segments, anchors to remain stable, and coverage manifests to account for covered or excluded source ranges. These checks are technical publication gates; they do not make LongScribe a legal authority or a guaranteed source of perfect transcription.
Speaker identity
Speaker names are not inferred from ASR alone. Automated diarization can detect that two passages sound like the same voice, but it does not know who the person is. Public real-person names should appear only when the record supports them, such as explicit self-identification, a named handoff, floor recognition, an official participant list, or comparable source evidence. If identity support is missing, LongScribe should avoid unsupported certainty by omitting the name, using an anonymous speaker cluster, or using a limited role description where that role is safely inferable.
Sources and provenance
LongScribe prefers primary sources for public transcript and dossier pages: the original public recording, official hearing or event recording, or another source URL that lets readers verify the page. Transcript pages link to the source URL when it is available. Citation dossiers connect editorial claims to timestamps, source segments and citation anchors so a reader can move from a LongScribe summary back to the supporting part of the source material.
Editorial summaries are not the original source. They are LongScribe's structured reading of the source record. When precision matters, consult the linked source and use the timestamps or anchors to verify names, dates, figures and quotations.
Copyright and rights
Public availability does not automatically mean unrestricted republication rights. LongScribe's public publishing model is rights-aware: some material can be reproduced more fully when the source record supports that treatment, while commercial or third-party material may require summaries, limited quotations, paraphrase-only handling, or no public publication. Uncertain rights should block publication or full reproduction rather than be treated as permission.
This is an operational publishing policy, not legal advice or a legal guarantee. Source owners and rights holders can report concerns through the issue-reporting channel below.
Corrections
Readers can report transcription errors, speaker-attribution problems, source issues, rights or copyright concerns, metadata mistakes, factual issues in editorial summaries, or broken citation links. Material corrections may update speaker names, timestamps, metadata, transcription wording, source links, rights notes or editorial summaries. If information is known to be unsupported, the correction should replace or remove that unsupported information rather than preserve it silently.
LongScribe does not currently display a public correction-history log on each transcript page, so this policy does not claim one. When a page needs correction, the priority is to make the current public page accurate and appropriately supported.
Accuracy and limitations
Automated transcription can contain errors. Accuracy varies with audio quality, accents, crosstalk, background noise, technical vocabulary, proper names and the quality of any source captions. Speaker attribution can be imperfect, and timestamps may be approximate depending on the source and processing path. Curated summaries can help readers navigate a source, but they are not a substitute for the source itself.
Report an issue
To report a problem, use the Report a problem link in the LongScribe app footer or the contact form in the app. Include the page URL, the source URL if relevant, the timestamp or section involved, and a short description of the issue. Reports can cover transcription wording, speaker identity, source/provenance, rights, copyright, metadata or editorial-summary concerns.