How JetExpat Works
JetExpat is an AI-generated publication with an automated verification step and a human approval gate. This page sets out exactly how a brief gets made, what we check, and what we cannot catch. If you are going to rely on our numbers, you should know how they got there.
Where the data comes from
- SEC filings — 10-Q and 10-K documents pulled directly from SEC EDGAR, the primary source. We parse the filing itself, not a summary of it.
- Market data — daily closing prices for every tracked ticker, collected after the US market close.
- TSA checkpoint throughput — weekly passenger volumes, as a demand signal.
- Industry news — monitored continuously for context around the filings and price moves.
Every figure we quote traces back to one of these. We do not source numbers from other commentary.
How a brief is written
- Detection — we monitor EDGAR for new filings by tracked companies. A new 10-Q or 10-K triggers a brief automatically.
- Parsing — the filing is broken into its financial statements and management discussion, with management's own wording preserved verbatim rather than paraphrased.
- Drafting — a large language model drafts the script from the parsed filing plus the market and demand data around it.
- Fact-checking — the draft is checked claim by claim against the source data, before any audio is made. See below.
- Human approval — nothing publishes automatically. A person reviews and approves each episode before it goes to the feed, the site, and YouTube.
What the fact-check actually does
This is the step we would want to know about if we were reading someone else's AI-written analysis, so here it is in detail.
Every draft passes through an automated verification layer that re-reads the script against the underlying filing and market data. It checks the things a language model is most likely to get wrong: stock prices and percentage moves, the direction and magnitude of changes, company and ticker identification, figures drawn from the financial statements, and quotes attributed to management. Management quotes are verified against the verbatim text of the filing, not against a summary of it.
The check runs before audio synthesis, and it can block the pipeline. If it finds a critical error, the episode does not proceed to audio and is held for manual review. That ordering is deliberate: it is much cheaper to stop a bad script than to unpublish a bad episode. Every verification report is stored and auditable.
What this does not guarantee
We would rather state the limits than imply there aren't any.
The fact-check verifies claims against source data. It does not verify judgement — an interpretation can be well-sourced and still wrong. It is itself an automated system and can fail; we have found and fixed cases where it did. Filings are complex and a summary is always lossy. And a brief reflects what was known when it was published; we do not silently rewrite old analysis when the world changes.
Treat JetExpat as a well-sourced starting point for your own work, not as a substitute for reading the filing.
Corrections
If we get something wrong, tell us: dev@jetexpat.com. We would much rather hear it from you than not hear it at all.
Material factual errors are corrected on the page, and the correction is noted rather than quietly applied. Where an error reached the audio, we say so in the episode record. Published audio is not always re-cut — re-synthesising an episode is not always practical — but the written record is corrected and the discrepancy is disclosed.
Independence
JetExpat LLC takes no payment for coverage and has no commercial relationship with any company covered. Coverage is determined by which companies file, not by who asks. If that ever changes — through sponsorship or advertising — it will be disclosed clearly on the page it affects.
More about the publication on the about page, or read the full disclaimer.