Methodology and publication scope

The public site now limits itself to three source classes: current BLS occupation pages, live public ATS job boards, and primary-source layoff disclosures such as SEC filings or company investor-relations releases.

If a data series cannot yet be reproduced from those source classes, it stays off the public charts and off the public page copy, except for a small set of clearly labeled high-confidence layoff reports.

• BLS Occupational Outlook Handbook pages for employment, wage, and projection fields.
• Live Greenhouse and Ashby boards for current role counts, company counts, and posting recency.
• Monitored layoff disclosures from primary documents, plus clearly labeled high-confidence reported events when official documents are not yet public.

The live layoff page publishes filing-grade or official company disclosures that can still be fetched from their source URLs. Those entries are marked Confirmed and remain intentionally narrower than the full market while the broader provenance pipeline is being audited.

A smaller set of Reported entries may appear when a trusted outlet such as Reuters, Bloomberg, The Wall Street Journal, Financial Times, The Information, or the Associated Press reports a layoff before the primary company document becomes public. Those events stay labeled Reported and do not roll into confirmed totals until a stronger source appears.

AI context is kept separate from the fact of the workforce reduction. A company may be investing in AI while cutting staff without saying AI caused the event.

• Confirmed: SEC filing, annual report, WARN notice, or direct company investor-relations statement.
• Reported: trusted financial or wire reporting, including Reuters, Bloomberg, The Wall Street Journal, Financial Times, The Information, or the Associated Press, that has not yet been matched to a public primary document.
• AI cited: the source text itself names AI as an investment priority, restructuring factor, or explicit driver.
• Not cited: the source documents the cut but does not attribute it to AI.

Synthetic displacement ratios, placeholder geography layers, and unaudited research dashboards have been removed from the public research and trends pages until their lineage is auditable end to end.

The occupation page now follows the same rule. If a modeled layer cannot be reproduced from audited public inputs, it stays off the public site until that work is complete.

The public occupation page publishes only directly sourced BLS fields: employment, wage levels, wage percentiles when available, projected growth, and annual openings. Those numbers come from the Occupational Outlook Handbook and related BLS release tables.

AI-exposure mappings, task-overlap estimates, posting-direction models, and transition guidance remain off the public calculator until their provenance, methodology, and reproduction steps can be audited to the same standard as the BLS inputs.

The public research and trends pages publish BLS occupation values and live ATS aggregates. The layoff page publishes confirmed disclosures from official company or SEC documents, plus a small number of clearly labeled high-confidence reported events when a primary document has not surfaced yet. If a series cannot clear that bar, it is removed from the charts rather than published with a weaker standard.

The same rule now applies to the occupation page. If a modeled layer cannot be reproduced from audited public inputs, it stays off the public site until that work is complete.