AIREAPER
Plate VI · How We Calculate Risk

Methodology.

How AIREAPER calculates AI displacement risk — and what we don't claim.

Data layers

Every score combines five independent data sources, each weighted by how directly it measures displacement risk:

Oxford / Frey-Osborne (2013)
Baseline automation probability across 702 occupations
WEF Future of Jobs Report 2025
Sector-level net displacement estimates
BLS JOLTS
Live US hiring and separation trends by industry
GDELT news monitoring
Real-time layoff and restructuring sentiment
SEC EDGAR full-text search
Workforce reduction language in 8-K and 10-Q filings
O*NET skills database
Task-level automation exposure for your specific role
Yahoo Finance / Market signals
Stock price trends and 30-day performance for 100+ publicly traded companies — used as a leading indicator of restructuring pressure

Source confidence

Not all sources carry equal weight. We assign confidence ranges based on how directly verifiable a signal is:

SEC filing / WARN Act / official government notice90–100
Official company press release80–95
Reuters / Bloomberg / FT / WSJ reporting75–90
Layoff trackers (Layoffs.fyi, TrueUp)60–80
Company careers page changes50–80
Union or labor board statements60–85
Employee forum posts (anonymous, unverified)20–55

What counts as verified

Government filings (WARN Act, SEC EDGAR), official company press releases, and reporting from major financial news outlets. These appear on AIREAPER as confirmed events.

What counts as a signal

Hiring trend changes, job posting volume shifts, GDELT news sentiment, and industry-level macro data (BLS JOLTS). These move your score but are presented as trends, not certainties.

What counts as chatter

Anonymous employee posts and community reports. These are shown as "employee chatter" or "rumor heat" — never presented as confirmed fact, and never the sole basis for a high-risk score.

How often scores update

Risk calculations pull live data on every assessment — BLS JOLTS, GDELT news, and SEC filings are queried in real time. Sector-level baselines (Oxford, WEF) update as new reports are published, typically annually.

What we do not claim

Every score is based on signals, not certainty.
AIREAPER does not predict individual termination decisions.
We separate verified events, risk signals, and employee chatter.
Rumors are never treated as confirmed facts.

Privacy / anonymity model

We separate your identity from public posts, minimize stored verification data, and never expose your email or verification data through client-facing APIs. Your company board identity is separated from your public posts.

Prefer raw data? View the live company Watchlist as a public spreadsheet.
Curious who built this and why? Read the story behind AIREAPER.