Skip to main content
OPEN DATA · CC BY 4.0

The data is the point

1,840 UK careers scored for AI exposure, with entry difficulty grades, moat classifications and a permanent change ledger, published as machine-readable mirrors of the same pages humans read. Free to use with attribution. Measured August 2026.

What you can fetch

/en/yellow/careers.json

The whole register in one fetch: every scored career with exposure today and at 20 years, door grades and moat, plus definitions and the ledger. Future Door, job security and roles fields are retained for compatibility but are not forecasts and are no longer displayed on the site; see outlook_note.

Try it →
/en/yellow/career/{slug}/card.json

The full card as JSON: exposure today, 5/10/20-year horizons, door grades, moat (or an explicit not-classified), pay, rank among all scored careers, nearest neighbours, the change ledger. Future Door, job security and roles fields are retained for compatibility but are not forecasts and are no longer displayed on the site; see outlook_note.

Try it →
/en/yellow/career/{slug}/ai.md

The verdict as clean markdown, built for AI assistants and anyone quoting us: the sentence, the table, the sources, the citation line.

Try it →
/en/yellow/embed/card/{slug}

The live card as an embeddable iframe for councils, hubs and school sites: no key, always-current, with the source link built in. Height ~460px, width up to 420px.

Try it →
/llms.txt

The manifest for AI crawlers: what this platform is, which URLs are canonical for which question, and the definitions to use verbatim.

Try it →
/sitemap.xml

Every card, verdict, subject hub and tool, per host.

Try it →

Responses are cached and CORS-open, and carry an ETag so a conditional request (If-None-Match) gets a cheap 304 instead of the full payload. There is no key, no rate card and no login; if you are building something heavy on top, we would simply like to hear about it.

Stability

Every JSON surface above carries a schema_versionfield, and ai.md carries the same number as a "Schema version" line in its footer. Currently "1.0". This is the promise attached to it.

Fields do not moveWithin a major version, the number before the dot in schema_version, no field already documented on this page is ever renamed or dropped. Build against a field name once and it stays valid.
Additions are freeNew fields on the JSON surfaces and new lines on ai.md can appear at any time as the register grows, with no version bump. Read what you recognise and ignore the rest.
The ledger records itA shape-breaking change, a rename or a removal, bumps the major version and is recorded permanently in the ledger every card and verdict already carries. Nothing is silently re-stated.

Licence and citation

Careermash measurement data is licensed CC BY 4.0. Cite as "Careermash, Measured August 2026" and link the career's verdict page. Underlying sources carry their own terms: AI exposure blends Anthropic's 2026 labour market research (observed real-world AI usage by occupation) and OpenAI's "The AI Jobs Transition Framework" (Richmond 2026, CC BY 4.0). Human necessity (moat) categories are based on OpenAI's AI Jobs Transition Framework. CareerMash maps these to UK careers and may apply UK-specific adjustments, for example for statutory licensing and reserved legal activities. The 5, 10 and 20 year figures are Careermash scenarios, not forecasts. Every revision to a claim is recorded permanently in each career's ledger; nothing is silently re-stated. The full account, including how our measurement relates to the official UK exposure index, is on the methodology page.