How the index is built
AIWorkJobs is a discovery layer, not the employer. This page explains what we do to the source data, what we refuse to invent, and how listings leave the index.
Last reviewed 2026-11-19
What the counts mean
Headline counts show active, unexpired, unique roles. The underlying database also contains expired rows and duplicate rows; those are retained as closed or duplicate records but are excluded from active counts. The public JSON feed reports the active total alongside the excluded expired and canonicalized duplicate counts.
Sourcing
We read active roles from a separately managed jobs database. The underlying records come from public job pages and authorized feeds operated by the platforms named on each listing. The web application has read-only access and never writes to that database.
Each role keeps its source platform, external identifier where provided, source application URL, and the source’s own pay and eligibility fields. We do not create placeholder jobs or publish a listing to fill a category page.
Normalization
We clean titles, engagement types, workplace types, pay periods, and country codes so roles can be compared. We preserve the source’s meaning. If the source does not state a country, the listing says eligibility is unspecified; we do not assume worldwide.
Pay keeps its original currency and basis. A per-task or fixed amount is never converted into an hourly rate. Implausible pay fields are withheld and flagged rather than displayed as fact.
Categories and specialties
The five top-level hubs (AI training, data annotation, AI evaluation, RLHF, and remote AI work) group the candidate experience. Specialty labels such as finance, law, healthcare, and language work come from the source platform’s category taxonomy, matched to each job in the database. They are used as filters, not as invented seniority or qualification claims.
Deduplication
A role is treated as a duplicate when the normalized title, company, location, and eligible-country set match another current role. The newest record with a usable application link becomes canonical. Other records stay reachable for old links but are excluded from search and the sitemap, and they point to the canonical page.
We do not merge roles only because the titles are similar. Different geographies, contract terms, or requisitions can be separate jobs.
Expiry
Every job row carries a source expiry timestamp. Expired roles leave the active index, active counts, related-role blocks, and sitemap, but remain on the site as closed records. You can browse them in the closed-roles filter and in the recently closed section on company and category pages. Closed pages have no Apply button and no job-posting structured data. This keeps old links useful without sending candidates to a dead application.
What we do not do
- We do not invent employers, salaries, countries, or deadlines.
- We do not submit applications or auto-apply on your behalf.
- We do not require an account or résumé to view an external application link.
- We do not sell placement in the organic job index.
- We do not present a platform logo as an endorsement or partnership.
Corrections
If a listing is wrong, closed, duplicated, or points to the wrong destination, send the job URL and the problem to support@aiworkjobs.com or use the contact form. We review reports against the source page and correct or remove listings that cannot be verified.
Machine-readable data
A public metadata feed is available at /jobs.json. It contains active, canonical roles with source, pay basis, eligibility, and links. Job facts remain subject to the source platform’s terms; attribute AIWorkJobs and the named source if you reuse the compilation.
