Methodology

How JobAware reads jobs from employers, decides which site each belongs to, keeps them current, and measures the figures we publish. Figures on this page were measured 21 Nov 2026, 08:04 UTC.

Where the jobs come from

We keep a registry of employers and the job board each one uses. Today JobAwareBot reads 12,601 employers’ boards.

  • Job board platforms. Most employers publish jobs through a hiring system with a public job board: Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Workable, Rippling, Comeet, Gem and others. We read those boards directly.
  • Careers pages. For employers with their own careers site, we read the page itself. Pages that only show jobs after running JavaScript are opened in a headless browser that also identifies itself as JobAwareBot.
  • Never other job sites. We don’t take listings from aggregators or other job boards, so every job links back to the employer, and applications go to the employer.

How often

Each board has its own schedule. The median board is read every 8 hours. The 50 employers at the top of the Attention Head AI Index are read more often; the status page shows how fresh their boards are right now. A new or changed posting reaches our sites at the board’s next read, and a posting that leaves the board closes at the next read. The status page shows how fresh the index is right now.

Which site a job belongs to

Each site has rules, written in code, that decide which jobs belong to it. They start from the job’s title. Where the title is too general to decide, they also read its team, its employer or its description: a “Senior Mobile Engineer” for which platform, an “Insights Analyst” who writes SQL or not. Each site documents its own rules and how accurate they are:

A job can belong to two sites. It then has one page, on the site its title points to, and the other site links to it. A “Data Scientist, LLM Evaluation” role is a data job about AI: it lives on Left Join and is listed on Attention Head.

Pay figures

The pay chart on the home page and the sites’ pay pages use the same method:

  • Which roles. Open roles whose posting states an annual range in US dollars: 16,720 today, 37% of all open roles.
  • One role, one value. Each role (employer and title) counts once, at the midpoint of its stated range. A place at the end of a title that repeats where it is posted (“Engineer - Madrid, Spain”) is not part of the title. A role posted in several cities with different ranges counts once, at the average of its midpoints.
  • Plausible values only. Midpoints below $20,000 or above $1,000,000 are left out, since a few boards put hourly rates in the annual field.
  • Levels. A role’s level comes from a level word in its title: “Senior”, “Staff”, “Principal” and so on. The home page charts senior roles, and staff, principal and above. Titles that name no level mix every level, so they are left out, as are junior, manager, director and executive titles, which are fewer.
  • The range shown. The 25th and 75th percentiles: the middle half of roles. The line inside is the median.
  • Small groups. A level with fewer than 10 roles on a site is not shown.

These are the ranges employers state, not what people are paid. Many postings state no range, and postings in other currencies are not included yet.

Closed jobs

When a job leaves the employer’s board, its page on our sites tells search engines it is gone (HTTP 410) and drops out of search results.

What we don’t do

  • We read only public pages, and we respect each site’s robots.txt.
  • We store no data about candidates.
  • We don’t rewrite postings. Titles, descriptions and pay ranges come from the employer.