Every job on Seekless comes from the employer’s own hiring system. We don’t buy feeds from job boards, and we don’t copy listings from LinkedIn, Indeed or Glassdoor. This post goes step by step through how a job gets onto Seekless, and says where the process falls short.
Step 1: finding the companies
Most companies post jobs through an applicant tracking system, or ATS: software such as Greenhouse, Lever, Ashby, Workday, SmartRecruiters or Workable that hosts the company’s job board. Seekless looks for those boards directly. Every night it searches three places for new ones:
•
Google’s Vertex AI Search, queried for board pages on those six systems.
•
Common Crawl, the public archive of the web, which turns up Workday, SmartRecruiters and Workable boards a search engine misses.
•
Y Combinator’s company list, through a community-maintained copy of it. For each company, Seekless looks for an ATS board or the company’s own careers page.
A board becomes a company on Seekless only after the ATS’s own API confirms it exists.
You can add a job yourself by pasting a link. If the link points at one of those systems, Seekless reads the job straight from it. If it points somewhere else, a LinkedIn post say, Seekless reads only the employer’s name and the job title from it, finds that employer’s own careers page, and crawls that. The job you end up with is the one on the employer’s site.
Job boards and aggregators (LinkedIn, Indeed, Glassdoor, Wellfound, Built In and others) are on a blocklist: none of them can ever be recorded as a company’s website.
Step 2: reading the jobs
Every other night, starting at 00:30 Eastern, every company is marked due, and a pool of crawlers works through them. Each company is read in up to three passes:
1.
The ATS’s own feed. For eight systems (Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Workable, Rippling and Gem), Seekless reads the job board’s data feed. This is the cleanest source there is: the employer’s postings exactly as their system stores them.
2.
The careers page, in a real browser. For every other site, Seekless opens the careers page in a browser, follows “view all jobs” links, presses “load more”, and collects the postings.
3.
Reading the page. It takes the structured job data a site publishes for search engines where there is some, then known career-site formats, then each job’s own page. A language model is the last resort, for pages none of those methods understand.
Step 3: checking what came back
•
Duplicate postings with the same ID are merged into one.
•
Every description gets a quality score from 0 to 1, from its length, structure, completeness, noise and encoding. Anything under 0.5 is never published, and borderline descriptions get a second check by a language model.
•
A posting confidently identified as not written in English is not published.
•
A “posting” that isn’t a job, such as “join our talent community” or “general application”, is recognized and taken down.
Step 4: understanding each job
Each posting is then read for about 35 structured details: seniority, whether it’s remote, hybrid or on-site, pay and how often it’s paid, the years of experience it asks for, travel, shifts, equity, the technologies and requirements it lists, and more. Those details are what let you search in a sentence, like “senior backend roles in Denver paying over $150k”, and filter precisely. Only a posting that has been read this way is published.
What we don’t do
•
No paid placement. Seekless has no such thing as a sponsored or promoted job. Results are ordered by how well a job matches what you asked for, adjusted for how recently it appeared and for signs that it’s no longer actively hiring. No company can pay to change that.
•
No reposting. We don’t republish other job boards’ listings, and when you apply, you apply on the employer’s own site.
Where it falls short
•
Freshness. The crawl runs every other day, so a listing can be up to about two days behind the employer’s board. Every job shows when we last saw it live; how that works is its own post, How we check that a job is still live. •
Recruiting agencies. Some boards belong to staffing agencies advertising their clients’ roles. Seekless screens every board for this and labels the ones it believes are agencies, but for now it doesn’t remove them automatically. So far one has been removed by hand.
•
Coverage. Discovery covers six ATS hosts, Y Combinator companies and the links people paste. A company with its own careers site and none of those connections can be missing. If you find one, paste a link to one of its jobs and it gets added.