Where Job Listings Come From: ATS Sources, Aggregators, and Reposts

A job listing from any employer big enough to run hiring software starts in the same place: a requisition opened inside an applicant tracking system. When a recruiter publishes that requisition, the ATS renders it wherever the employer has pointed its careers page, usually a public board the vendor hosts. Everything else you see is downstream. Aggregators ingest those postings through vendor APIs, employer feeds, or crawling; staffing agencies re-post client roles under their own name; and listings stay visible after the underlying requisition closes, because closing a req and removing every copy of the posting are different actions. The useful skill is tracing a listing back up that chain, and the URL you apply through usually tells you most of it.

Layer one: the employer's ATS

The requisition is the original. It carries the internal job id, the hiring team, the approval state, and the pipeline that applications land in. The public listing is a rendering of it, which is why the ATS-hosted page is the closest thing to a source of truth a candidate can see from outside. These pages are recognizable by their hostnames: greenhouse.io, lever.co, ashbyhq.com, smartrecruiters.com, workable.com, teamtailor sites, or a company subdomain at breezy.hr, recruitee.com, personio.com, or myworkdayjobs.com. The pattern is broadly consistent across vendors. The employer's name sits in the host or the first path segment, and a posting id sits further along. It breaks in one common case worth knowing: an employer can front its ATS with its own careers domain, so a company URL carrying a job-id parameter is often an ATS posting wearing the employer's colors.

Two useful consequences. First, if the apply flow leaves the board you were reading and lands on one of those vendor domains, your application goes into an applicant tracking system rather than into a re-hosted form on someone else's site. That tells you where the application lands. It does not tell you the requisition is still open, and it does not tell you whose ATS it is: staffing firms run their own boards on the same vendors, under their own name. Second, the posting id in that URL is a stable handle for that copy. Matching ids across two boards is good evidence you are looking at one role listed twice. Different ids are weaker evidence of anything, because one requisition can be published as several posts.

Layer two: how a posting reaches an aggregator

Aggregators do not have the requisition. They have a copy of the listing, acquired one of a few ways, and the way it was acquired shapes everything about how accurate that copy stays.

• Vendor job board APIs. Greenhouse, Lever, Ashby and others publish documented endpoints that return an employer's open roles as structured data, readable without credentials. This is the cleanest path: the employer's own fields, the employer's own text, and an apply link pointing back at the ATS.
• Vendor feeds in RSS or XML. Some systems expose the same thing as a feed rather than a JSON API. Equally clean, usually with fewer fields.
• Employer-supplied feeds. Large boards accept a feed the employer maintains and submits directly. Also clean, though the employer controls the refresh cadence.
• Crawling. A board discovers postings by fetching careers pages and extracting fields out of HTML. Serviceable, but extraction is lossy, and a page that stopped being maintained keeps returning content long after the role closed.
• Paid distribution and partner syndication. A posting is redistributed through a network, sometimes several hops from the employer, sometimes with the title, location, or company name rewritten on the way.
• Manual posts by third parties. A recruiter at a staffing firm posts a client's role under the firm's name, often without naming the client at all.

Coverage gaps follow from this directly, and they are mostly a plumbing fact rather than a hiring fact. Not every system offers an open public endpoint. Some route bulk access through a dedicated partner feed the vendor provisions on request, which each employer then opts into, so a board without that arrangement simply cannot see those roles. Others expose postings only as pages meant for human readers. When a company's openings never appear on any independent board, the company is usually still hiring. The door downstream is just narrower than it looks.

What a copy loses

Once a listing is copied, its metadata belongs to the copy. The posted date on an aggregator is typically the date that aggregator first saw the listing, not the date the role opened, which is why one job reads as posted today on one site and months ago on another. It gets subtler than that: some vendor endpoints return only a last-updated timestamp rather than a first-published one, so even a board reading the employer's own feed may be showing you when the posting last changed. An honest board says which one it is showing. Many say nothing, and the field is simply labeled posted.

The rest of the record erodes the same way. Titles get normalized into the board's own taxonomy, so a specific internal title becomes a generic one. Locations get simplified, and a hybrid arrangement becomes remote or becomes a headquarters address. Compensation gets inferred, estimated, or dropped. None of this is dishonest by itself. It just means an aggregator's version of a posting is a derived artifact, and derived artifacts drift from the thing they were derived from.

How to tell you are looking at a repost

There is no certain test from the outside, but the signals stack up fast, and every one of them is checkable in under a minute.

• Follow the apply button before you write anything. If it opens a form hosted by the board rather than the employer's ATS or careers domain, your application may be entering a database that is not the employer's pipeline.
• Look for the role on the employer's own careers page. If a company publishes its openings and this one is not among them, the listing is stale, agency-sourced, or bait.
• Note the posting id in the URL. It will not tell you whether a listing is legitimate, but it lets you match duplicates across boards.
• Read the company field. A description in place of a name (a leading fintech, a Fortune 500 client) means an intermediary sits between you and the employer.
• Compare the text against the employer's own version. A rewritten title, missing sections, or a tidier salary line all say the copy passed through someone else's normalization.
• Treat a posted date that keeps resetting as weak evidence either way. It can mean the posting was edited, refreshed by the board, or re-listed, and it rarely means a role reopened.
• Watch for contact bait. A listing that asks you to email a resume to an individual address, or to register somewhere else before showing details, may be collecting contacts rather than filling a role. Verify the employer before you reply.

Why closed roles stay up

Be precise here, because the folklore runs well ahead of the evidence. The mechanical explanations are unglamorous and sufficient on their own: closing a requisition and removing the posting from every place it was syndicated are separate actions, feeds refresh on their own schedules, crawled pages keep returning content until something changes them, and cleanup is nobody's job. Separately, there are legitimate practices that keep a posting live without a specific opening behind it, such as evergreen requisitions used to build a pipeline for roles a company hires continuously.

Notice which boards can even tell. A board reading an employer's complete list of open roles can watch a posting disappear from that list and mark it closed. A board holding a crawled copy has nothing to compare against: absence looks identical to a fetch that failed, so the safe move is to leave the listing up. That asymmetry, not malice, is the reason a copy has no reliable way to know a role closed.

What this changes about how you apply

Provenance is triage. It does not make a role better or worse, but it tells you where your effort is going. A tailored resume and a real cover note are expensive to produce, and they are worth spending on listings you can trace to an employer's own pipeline. A listing you cannot trace deserves the cheap version of your attention: verify before you invest, and go find the employer's careers page rather than trusting the copy in front of you.

It is also the honest way to judge a job board as a category. A board that reads an employer's own hiring system and links you back to that posting is doing something structurally different from one that re-hosts an apply form and rewrites the title. Our sibling jobs property, YouSourced, is built on the first side of that line: it reads companies' own instances of the hiring systems named above, and the link it hands you is the posting URL that board reports, so applying means leaving us. Federal roles are the one exception, and they are labeled as such: those come from the USAJOBS search API and link back to usajobs.gov. Its sourcing policy page (linked below) names every system it reads. Do not take any of that on faith. The test works the same on any board: click apply, and look at where you land.

The underlying idea outlives job search. A listing is a record with a source, and every copy is a claim about that source made by someone with their own incentives. Reading the chain back to the origin is the whole skill: which system holds the original, who copied it, what the copy changed, and whether the original is still true. Job boards are just where most people meet that problem with real money on the line.

Sources

• Greenhouse Job Board API documentation (public, credential-free employer boards) ↗
• Lever postings API documentation (the public postings endpoint) ↗
• Ashby job postings API documentation ↗
• Ashby dedicated partner job feeds (how bulk access is provisioned, and opted into) ↗
• YouSourced sourcing policy: every hiring system the feed reads ↗