How the check works
What we read, what we can't see, and how the stats are measured.
Job links from employers' own systems
If the link is to a Greenhouse, Lever, Ashby or Workday posting, we ask that system's public job API for the posting. That tells us whether it's still open, when it was published, and what it says. We then look at:
- Age. Postings open more than 45 days are flagged.
- Pay. Whether a pay range appears in the text. Remote US roles without one are flagged, because several states require ranges in postings open to their residents.
- Wording. Talent-pool language (“future openings”, “join our talent community”), roles posted for an unnamed client, very short descriptions, and personal webmail contact addresses.
- Scam patterns. Interviews over chat apps, equipment checks, fees, and requests for a Social Security number or bank details up front.
- History. If the posting is in our index, when we first saw it and whether the same title has been reposted under new IDs.
LinkedIn, Indeed and other job boards
We don't open job-board pages. Instead you give us the company and title, and we look for the role on that company's Greenhouse, Lever and Ashby boards, trying a few likely board names. If we find it, you get the full check on the employer's own posting.
If we don't find it, that isn't proof the job is fake. Many large employers use Workday, iCIMS, Taleo or their own sites, which we can't search by company name yet. Look for the role on the company's careers page yourself, starting from their website rather than a link in the listing.
Pasted text
Text you paste is checked against the scam and wording patterns above, then thrown away.
How the stats are measured
Our scanner reads the public job boards of a list of employers, mostly technology and security companies, on a regular schedule. For every posting it records when we first and last saw it, when it closed, its text and its pay range. A posting that closes and comes back under a new ID with the same company and title counts as a repost.
We publish aggregates only and don't name companies. Figures from other organizations are labelled with their source, date and method.
Company posting records
A company's record uses only postings we've seen on its own job board. Its transparency score, from 0 to 100, is published once we've tracked it for 30 days and seen at least 10 postings:
- 30% for the share of postings that list a pay range
- 30% for the share of open postings that are 60 days old or newer
- 20% for the share of postings not reposted under a new ID
- 20% for the share of postings without talent-pool or “future opportunities” wording
The hiring follow-through score starts as that transparency score. Once at least 10 different applicants have reported on a company through Did they reply?, the share who heard back from a person or got a decision (a rejection, an interview or an offer, not an automatic receipt) counts for 40% and posting transparency for 60%. Each connection can file one report per company, and a new report replaces the old one.
We also show three facts that don't change the score, because each has ordinary explanations:
- WARN Act layoff notices filed in the last 6 months, from the WARN Feed dataset of state notices (CC BY 4.0), matched by company name. A notice can cover one site or division while the rest of the company hires.
- Open postings whose description is identical to at least two others apart from the title. Companies often post one role per city.
- Postings that say the role depends on winning a contract, common with government contractors.
These are measurements of postings and applicants' reports, not conclusions about intent: a role can stay open because it's hard to fill. Companies can send corrections or a statement from their record page.
News and confessions
The news page lists headlines and links collected automatically from public news searches and public forum feeds, filtered for hiring topics. We show only the headline, source and date, and link to the original. Forum posts are unverified accounts by their authors.
New headlines can be sorted by Jev, a classification model from TypeSafe, which answers two questions about each one: what it's about, and whether the speaker hires, is job hunting or is reporting. Confident answers decide whether a post is listed and under which heading; unsure ones stay where the keyword filter put them. Only the public headline and its source are sent, never anything about visitors.
Resume X-ray
Your file is read in memory and discarded when the page is sent. We extract text the way common applicant systems do (tables read cell by cell, text boxes and page headers skipped), then check structure, contact details, dates and each experience line. The match against openings counts how often current, ghost-screened postings for your role mention each skill. Written recommendations come from Claude (Anthropic's AI), which receives the extracted text for that one request and is instructed not to invent facts; any number it uses that isn't in your resume is replaced with a placeholder.
Corrections
If you think a check got something wrong, email hello@resume-pilot.io with the link and we'll look at it.
Check a posting before you apply
Paste a job link and we'll tell you whether it's live on the employer's own careers system, how long it's been open, and whether it shows scam signs. It's free.