Ghost Jobs in 2026: Is That Technical Req Real?

A ghost job is a posting the employer has no active intention of filling. It is not usually a scam and not usually a fake company — it is a real organization advertising a role that no one is hired against. For a software engineer, the cost is not the application, which takes minutes. The cost is the four-round loop and the eight-hour take-home you complete for a req that was never going to close.

Two things changed in 2026. Prevalence estimates got worse and less consistent, and state legislatures started writing disclosure rules that would put the employer's own hiring timeline on the face of the ad. This article covers what those laws would require, why the published prevalence numbers disagree so much, and how to limit your exposure at the point where it actually costs you something — the interview loop.

How common are ghost jobs, and why do the numbers disagree?

The honest answer is that the range is wide and the definitions are not the same.

Greenhouse analyzed postings on its own platform and reported in January 2025 that between "18% and 22% of job postings last quarter" were ghost jobs, which it defined as "listings published online with no intention of actively hiring." Some sectors ran far higher, with construction at 38%. Separately, HR Dive's June 2026 coverage of New York's legislation cited MyPerfectResume research putting "nearly 1 in 3 U.S. job listings" as not resulting in a hire.

Those two figures are measuring different things, which is why they diverge. "No intention of hiring" is a claim about employer intent. "Did not result in a hire" is an outcome, and it sweeps in reqs that were genuinely open and then died — budget pulled, headcount frozen, an internal candidate promoted, the hiring manager left. A posting can be entirely sincere on Monday and dead by Friday without anyone having deceived you.

Treat the widely circulated single-number claims with suspicion. Most originate in vendor surveys with small samples, self-selected respondents, and definitions chosen after the fact. The defensible statement is narrower: somewhere between roughly one in five and one in three advertised roles will not produce a hire, and you cannot tell which from the posting text alone. That uncertainty, rather than any specific percentage, is what should drive your behavior.

What the new disclosure laws would actually require

This is the part that is genuinely new, and it is the reason the topic is live right now rather than a perennial complaint.

New York, S8877. The legislature passed the bill on 2 June 2026. Reporting through late July 2026 describes it as awaiting Governor Kathy Hochul's action, and as of 10 September 2026 we found no reporting that it had been signed or vetoed. If enacted, it would take effect immediately. Its mechanics:

• It applies to employers with 100 or more employees and to third-party job posting platforms.
• Every ad must carry one of three disclosures, in bold capital letters: that the role is a current vacancy the employer intends to fill by a stated date; that it is not a current vacancy and will be filled no sooner than a stated date; or that it is not a current vacancy and the employer is collecting resumes for future review.
• Postings must come down within two weeks of the role being filled.
• Enforcement sits with the New York Department of Labor, at $2,500 per non-compliant publication, escalating for each additional 30-day period the ad stays uncorrected.

Pennsylvania, HB2321. Introduced 26 March 2026 by Representative Jim Prokopiak and, as of this writing, still in the House Labor and Industry Committee. It is broader than New York's in two ways that matter to a technical candidate. It would require employers to disclose the hiring timeline, intended hire and start dates, salary range, "the extent of AI use in hiring, and how many times the role has been posted in the previous year." It would also cap applicant-data retention at one year and penalize employers that "mine or sell applicant information" at up to $25,000 per applicant. Its threshold is lower, at 50 employees.

That repost-count provision is the single most useful disclosure either bill contains. A req advertised four times in twelve months tells you something no amount of reading between the lines of the job description will.

Note what none of these laws do: they do not require anyone to hire you, and they do not make a posting binding. They are labeling rules. They also arrive alongside a broader wave of hiring-transparency obligations — several states now require AI-use disclosure in screening, which we cover in our guide to what AI hiring rules entitle you to know.

A ghost job is not a scam, and the difference matters

These get conflated constantly, and the response to each is different.

A ghost job is posted by a real employer with a real careers page. Your data goes into a real applicant tracking system. Nobody hires you, and you have lost time. Pennsylvania's data-mining provision exists because some of those postings do exist mainly to harvest resumes, but the employer is who it says it is.

A fraudulent posting is a different animal: an impersonated company, a "recruiter" moving you to an off-platform chat, an interview process that exists to extract credentials, personal data, or an up-front payment. In technical hiring the sharpest version is the weaponized take-home — a repository you are asked to clone and run that carries a payload. If your instinct is that a process feels wrong rather than merely slow, that is a different checklist, and we wrote it up separately in how to check whether a take-home assignment is malware.

Slow, disorganized, and unresponsive is the signature of a ghost job or a dying req. Urgent, off-platform, and asking you to run unfamiliar code is the signature of fraud.

Signals that actually correlate at the interview stage

Generic ghost-job checklists lean on posting age, which is weak evidence — plenty of real reqs sit open for months because the bar is high. These are more diagnostic once you are in contact with a human:

• The recruiter cannot name the hiring manager or the specific team. For a funded req, both exist and are not secret.
• There is no described loop. A real process has a shape: screen, technical, system design, panel, and someone can tell you roughly how many rounds and over what period.
• A take-home arrives before any conversation with an engineer. Front-loading unpaid work ahead of any human assessment is the pattern most worth resisting, and it is worth reading our rules for take-home assignments before you accept one.
• The language is pipeline language. "Always accepting applications," "building a bench," "future opportunities," and "evergreen req" are, under New York's proposed scheme, exactly the postings that would have to say so in bold capitals.
• Scheduling slips repeatedly with no explanation. Reqs that lose their funding rarely get formally cancelled; they just go quiet.
• The same posting reappears with a new req number. This is what Pennsylvania's repost disclosure would surface directly.

None of these is conclusive alone. Two or three together justify capping your investment rather than withdrawing.

Bound the cost instead of trying to guess

You cannot reliably identify a ghost job in advance, so optimizing for detection is the wrong goal. Optimize for exposure instead, because applications are cheap and preparation is not.

Ask the timeline question early and plainly: is this a funded, currently open req, and when does the team expect someone to start? Most recruiters answer honestly. Evasion on a direct, reasonable question is itself information. If New York's bill becomes law, this question gets answered on the ad before you ask it, at least for larger employers advertising into that state.

Cap unpaid work at a fixed number of hours you decide in advance, and say so. "I can spend about four hours on this — I'll submit what I have at that point" is a normal professional boundary, and a real hiring team will accept it.

Prepare for the format rather than the company where you can. Time spent on system design fundamentals or on your own project narratives transfers across every loop; time spent building a bespoke deliverable for one employer does not. When you are genuinely deep in a real loop and preparing seriously, Control is a desktop AI interview assistant for Windows and macOS with a free allowance of five messages and two minutes of voice, so you can evaluate whether it fits your format before paying for anything. It will not tell you whether the req is real. Nothing will.

What none of this fixes

A real req can still evaporate for reasons no disclosure rule reaches. Headcount gets frozen mid-loop. An internal candidate is chosen after four external rounds. A reorganization deletes the team. Under New York's bill, an employer that genuinely intended to fill a role and then could not would have made a truthful disclosure and still wasted your month.

The laws also carry meaningful gaps. New York's threshold exempts employers under 100 people, which is most startups. Pennsylvania's bill has not left committee. Neither imposes a private right of action in the reporting reviewed here — enforcement runs through a state agency, so an individual candidate's remedy for a mislabeled ad is a complaint, not a lawsuit. And a determined employer can satisfy a labeling rule while still running a process that goes nowhere.

The short version

Between roughly one in five and one in three advertised roles will not produce a hire, the published estimates disagree because they measure different things, and no posting-level tell reliably separates the categories. What is changing is disclosure: New York's S8877 passed the legislature on 2 June 2026 and would force a bold, explicit statement of hiring intent plus a two-week takedown after a role is filled, and Pennsylvania's HB2321 would go further by requiring repost counts and AI-use disclosure.

Until those are law where you are applying, the workable strategy is unglamorous. Ask directly about funding and timeline, treat evasion as data, cap unpaid work at a number you set beforehand, and put your preparation into skills that transfer across loops rather than artifacts that serve one employer. If you want the most useful next read, start with what the 2026 AI hiring rules entitle you to know — the same transparency push driving ghost-job legislation is what gives you the right to ask how you are being screened.