If you are applying to tech roles in 2026, roughly one in every two postings you tailor a resume to has no present intention to hire. That is the finding from the Columbia Law Review's November 2025 "Ghost Jobs" forum piece, and it reframes the whole application game: the bottleneck is not your resume, it is the postings you spend the resume on.
The correct response is not more applications. It is a verification pass before you touch the tailoring.
What counts as a ghost job in 2026
A ghost job is a listing by a real company for a position that does not actually exist or for which there is no present intention to hire. That is the Columbia Law Review's November 2025 definition, and it is the one to use because it separates ghost postings from merely slow ones.
The scale is worse than most job seekers assume:
- 48% of information and tech postings are ghosts (Columbia Law Review, Nov 2025).
- 60% in government roles.
- 93% of HR professionals say their employer posts ghost jobs, with 45% saying "regularly" (LiveCareer, March 2025, n=918).
- 8 to 4: in 2019, ten postings produced about eight hires. By 2024, ten postings produced about four.
- 9 hours: average time a job seeker spends on a single ghost-job application cycle (Jobright.ai analysis of 4.4M applications, 2025).
Columbia Law Review's November 2025 forum analysis, with government at roughly 60%.
Congress has explicitly acknowledged there are no official ghost-job statistics (Congressional Research Service, April 2025), so every number here comes from private platforms or academic work. The convergence is what makes it credible: Greenhouse's internal platform data has hovered between 18% and 22% platform-wide in any quarter, and Glassdoor's LLM-driven analysis puts the figure at up to 21%. Tech and government drag the average up.
Why hires-per-posting collapsed, and why it matters for your inbox
The denominator inflated while the numerator did not. Postings kept coming; hires did not. That is the mechanism behind every ghost-job statistic you will read this year.
BLS February 2026 JOLTS shows the receipts:
- 6.9M job openings, down from a revised 7.2M in January.
- Hires rate of 3.1, the lowest level since January 2011, matching the COVID-low of April 2020 (Indeed Hiring Lab).
Compounding that, ghost postings are rarely malicious. The Columbia Law Review is direct about it: they exist because of organizational incentives nobody has a reason to fix. Recruiters build pipelines for headcount that is not yet approved. Companies leave careers pages full so customers and investors read momentum. The role is real next quarter, not this one.
This is the part the "just apply to more jobs" advice misses. If the denominator is inflated with pipeline-building and PR postings, doubling your applications guarantees half your incremental effort is waste. Verification is not optional anymore. It is the highest-leverage move in a 2026 job search.
The recruiter-to-candidate ratio nobody checks
The single strongest ghost-job predictor is how thinly staffed the hiring team is relative to the candidate pool, and you can check it in 30 seconds on LinkedIn. In Refolk's index of professional profiles, the tech-side math is punishing on its own, and the government-side math explains the 60% ghost rate.
| Segment (US) | Count | What it means |
|---|---|---|
| Software Engineers + Senior Software Engineers | 521,918 | Baseline candidate supply behind the 48% ghost rate |
| Technical Recruiters + Talent Acquisition | 22,994 | Humans staffed to move those postings through |
| Recruiters at Government Administration employers | 4,449 | Recruiter capacity behind the ~60% ghost-rate sector |
| Candidates per tech recruiter | ~22.7 | 521,918 ÷ 22,994 |
| Tech SWE candidates per government recruiter | ~117 | 521,918 ÷ 4,449 |
Twenty-two candidates per recruiter is a functioning market. One hundred and seventeen is not. That is why postings languish in government and in any tech org that runs lean on talent acquisition. Before you tailor, open the company on LinkedIn, filter employees by "recruiter" or "talent acquisition," and compare that count to their open roles. A company with 40 open engineering postings and one recruiter is running a pipeline, not a hiring process.
Companies staffing real technical recruiting at scale in Refolk's index include Snowflake, Blue Origin, Google, Figma, Datadog, Ashby, and Glean. That is not a whitelist, but it is a useful reference point for what a genuinely staffed hiring org looks like.
Twenty-two candidates per recruiter is a functioning market. One hundred and seventeen is not.
The 5-minute verification checklist
Run this before you spend an hour tailoring. If a posting fails two or more checks, move on.
1. Repost age and reposting cadence (60 seconds)
Check LinkedIn's "Posted X days ago" and search the exact job title plus company on Google. If the same posting appears with dates spanning more than 60 days, or has been reposted three or more times in a quarter, treat it as a pipeline posting. Real roles close.
2. Recruiter headcount (30 seconds)
On the company's LinkedIn page, filter people by "recruiter," "talent acquisition," or "sourcer." Compare to the number of open roles on their careers page. Anything worse than one recruiter per 25 open roles in tech, or one per 50 in non-tech, is a red flag.
3. Hiring-manager signal (60 seconds)
Search LinkedIn for the likely hiring manager: the director or manager of the function on the team you would join. Look for recent posts about hiring, team growth, or a specific project. Silence on the topic while the posting is live is a real signal. Managers who are actively hiring almost always talk about it.
4. Compensation transparency (30 seconds)
If the role is in a pay-transparency state (California, Colorado, New York, Washington, Illinois) or in Ontario after January 1, 2026, the posting must include expected compensation. Missing salary in those jurisdictions is either a legal violation or a ghost. Ontario penalties run up to CAD $100,000 per violation, so employers with real roles comply.
5. Employee-count trajectory (90 seconds)
LinkedIn's company page shows headcount growth over 6 and 12 months. Companies that have shrunk 10% or more in the last year and still list dozens of "open" roles are almost always running pipeline postings or PR-driven listings. Cross-reference with any recent WARN filings.
Once a posting clears all five, tailoring is worth the hour. That is where Refolk becomes the useful tool: paste the verified posting, and Refolk writes your resume from your own history rewritten for that role, drafts the cover letter, and scores how well you actually fit. The verification decides whether to apply. Refolk decides how.
Ghost-job rate by industry, and what to do with the number
Tech at 48% and government at 60% are the two sectors that break the platform average, and knowing your sector's number is the difference between rational application volume and burnout.
| Sector | Ghost-job rate | Source | Practical implication |
|---|---|---|---|
| Government | ~60% | Columbia Law Review, Nov 2025 | Expect 45+ day cycles. Prioritize postings with named hiring manager. |
| Information / Tech | ~48% | Columbia Law Review, Nov 2025 | Recruiter-headcount check is mandatory. |
| Platform average (all sectors) | 18 to 22% | Greenhouse internal data; Glassdoor LLM analysis | Non-tech, non-gov roles are safer to apply to at volume. |
| HR-reported (any employer posts ghosts) | 93% | LiveCareer, March 2025 (n=918) | Assume every company posts at least some ghost roles. |
The reconciliation matters. When you read "only 20% of jobs are ghosts" on a platform blog, that is the mean across all sectors. If you are an engineer or a policy analyst, your personal ghost rate is closer to 50%. Weight your effort accordingly.
At ~117 tech-adjacent candidates per government recruiter in Refolk's index, government roles will look open for months without moving. This is not always ghosting in the malicious sense, it is a staffing reality. The workaround: apply only to government postings with a named hiring official, a closing date within 30 days, and a specific position control number. Everything else can wait.
Ontario's January 1, 2026 rules, and what US applicants can borrow
Ontario is the first jurisdiction with hard enforcement, and its rules give job seekers a checklist even US applicants can use as a proxy for legitimacy. As of January 1, 2026, Ontario employers must:
- Include expected compensation in every posting.
- Disclose any AI used in hiring decisions.
- Notify interviewed candidates of a decision within 45 days.
- Stop requiring "Canadian experience."
Penalties run up to CAD $100,000 per violation, which is high enough to make disclosure credible. If you are applying to a company with Ontario operations, the Canadian posting is a tell for how they run US hiring too. A company that posts a compliant, salary-disclosed Ontario role and a vague US one is signaling which market they take seriously.
California's AB 1251 (2025 to 2026 session) would go further: it would require private employers to state in every posting whether it is for an actual vacancy, with false listings treated as unfair competition and civil penalties of $100 to $10,000 per violation. Watch that bill.
Jobright.ai analysis of 4.4 million applications, 2025.
What a verified-posting workflow actually looks like
The workflow is verify, then tailor, then apply, in that order. Most job seekers reverse the first two steps and pay for it in hours.
Here is the concrete sequence:
- Screen 20 postings in 30 minutes using the 5-check list above. Kill the ghosts.
- Rank the survivors by recruiter-headcount ratio, hiring-manager activity, and pay transparency.
- Tailor the top 5, not the top 20. This is where Refolk pays for itself: paste the posting, get your resume rewritten from your own history for that specific role, cover letter drafted, and a fit score that tells you which of the five to send first.
- Track responses by ghost-signal score, not by title. Over four weeks, you will see that clean postings reply and dirty ones do not. The pattern is louder than any single application.
Some researchers, including the Columbia Law Review analysis, argue that a slice of ghost postings function as resume harvesting: names, emails, and work histories collected by a listing that was never real, with possible FTC enforcement implications. You cannot always tell those apart from pipeline-building postings, but the verification checklist catches both. If a company fails three of the five checks, do not hand them your work history.
FAQ
How do I spot a ghost job posting on LinkedIn specifically?
Look for four LinkedIn-specific tells: the posting has been up more than 45 days without a "reposted" label, the company's recruiter headcount on their People page is thin relative to open roles, the "Meet the team" section is empty or shows no one from the hiring function, and applicant counts climb past 200 without the posting closing. Any one of these is noise. Two or more is a ghost.
Are fake job listings illegal in the US?
Not federally, and not in most US states yet. Ontario's amended Employment Standards Act made them enforceable as of January 1, 2026, with penalties up to CAD $100,000. California's AB 1251 would do something similar, with civil penalties of $100 to $10,000 per violation. Until state laws catch up, the FTC angle around resume harvesting flagged by the Columbia Law Review is the only federal lever, and it is untested.
Should I still apply to ghost jobs just in case?
No, and the math is the reason. At a 48% ghost rate in tech and an average 9-hour application cycle, applying without verification means about 4.3 hours of every 9 goes to nothing. That is a full workday per week wasted for someone applying to five roles. Spend the same 4.3 hours verifying and tailoring five verified postings instead.
Why is the ghost-job rate 48% in some studies and 20% in others?
Because they measure different things. Greenhouse's platform data (18 to 22%) and Glassdoor's LLM analysis (up to 21%) are averages across every sector on the platform. The Columbia Law Review's 48% is specific to information and tech, and 60% is specific to government. Tech and government pull the mean up. Your personal ghost rate is the sector number, not the platform average.