Texas Just Subpoenaed LinkedIn Premium. 27% of Its Jobs Are Ghosts.

If you have paid $39.99 a month for LinkedIn Premium and watched your applications vanish into a void, a state attorney general now agrees the void might be the point. In July 2026, Texas AG Ken Paxton opened a formal investigation into LinkedIn over whether Premium subscribers were sold access to a marketplace where between one in five and one in three listings never intended to hire anyone.

This is the first serious state-level enforcement action against a job board over posting quality, and it changes how a rational job seeker should spend their week.

What Ken Paxton actually filed

Paxton's office issued a Civil Investigative Demand against LinkedIn, targeting whether Premium subscribers were misled about the authenticity of the job marketplace they were paying to access. The demand compels documents, data, and internal communications on LinkedIn's advertising, marketing, verification practices, and Premium representations.

The core numbers behind the Texas AG LinkedIn investigation:

• LinkedIn: over one billion registered users worldwide, roughly $17.8 billion in FY2025 revenue from Premium, recruiter licenses, and advertising.
• Premium Career: $39.99/month, or $479.88 annualized.
• Premium Business: $69.99/month, or $839.88 annualized.
• LinkedIn's public response: policies require postings to be "authentic and accurately represented," and many listings display company response time and whether candidates are being reviewed.

Paxton's theory, read between the lines, is a classic deceptive-trade case. Premium sells applicant insights and "top applicant" positioning. If a large share of the underlying inventory was never going to close, the InMail credits and priority placement were spent on listings that could not convert regardless of how strong the applicant was.

The 27% number, and why it is not the whole story

The headline figure - 27.4% of U.S. LinkedIn listings behaving like ghost jobs - comes from ResumeUp.AI's analysis of postings that stayed live past 30 days across 30, 60, 90, 120, and 233-day windows since January 2025. It is a staleness proxy, not a fraud measurement. That distinction is exactly what the Texas AG's case will hinge on.

The honest range across every serious dataset for ghost job statistics 2026 sits between 18% and 33%, and the methodology explains the spread.

Three different questions are being asked. Ashby's 18% counts ATS outcomes: did this req close? Clarify Capital's 33% asks employers to self-report intent. JOLTS-derived rates measure openings that never produced a hire. None of them is wrong. The true rate for any given sector likely lives in the 20% to 35% band.

The city breakdown nobody quotes

ResumeUp.AI's city data is where the pattern gets uncomfortable. Los Angeles leads at 30.5%, followed by Philadelphia at 30.1%, Indianapolis at 27.8%, New York at 26.7% (on 23,000 postings, the highest raw volume), and San Francisco at 26.0%. Seattle is the cleanest large market at 16.6%. If you live in LA or Philly, roughly one in three of the jobs you scrolled past this morning was already stale.

Why the ghost rate stopped being cyclical in 2021

Ghost postings became a structural feature of the U.S. labor market five years ago, not a temporary distortion. Openings and hires tracked together for a decade after the Great Recession. Then in 2021 openings spiked toward 11 million while hires stayed at 6 to 7 million, and the 28% to 32% gap has held ever since.

The mechanism is the recruiter workload math, and it is where Refolk's index becomes uncomfortable reading for LinkedIn.

• U.S. professionals currently holding a Recruiter or Talent Acquisition title in Refolk's index: ~91,748
• Of those, professionals who explicitly surface Texas in their profile keywords: ~258
• June 2025 hires reported by BLS: 5.2 million
• Implied load if every recruiter had to close every hire: ~57 hires per recruiter per year

No recruiter closes 57 hires a year at any level of seniority worth naming. The math only works if a meaningful share of postings were never meant to close: pipeline building, budget signaling, competitive intelligence, always-on listings from staffing firms like Robert Half and RCM Health Care Services. Ghost postings are not a bug. They are what makes the recruiter workload survivable.

Ghost postings are not a bug in the marketplace. They are what makes the recruiter workload survivable.

Is LinkedIn Premium worth it in this market?

For most job seekers in 2026, the answer is no, and the reasoning has changed. Two years ago the argument against Premium was soft: InMail response rates are mediocre. Now the argument is quantitative: you are paying for priority access to inventory where roughly one in four listings will never hire, and a state AG is actively investigating whether that constitutes deceptive trade.

Run the math on your own applications. Jobright.ai's analysis of 4.4 million applications puts the average application cycle at 9 hours of research, tailoring, follow-up, and waiting. If you send 100 applications a year, a 27.4% ghost rate means you are burning roughly 247 hours on postings that were never going to close. That is six full working weeks.

Premium's value proposition depends on three assumptions:

1. The underlying inventory is real.
2. "Top applicant" placement affects hiring outcomes.
3. Applicant insights, including how many applied and how you rank, reflect real competition.

The Paxton investigation puts all three under pressure. If 27% of listings behave like ghosts, the "top applicant" badge is doing work on postings where nobody ever wins. That is the deceptive-trade theory Paxton is likely testing, and it is why the annualized $479.88 to $839.88 spend deserves a hard look before you renew.

The better use of that budget is time, not tools. If the marginal cost of an application dropped low enough that you could apply to five real openings in the time it now takes to tailor one, the ghost rate stops mattering. That is the specific friction Refolk removes: paste the posting, get your resume rewritten for it, cover letter drafted, and a fit score that tells you whether the role is actually worth your nine hours before you spend them.

How to spot fake job postings before you apply

Check the age, the pattern, and the poster. Every reliable signal for how to spot fake job postings is public information that takes less than 60 seconds to verify.

• Age check. Anything past 30 days on LinkedIn is in the ResumeUp.AI ghost band. Past 60 days, treat it as a lead-gen exercise unless the company is Fortune 500 with a known slow cycle.
• Repost pattern. Search the exact title plus the company name. If the same req has been closed and reposted three times in six months without a hire announcement, it is a pipeline listing.
• Poster identity. A req posted by "LinkedIn" or an unnamed recruiter is weaker signal than one posted by a named person whose profile shows recent hires in that function.
• Staffing firm concentration. One 2025 analysis found a handful of staffing firms were responsible for over 1.2 million suspicious postings across major job boards. If the poster is a staffing agency with thousands of open reqs, discount aggressively.
• Response-time badge. LinkedIn now displays company response time on many listings. "Actively reviewing applicants in the past week" is real signal. Absence of that badge on a 45-day-old posting is not.
• Salary band presence. In states without pay-transparency laws, a listing with a specific band is more likely to be a real hiring req than a vague "competitive" line.
• Headcount range. LiveCareer's data pins the worst offenders at 1,001 to 5,000 employees, around 25% ghost rate, particularly in tech, publishing, and software. Sub-200-person companies and Fortune 100s both post cleaner listings, for opposite reasons.

The other 67% problem, per a 2025 analysis: 67% of applications receive no response at all, not even an automated rejection. Silence is not evidence of a ghost job on its own. But silence plus a stale posting plus a staffing-firm poster is a strong three-signal stack. Skip it.

Where to spend the hours you get back

Reallocate application effort toward postings with response-time badges, named human posters, and companies actively announcing hires in the function you target. The federal Truth in Job Advertising and Accountability Act, currently in play in 2025, would eventually force employers to disclose intent to hire, salary bands, and post-removal deadlines under DOL and FTC enforcement. Until it passes, the burden of filtering is on you.

Three moves that compound:

1. Cut your application count in half, double your tailoring depth. The ghost rate makes volume strategies economically worse each year. A 40-application month of tailored, response-time-verified submissions beats 100 spray-and-pray applications.
2. Move the "am I a fit?" question upstream. Deciding after you have already written the cover letter is the expensive way. Refolk scores how well your history actually matches the posting before you draft anything, so the nine-hour cycle collapses to about twenty minutes on the roles worth pursuing.
3. Post where recruiters actually work. Refolk's index shows ~91,748 U.S. recruiters and TA professionals, concentrated at firms like HCA, Robert Half, and RCM Health Care Services in healthcare, and dispersed across mid-market tech elsewhere. Being findable to that population is more efficient than out-applying the ghost rate on LinkedIn.

The Texas investigation may or may not produce a settlement. Either way, the number that matters to you is not the AG's headline. It is the 247 hours a year you are currently trading for outcomes the marketplace was never going to deliver. Cut the ghost applications, and the search gets shorter whether Paxton wins or not.

FAQ

Is LinkedIn Premium worth $39.99 a month right now?

For most active job seekers, no, and the calculation got worse in July 2026. You are paying an annualized $479.88 (Career) or $839.88 (Business) for priority access to a marketplace where roughly 27% of listings behave like ghost jobs and a state AG is investigating whether that constitutes deceptive trade. If you can identify three specific InMails per month you would not send without Premium, and those recipients are hiring managers at companies with recent hire announcements, keep it. Otherwise cancel and redirect the budget toward tools that reduce your per-application time cost.

What does "ghost job" actually mean, legally and practically?

A ghost job is a posting that is live and accepting applications but is not actively being filled. Practically, that covers reqs kept open for pipeline-building, budget-approved roles that got frozen, competitive intelligence postings, and staffing-firm listings that exist to collect resumes. Legally, there is no federal definition yet; the proposed 2025 TJAAA would create one by requiring employers to disclose hiring intent and post-removal deadlines. Paxton's Texas case will test whether existing state deceptive-trade statutes already cover the practice.

How do I tell if a specific LinkedIn posting is a ghost job?

Check three things in under a minute: posting age (over 30 days is the ResumeUp.AI ghost band), poster identity (a named recruiter with recent hires in that function is stronger signal than an unnamed account), and repost history (search the exact title plus company). A response-time badge showing recent applicant review is the single strongest positive signal LinkedIn currently exposes. Absence of that badge on a stale posting from a staffing agency is a three-signal skip.

What should I do differently if the ghost rate is really 27%?

Two changes. First, stop optimizing for application volume; the ghost rate makes spray-and-pray economically worse each year. Second, move the fit decision upstream, before you spend nine hours per application. The point is not to apply to fewer real jobs. It is to stop spending real hours on postings that were never going to close. Cut ghosts out of the funnel and a 100-application year becomes a 73-application year with the same number of real shots at a real offer.