Job Listing Quality Statistics 2026: We Scored 1,000 Listings

The consensus on LinkedIn right now is that the job market is effectively broken. You see the same posts every day: highly qualified engineers shipping 400 resumes into the void and getting zero interviews.

People are convinced companies are posting what some call "ghost jobs" — roles that look active but have zero budget or hiring intent attached to them.

But complaining on social media doesn't solve anything. We wanted the actual data on listing quality. So we built a scoring engine that checks 51 quality signals per listing, and we validated it in our V3 Tuning Validation Study.

We extracted a stratified sample of 1,000 real job listings pulling from major Applicant Tracking Systems and ran them through three independent models:

1. Subspace V3 Engine: Our 51-signal deterministic engine.
2. Llama 3.3 70B: Meta's 70-billion parameter frontier model.
3. Gemini 2.5 Flash: Google's efficient multi-modal LLM.

Here's the math on what's actually happening.

Proving the Engine Works

If we're going to claim a percentage of listings are suspect, the model has to be solid.

To validate our engine, we ran a bake-off. We took the 1,000 jobs and had both Llama 3.3 70B and Gemini 2.5 Flash analyze the raw context and render a verdict.

Our automated, multi-signal deterministic engine maintained a 0.904 correlation with Llama 3.3 70B, and a 0.877 correlation with Gemini. The models agreed on the risk level (within 30 percentage points) 98.2% of the time.

It works. And it does it at a fraction of the cost of running a heavy LLM.

The Scale of the Problem

We didn't just build a naive scraper that looks at "Days Since Posted." Companies game that routinely. We built an engine that evaluates dozens of behavioral markers per listing — structural metadata, cross-platform persistence, and posting patterns.

~32% of all active listings we scanned flagged as Low Quality (60%+ concern score).

Based on what we've seen, that means roughly 1 in every 3 jobs you come across may not represent a genuine, actively-filled opening.

Why Do Low-Quality Listings Exist?

When we parsed the data by ATS source, we found that low-quality listings aren't evenly distributed. They cluster around specific company behaviors and platforms:

1. The Pipeline Hoarders: All three models agreed that certain recruiting platforms inherently carry higher risk. Listings from ATS vendors heavily favored by high-growth startups averaged a 67-72% concern score, significantly higher than enterprise-focused software. Why? Because startup-focused platforms make it incredibly easy to maintain "always open" pipeline requisitions to harvest resumes, even during hiring freezes.
2. The "Look at Our Growth" Mirage: Older listings are overwhelmingly low quality. On fresh listings (0-7 days), the average concern score is around 39%. By the time a listing hits 90+ days, it jumps to 67-73%. Companies leave these up to look like they have momentum.
3. The Salary Omission Trap: Only 30% of the listings we analyzed actually posted a salary. The remaining 70% were heavily penalized by both Subspace and Llama. Salary omission strongly correlates with organizational opacity, which correlates directly with low listing quality.

Stop Applying to Low-Quality Listings

The high-volume "spray and pray" approach doesn't hold up when a third of the board is noise. A quality filter helps.

1. Stop applying to jobs that have been up for 45+ days.
2. If the company is constantly raising capital but the job description lacks specific project details, be highly skeptical.
3. If the role has no salary band (especially in a state that requires it), treat it as an unfunded wishlist until proven otherwise.

That said, doing this manually for every application isn't realistic. I took the engine from this research and shipped it as a free tool — drop any URL into thesubspace.io and it runs the same 51-signal check in about three seconds.

— Vivek