How Credibility Scoring Works: Inside JobIntel's Intelligence Engine

Not all job listings deserve your time. That sentence has been the through line of everything I have written since launching this blog. Ghost jobs waste your applications. Duplicate listings inflate the market. Stale postings lead to silence. The question was always: how do you tell the difference before you invest?

Credibility scoring is the answer I built. Every listing that passes through JobIntel receives a score from 0 to 100 - a probabilistic assessment of whether that posting represents a real, active, fillable role. This post explains exactly how it works, what signals drive the score, and where the system falls short.

I am publishing the methodology because transparency is the point. The methodology borrows from the same enterprise evaluation frameworks I outlined in my books - the same structured approach to separating signal from noise that enterprise teams use for vendor selection and technology adoption, applied to job listings. The job market is full of black boxes - ATS systems that filter without explanation, AI screening tools with documented bias, JOLTS numbers inflated by duplicates. Job listing credibility scoring should not be another one.

The problem credibility scoring solves

The numbers tell the story. 18-27% of online job listings are ghost jobs, according to converging studies from Greenhouse, ResumeBuilder, and Clarify Capital. 40% of companies posted at least one fake listing in the past year (ResumeBuilder survey of 1,600 hiring managers). JOLTS job openings fell to 6.542 million in December 2025, and even that number overstates reality because JOLTS does not deduplicate across job boards.

Meanwhile, 72% of job seekers say the process has harmed their mental health (Resume Genius). The typical path to an offer requires 100-200+ applications. At that volume, every application sent to a ghost listing or an expired posting is time and emotional energy spent on nothing.

The ghost job problem I first documented and the way duplicate listings inflate the market are not abstract problems. They are the reason your search feels harder than the headline numbers suggest.

The signals: what feeds the score

Credibility scoring evaluates each listing across multiple signal categories. No single signal is definitive - the score reflects the weight of evidence across all inputs.

Temporal signals

Listing age is the strongest single predictor. According to LinkedIn data, candidates who apply within the first week of posting are 4x more likely to receive a response. Response rates decline significantly after the first two weeks. Industry benchmarks suggest typical hiring timelines of 30-45 days for tech, 20-30 days for healthcare, 45-60 days for finance, and 60-90+ days for government.

A listing open for 60+ days without updates is well past expected hiring timelines for most industries. The listing age analysis I published showed how "days on market" - a concept borrowed from real estate - reveals which listings are genuinely active.

Update frequency matters. A listing that has been refreshed, edited, or reposted recently shows active management. A listing unchanged for months suggests it has been forgotten or was never meant to fill.

Content signals

Description specificity separates genuine hiring intent from placeholder postings. Real roles have specific technical requirements, named frameworks, defined team structures, and clear responsibilities. Ghost postings use evergreen language: "fast-paced environment," "self-starter," "wear many hats." The less specific the description, the lower the credibility signal.

Salary transparency correlates with hiring intent. Roughly 60% of postings on Indeed now include salary information, up from 18% in 2020. In states with salary transparency laws, the absence of a salary range is not just a red flag for compensation - it is a credibility signal. Companies investing in transparent, compliant postings are more likely to be actively hiring.

Requirements realism. A listing demanding 10 years of experience with a technology that has existed for three years is not a real hiring profile. Unrealistic requirements correlate with ghost postings designed to "prove" no qualified candidates exist.

Source signals

Cross-platform presence. A listing that appears on the company's own careers page and on aggregator boards is more credible than one that lives only on third-party sites. The duplicate detection system identifies when a role appears across multiple platforms - the pattern of where it appears tells a story.

Direct employer versus staffing agency. Staffing agency listings have different credibility profiles than direct postings. Neither is inherently better or worse, but the signals differ and the scoring adjusts accordingly. A staffing agency posting the same role across five boards is normal business practice - a direct employer doing the same thing may signal desperation or a ghost posting designed for maximum visibility without hiring intent.

Company posting history. A company that regularly posts, fills, and closes listings demonstrates genuine hiring activity. A company with dozens of open positions that never close - month after month - shows a pattern consistent with pipeline building rather than active recruitment.

Scoring logic

Signals are combined into a composite score where stronger predictive signals carry more weight. Listing age and description specificity are weighted most heavily based on their correlation with genuine hiring outcomes. Salary transparency and source verification provide supporting evidence. The result is a 0-100 score distributed across three tiers:

High credibility (80-100): likely real, active roles with specific descriptions, reasonable salary ranges, and verified employer activity
Medium credibility (50-79): mixed signals; may be genuine but shows some risk factors
Low credibility (0-49): multiple ghost indicators present; proceed with caution

What the scoring gets wrong

I would not trust a methodology that claims perfection. Here is where credibility scoring has known limitations.

Government and academic roles. Government hiring timelines of 60-90+ days mean a legitimate federal position can look stale by private-sector standards. The scoring adjusts for sector, but the adjustment is imperfect. A 90-day-old government listing is not equivalent to a 90-day-old tech listing.

Internal postings made public. Some companies post roles externally as a compliance requirement when they have already identified an internal candidate. These postings are technically real but functionally unavailable. No scoring system can detect internal pre-selection.

New companies. A startup with no posting history has less signal for the algorithm to evaluate. Absence of evidence is not evidence of absence - but it does mean lower confidence.

The probabilistic caveat. A credibility score of 35 does not mean the listing is definitely a ghost. It means the listing displays multiple characteristics consistent with ghost job patterns. Some low-scoring listings are real. Some high-scoring listings are not. The score improves your odds of investing time wisely - it does not guarantee outcomes.

What this means for your search

The practical framework is straightforward.

High-credibility listings (80-100) deserve investment. Tailor your application. Write the cover letter. Research the company. These listings show the strongest signals of genuine hiring intent. The time cost analysis showed that targeted applications to high-quality listings produce dramatically better results than volume applications to everything.

Medium-credibility listings (50-79) deserve evaluation. Check the company's careers page directly. Look at listing age. Evaluate whether the mixed signals have an explanation. Apply if the role fits, but do not invest maximum effort.

Low-credibility listings (0-49) deserve skepticism. Not all are ghosts - some are legitimate roles with poor posting practices. But the signal-to-noise ratio is unfavorable. If you are sending 100+ applications and half are going to low-credibility listings, you are spending half your effort on the worst odds.

This is the intelligence layer I built JobIntel to provide. Every listing scored, the methodology disclosed, and the limitations acknowledged. The goal is not to replace your judgment - it is to give you data for that judgment.

Try JobIntel free at jobintel.com. See which listings are real before you apply.