In July 2025, LinkedIn quietly disclosed a number that should have made everyone rethink their job search: 11,000 applications per minute, up 45% year-on-year (eWeek).
Katie Tanner, an HR consultant, posted a single remote role and got 1,200+ applications within days. She pulled the listing and was still sorting submissions three months later.
This is not a hiring market. It's a queue nobody can read.
If you've been feeling like your applications are disappearing into a void, they are. The gap between "I submitted 300 applications" and "I got three interviews" isn't a motivation problem. The funnel is structurally broken. Applying harder rewards the wrong behaviour.
The application tsunami
The numbers compound. LinkedIn's 45% YoY surge sits on top of a 31% rise in H1 2024. Recruiters now handle 2.7× more applications and 56% more open requisitions than they did three years ago (HyperApply). The average LinkedIn Easy Apply CV gets 8.4 seconds of attention. That's not a typo.
The aggregate effect is brutal. Overall job-application response rates have fallen roughly 3× since 2021 (uppl.ai). In 2025, candidates averaged 42 applications per interview, a conversion rate of about 2.4% (The Interview Guys).
The standard advice ("just apply to more roles") is now counterproductive. In a 2025 meta-analysis of 27 hiring studies, candidates who applied to 21–80 roles had a 30.9% offer rate. Those who applied to 81 or more had 20.4%. Past a point, volume hurts outcomes.. time spent blasting is time not spent on the things that actually work.
There's a subtler problem with the LinkedIn-or-Indeed playbook. 57% of jobs marked "expired" on LinkedIn and 50% on Indeed are still open on the employer's career site. Aggregators decay in both directions. They surface stale postings and hide live ones. You're not even seeing the real market.
The ghost-job problem
On top of the volume crisis sits a signal crisis. A big share of the postings you're applying to don't correspond to real hiring intent.
- Hunter Ng's 2024 arXiv paper analysed Glassdoor postings with LLM classification and found up to 21% were ghost jobs (arXiv:2410.21771).
- Revelio Labs tracked hires-per-posting from 0.75 in 2018 to under 0.5 in 2023. Half of postings now never result in a hire (Revelio).
- Clarify Capital surveyed 1,000 employers in 2025 and found nearly one in three admitted to posting roles they did not intend to fill. Around 1-in-5 keep roles "open" for optics. Nearly 1-in-5 use AI to auto-repost inactive listings (Clarify Capital).
- Greenhouse's own 2024 State of Job Hunting report disclosed that 18–22% of postings each quarter are ghost jobs, and 70% of Greenhouse clients posted at least one ghost job in Q2 2024 (Greenhouse). That's on Greenhouse, right? A paid, reputable ATS. The noise is inside the clean channels too.
The uncomfortable implication: going direct to the employer's ATS isn't enough on its own. You still need filtering.
Ghost postings now distort macroeconomic indicators. The Beveridge curve, which economists use to relate vacancies to unemployment, has been visibly shifted by fake listings since 2021. Bad data at the posting level is deforming the models governments use to read the labour market.
Bots applying to bots
Humans are mostly not in the first screening loop anymore. On either side.
On the candidate side, Gartner's Q4 2024 survey of 3,290 applicants found 39% used AI to write or tailor their applications. Other 2025 surveys put the number at 65% in some segments (CNBC). 64% of recruiters now report a surge in lookalike AI-generated CVs. 73% have seen AI-written applications containing false information (HR Dive). Gartner projects that by 2028, one in four candidate profiles will be fake.
That's not hypothetical. In May 2024, the US DOJ charged a ring that had placed North Korean operatives as remote IT workers at 300+ US companies, generating $6.8M in revenue for sanctioned entities. Every one of those hires cleared somebody's screening process.
On the recruiter side, 72% of HR teams use AI in hiring, up from 58% in 2024 (HireVue 2025 AI Report). HireVue alone has run 70 million+ AI-scored video interviews and services 60%+ of the Fortune 100. AI screening tools classify a CV in roughly 0.3 seconds.
So the first human who sees your application is often the third or fourth layer down, after an auto-screen, a knockout-question filter, and a scoring pass. Hiring in 2026 is bots negotiating with bots. Humans arbitrate the survivors.
A meme worth killing: "75% of CVs are auto-rejected by ATS formatting" is nonsense. The number traces to a 2012 marketing claim from Preptel, a now-defunct resume-optimisation vendor. A 2025 recruiter survey found 92% of ATS systems do not auto-reject for formatting (The Interview Guys). What rejects you is a knockout question, an LLM screen, or a human giving you 8 seconds because they have 500 others in the queue. Stop wasting hours keyword-stuffing.
What the data says works in 2026
Honestly, job-hunting isn't hopeless. The gradient has moved.
1. Go direct to company ATS pages
The response-rate gap between channels is enormous.
- Company career pages and direct-to-ATS applications: ~11%
- Google Jobs: ~11.3%
- LinkedIn: 3–4%
- Generic job boards: under 2%
This gap exists for a structural reason. Greenhouse, Lever, Workable and similar ATS boards post roles before aggregators index them, sometimes by days. Most applications arrive in the first 72 hours. Candidates hitting the posting at source are competing in a smaller, earlier pool. Job boards produce roughly 90% of applicants but only ~38% of hires (Pinpoint). At source, the ratio flips.
2. Warm intros, seriously
A single referral is worth roughly 40 cold applications. Referred candidates are 4–5× more likely to be hired. Recruiter-sourced candidates are 8× more likely than cold applicants. Referrals make up 2% of the applicant pool but 11–30%+ of hires (Pinpoint; The Interview Guys).
If you have ten hours a week to spend on your search, seven of those hours are better spent messaging past colleagues, asking for intros, and writing thoughtful DMs to two or three hiring managers than on LinkedIn Easy Apply. It doesn't feel like job-searching. That's why most people skip it. It works because most people skip it.
3. Tailored beats volume, with a hard cap
Again: 21–80 targeted applications → 30.9% offer rate. 81+ → 20.4%. Pick a smaller set of roles. Write a specific cover letter. Reference something real about the company. Set a ceiling, not a floor.
4. Show evidence
CodePath's December 2025 employer-readiness data found 38% of tech employers rank portfolio work and side projects as the top signal, and 34% rank public GitHub. Both rank above degree prestige (Fortune). In a world where AI can generate a plausible CV in thirty seconds, a live artefact you made (that compiles, ships, has users) is worth more than a line of text claiming you can make it.
The awkward conclusion
You need to read more postings than ever, because the good ones live on company ATS boards no aggregator fully indexes. And you need to apply to fewer, because blasting past 80 tanks your odds.
Covering 100+ independent Greenhouse, Lever, Workable and Reed boards by hand. Filtering out the 20% that are ghost postings. Figuring out which of the remaining 1,000 listings that week are actually for you. That's a filtering problem. Most candidates never solve it. They default to LinkedIn because it's the one place that puts a million jobs in one interface, and they pay for that convenience with a 3% response rate.
This is roughly the problem Flint was built to solve. We pull directly from 100+ company ATS boards (Greenhouse, Lever, Workable, Remote OK, Adzuna, Reed, more), score every role against your CV across six dimensions (title, skills, salary, location, seniority, industry), and surface the handful that are actually worth your time.
Read wider. Apply narrower. Spend the hours you save on warm intros.
You can drop your CV into Flint free, in about 60 seconds.