LazyApply's 23% Ban Rate: The Auto-Apply Signals Recruiters Blacklist

Auto-apply used to be a recruiter annoyance. As of mid-2026 it is a career liability: LinkedIn's new velocity detection restricts roughly one in four bulk-automation users inside 90 days, and a recruiter-maintained blacklist of 461 plugins now names LazyApply directly. If you are running a bot on your own LinkedIn account, the ban and reputational risk transfers to you, not the vendor.

The 23% number, and why it is 57x the baseline

LinkedIn now restricts about 23% of bulk-automation users within 90 days, compared to a category baseline near 0.4%. That is a 57.5x relative risk multiplier for the same account behavior that six months ago went untouched.

The mechanism is architectural. LazyApply and most competitors ship as a Chrome extension that runs on your own machine, over your own residential IP, using your own logged-in session. There is no IP rotation, no session throttling, no browser fingerprint randomization. When LinkedIn's classifier flags a session as non-human, it does not flag the vendor. It flags you.

Enforcement under LinkedIn's User Agreement §8.2 is tiered:

• First offense: a 24 to 48 hour restriction and a warning.
• Repeat offense: permanent account deletion.
• Appeal path: fewer than 10% of appeals succeed.

A permanently deleted LinkedIn profile is not a minor inconvenience in a market where the average job search now takes 55 to 71 days and the average posting draws roughly 242 applicants. It is a hole in your work history that recruiters, sourcers, and your own network use to find you.

What "human-impossible velocity" actually detects

Human-impossible velocity is LinkedIn's 2026 label for a session pattern no real person could produce. It is a fingerprint, not a single threshold, and slowing your bot down does not defeat it.

The documented trigger cited most often is 100+ applications per hour. LazyApply's Ultimate tier advertises up to 1,500 applications per day, which over an eight-hour run averages 187 per hour, roughly 1.9x the trigger. But the classifier also weighs signals that have nothing to do with hourly count:

• Clicks and page transitions faster than any human could read the page.
• Sessions that never time out, never idle, never scroll erratically.
• Bursts at 3am local time, or perfect regularity across a 12-hour window.
• Identical dwell time on every Easy Apply modal.
• Screener answers pasted with zero keystroke variance.
• The exact same resume file hash uploaded to hundreds of postings in a row.

Any of those signals, on their own, will not get you restricted. Stacked together for 90 days, they will. This is why "just apply slower" is not a fix. The pattern has to look human end to end, which is roughly incompatible with a Chrome extension optimizing for throughput.

The 461-plugin recruiter blacklist

Josef Kadlec's Complete List of Blacklisted LinkedIn Plugins Vol. 3 catalogs 461 flagged tools, LazyApply among them, and it functions as de-facto shared intelligence across recruiter networks.

The list is not an official LinkedIn document. Kadlec is a former ethical hacker and digital forensic examiner who co-founded GoodCall, a recruiting agency with 140+ employees, and he speaks at industry events. His list has grown from 83 plugins to 461, roughly a 5.5x expansion in the surveillance surface recruiters actively watch.

That matters because the audience for the list is large. In Refolk's index of professional profiles, about 90,720 people in the United States currently hold "Recruiter" in their title, and about 6,551 do in the United Kingdom, a 13.8x US-to-UK ratio. These are the people who share screenshots in private Slacks, tag repeat submitters in ATS notes, and remember names.

The five signals that get an application flagged

Recruiters and ATS platforms flag auto-apply submissions on five behavioral fingerprints. Each one is visible without any algorithm running.

1. Duplicate CVs across a shared ATS

Modern ATS platforms including Greenhouse, Workday, Lever, and Ashby show recruiters when a candidate has already submitted to the same company, and increasingly across companies on the same platform. A candidate who fires 12 identical CVs at one employer in a week gets a permanent tag on their name in recruiter memory. No LinkedIn ban required. This is the sleeper risk that most auto-apply reviews miss.

2. Identical screener answers

Every posting asks "Why this role?" or "Describe a relevant project." Bots paste the same 400-character block into all of them. Recruiters see the same answer twice and stop reading. This is the exact friction Refolk is built to remove: paste the posting, get your own resume rewritten against it, get a cover letter drafted from your real history, and get scored on how well you actually fit before you hit submit.

3. Applications to stale or already-filled roles

In one recruiter test, 31 of 180 Sonara auto-submissions went to jobs that had already been filled internally, postings that were two to three weeks stale. Bots do not know a role is dead. Recruiters see the timestamp and infer you did not know either.

4. 3am submission clusters

Human job seekers apply between 9am and 11pm local time, with a lunch spike and an evening peak. Bots apply on a cron. A 3am submission burst against your local timezone is one of the cheapest signals to detect.

5. CAPTCHA-broken partial submissions

CAPTCHA walls on Indeed, ZipRecruiter, and Glassdoor now regularly break LazyApply mid-run, leaving half-submitted applications and error states that recruiters see as ghosted or incomplete. A trail of broken submissions across three job boards in one afternoon is its own fingerprint.

Why volume does not buy you interviews

Auto-apply saves time, not interviews per attempt. LazyApply-style volume produces roughly the same 0.4% interview rate as the market baseline, which means the marginal 900th application in a day converts at the same rate as the first.

That is a hard result to internalize when a tool promises 1,500 submissions per day. The pitch sounds like leverage. The math is not. If your baseline conversion is 0.4%, running 1,500 applications yields six interviews on paper, and the LazyApply Trustpilot rating (~2.1 to 2.4/5, with 56% one-star reviews per a March 2026 analysis) is populated by users who did not see those six materialize because their CV was generic, the postings were stale, or their account was restricted before the numbers landed.

The interview-rate ceiling moves when the resume and cover letter change per posting, when the fit is scored before submission, and when the recruiter reads something that sounds like it was written for them. That work is slow by hand, which is the whole reason people reach for bots.

How to move fast without producing bot signals

You can send a lot of applications without triggering velocity detection or landing on a recruiter blacklist. The rule is that every application has to look, to both the classifier and the human reader, like a real person made a decision.

Practical constraints that keep you inside the human envelope:

• Cap yourself at 15 to 25 applications per day, sent across a normal waking window.
• Never re-upload the same PDF file hash. Regenerate the resume per posting, even if the changes are small.
• Write, or have written, a specific first sentence per cover letter that names the team or product.
• Answer screener questions in your own voice, and vary them across postings.
• Skip postings older than 21 days unless the company has a known slow pipeline.
• Never run a browser automation extension logged into your primary LinkedIn.

The generation step is the bottleneck. A tailored resume plus a real cover letter used to take 20 to 40 minutes per application, which is why people reached for bots. Refolk collapses that into paste-the-posting-get-the-file, and returns a fit score so you can drop the ones you would not clear anyway. You end up sending fewer applications at a higher conversion rate, on your own IP, with no fingerprint a classifier can chain into "human-impossible velocity."

What to do if you are already restricted

If LinkedIn has already restricted your account, the playbook is short and time-sensitive. Fewer than 10% of appeals succeed, so the odds favor prevention, but there are still moves that help.

1. Uninstall every browser extension that touches LinkedIn, not just the auto-apply tool. Screenshot the extension list before you do.
2. Change your password and sign out of all sessions.
3. File the appeal within 24 hours, in your own words, and do not mention the tool by name.
4. Assume the appeal fails. Start building a parallel presence: a personal site with your work history, a GitHub or portfolio link, and outreach direct to hiring managers through email rather than the LinkedIn Easy Apply funnel.
5. If the account is permanently deleted, do not open a second account under a variant of your name. LinkedIn detects and merges duplicates, and the second deletion is faster than the first.

The reputational tail is longer than the algorithmic one. A restriction eventually expires or the account eventually rebuilds. A recruiter who tagged you as a repeat auto-submitter in an ATS note six months ago still sees that note when your name comes up in a new search. The only fix for that is time and a different pattern of applying.

FAQ

Is LazyApply actually banned by LinkedIn?

LazyApply is not banned in the sense that the vendor cannot ship the extension, but its usage pattern gets user accounts restricted at roughly 23% within 90 days versus a 0.4% baseline for non-automation users. It is also named in Josef Kadlec's Complete List of Blacklisted LinkedIn Plugins Vol. 3, alongside 460 other tools that recruiters share as informal intelligence. The tool keeps operating. The users take the hit.

Can recruiters actually see that I used an auto-apply bot?

Yes, in two ways. Algorithmically, ATS platforms like Greenhouse, Workday, Lever, and Ashby show recruiters duplicate submissions from the same candidate, sometimes across companies. Behaviorally, identical screener answers, submissions to already-filled roles, and 3am clusters are all visible to any recruiter who opens the application. They do not need a bot-detection tool to spot the pattern.

Will slowing down my bot to under 100 applications per hour make it safe?

No. LinkedIn's "human-impossible velocity" detection is a fingerprint, not a single threshold. It weighs click speed, session length, scroll patterns, time-of-day distribution, and dwell time on modals. A bot capped at 50 applications per hour that still clicks faster than a human can read and runs at 3am local time will still get flagged. The pattern, not the count, is what fails.

What is a safer way to apply to a lot of jobs quickly?

Cap daily volume in the 15 to 25 range, generate a genuinely tailored resume and cover letter for each posting, answer screeners in your own words, and skip stale listings. The bottleneck used to be the tailoring, which is why people reached for bots in the first place. Refolk does that step per posting from your own history and scores the fit before you submit, so you spend your daily budget on applications that could plausibly convert instead of on volume that lands you on a blacklist.