The short answer: roughly 100 invitations per rolling week on a free account, up to about 200 on Recruiter, with a practical daily ceiling between 15 and 50 depending on how many connections you already have. There is also a second cap almost nobody plans for: LinkedIn starts blocking new invitations once you have around 400 pending, no matter how far under the weekly limit you are.
LinkedIn publishes none of these numbers. What follows is what we enforce at CampaignStack across every connected account, why each number is what it is, and where the real bans come from (spoiler: it is usually pacing, not volume).
The weekly cap: 100 to 200, rolling
Since LinkedIn tightened invitations in 2021, the weekly cap has been the binding constraint:
| Account tier | Weekly invitations (approx.) |
|---|---|
| Free | 100 |
| Premium | 150 |
| Sales Navigator | 200 |
| Recruiter | 200 |
Two details matter more than the headline number:
- It is a rolling 7-day window, not a Monday reset. What you sent six days ago still counts against today. Tools that "reset" on a calendar week routinely push accounts over.
- Paid tiers buy tolerance, not immunity. A Sales Navigator seat roughly doubles the free cap. It does not exempt you from restriction if your pattern looks automated.
The daily number: about 3% of your network, between 15 and 50
Spreading 100 invitations over a week gives you ~14 per day, but a flat daily number is the wrong model. LinkedIn's tolerance scales with account maturity: what looks normal for someone with 4,000 connections looks aggressive for someone with 400.
The formula we enforce, adapted from years of community measurement (Dux-Soup published the original version):
daily limit = clamp(15, connections x 3%, 50) x tier multiplierAn account with 500 connections gets 15/day. One with 1,500 gets 45. Nobody gets more than 50/day as a base, because the weekly cap makes higher daily numbers pointless anyway. Paid tiers scale that base: 1.5x on Premium, 2x on Sales Navigator and Recruiter.
Two things then eat into that number before anything gets sent.
Scheduled work never gets the whole budget. Automations that run on a timer are capped at 80% of the day's write allowance. The last 20% is held back for the actions that fire in response to something real: a reply that needs answering, an invitation that just got accepted. An account that spends its entire budget on scheduled sends has nothing left for the conversations those sends started, which is both bad outreach and a worse pattern.
What you did outside the tool counts too. Every morning we measure what the account actually did on LinkedIn the previous day and split it into what came through us and what did not. Invitations you sent by hand, or through another tool still connected to the same account, get subtracted from the ceiling. If you sent 20 invitations manually yesterday, today's automated allowance drops by 20. LinkedIn counts the account, not the tool, so a budget that only knows about its own sends is not a budget.
The cap nobody budgets for: 400 pending invitations
Invitation acceptance rates in cold outreach commonly run 20-40%. That means 60-80% of what you send piles up as pending. Send 100/week at a 30% acceptance rate and you accumulate ~70 pending invitations per week; in six weeks you are past 400 and LinkedIn quietly stops letting you invite anyone.
The fix is withdrawal hygiene: regularly withdrawing your oldest sent invitations. Two rules keep it safe:
- Never withdraw anything younger than 7 days. Recent invitations still get accepted; a two-week-old one almost never does.
- Withdrawals should not eat your invitation budget. Withdrawing relieves the exact pressure the budget protects against. (CampaignStack budgets withdrawals separately, at 1.2x the connection-request limit, and sweeps daily per account.)
New and small accounts: the rules change
Two situations where the numbers above do not apply:
Under 300 connections: send nothing. Small accounts are the first ones LinkedIn restricts, because a small network sending volume invitations is the canonical spam signature. Below 300 connections we set every write budget to zero and halve the read budgets, then let the account grow on its own. The same block applies to an account we cannot read a connection count for that was linked less than 30 days ago, because an unknown number on a new account is far more likely to be small than large.
New to automation: start at 10%. An account that has never automated starts at 10% of its target budget and adds 4 percentage points for every day it actually sends. That is 23 days to full volume, call it three weeks, not the "ready by Friday" most tools imply.
The word "active" is doing real work in that sentence. Idle days do not pause the ramp, they reverse it: after a one-day grace we multiply by 0.82 for each additional silent day, so a week off costs more than half the progress. This is also why we keep sending through weekends at 30% of the weekday rate instead of stopping. A quiet Saturday and Sunday back to back reads as two idle days and starts the decay, which is a strange way to punish an account for looking like a person who takes weekends off. Reduced volume keeps the ramp climbing; zero volume does not.
Volume is half the story. Pacing is the other half.
Most restrictions we see are not volume violations. They are pattern violations:
- Bursts. Fifteen invitations in fifteen minutes reads as a script even though fifteen per day is fine. Spread sends across business hours (8 AM to 6 PM in the account's local timezone) with irregular gaps.
- Weekend volume. Full-speed Saturday outreach from a "sales professional" account is a tell. We cut weekend writes to 30% of the weekday number and reads to 70%. Writes get the harder cut because a Saturday invitation is strange in a way that a Saturday profile view is not.
- Same-second scheduling. If a tool dispatches every queued invitation at the same minute daily, the timing itself becomes the fingerprint. This exact pattern (a day's requests landing in one 15-minute window at the same time every day) is what triggers security checkpoints.
What happens if you trip a limit
LinkedIn escalates in stages rather than banning on the first offence, and the second stage is the one that surprises people: an identity verification lock that asks for a government ID and can only be cleared by the account owner. We wrote that ladder up in full, including what each stage looks like from the inside and how to tell a restriction from an ordinary glitch, in Will this get my LinkedIn account banned?
The short version for budgeting purposes: one security check is a warning, not a death sentence. Pause that action family for 24 hours, drop the ramp, resume slower. Ignoring the warning is how accounts die.
The honest caveat
All LinkedIn automation, including ours, is against LinkedIn's User Agreement. Safety limits reduce the risk; nothing eliminates it. Any vendor telling you otherwise is selling you their confidence, not their engineering. The numbers in this post are what we consider defensible in 2026, they are enforced automatically on every CampaignStack account, and we revise them when LinkedIn's behavior changes.
Numbers reflect what CampaignStack enforces in production as of September 2026. We revise this post when the limits change.
Every number in this post is enforced automatically on every connected account, with the ramp, the weekly pacing and the withdrawal sweep running without you touching them.
Try CampaignStack free