TL;DR

• LinkedIn publishes no weekly invitation limit. What it documents are the triggers: invitations that are ignored, left pending or marked as spam, and many invitations in a short time.
• Warm-up is not a loophole. LinkedIn's User Agreement prohibits unauthorized automation outright; a careful ramp lowers the odds of a restriction, it does not make automation compliant.
• Get the profile ready first: real photo, accurate identity, verification badge, two-factor authentication, and a few weeks of normal manual activity.
• Ramp slowly: about 5 invitations a day in week one, adding a handful per week, and hold steady around 20 a day. Vendor schedules disagree by up to 3x; the conservative end costs you little.
• Watch acceptance rate and pending invitations more than volume. Those are the signals LinkedIn actually names.
• One steady IP and device, uneven human pacing, and a hard stop at the first warning sign.

Every LinkedIn automation guide has a warm-up schedule, and no two agree. One ramps to 60 connection requests a day by week four; another says 20 is the ceiling for good. The reason is that LinkedIn never publishes the number. This guide separates what LinkedIn actually documents from what vendors repeat, then turns both into a warm-up plan you can defend: what to fix before the first message, how fast to ramp, what to watch, and what to do when LinkedIn pushes back.

What does LinkedIn actually say about automation?

Start here, because most warm-up guides skip it. LinkedIn's User Agreement (in force since November 3, 2025) says members may not "use bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages" or "bypass or circumvent any access controls or use limits." The help page on prohibited software names "crawlers, bots, browser plug-ins, or browser extensions that scrape, modify the appearance of, or automate activity on LinkedIn's website," and warns that accounts using them risk being "restricted or shut down."

Two consequences follow. First, LinkedIn draws no line between a "safe" and an "unsafe" tool: extensions, desktop apps and cloud tools are covered alike, whatever their marketing says. Second, no warm-up schedule makes automation compliant. What a warm-up does is lower the chance that your account's behavior looks abnormal, which is how enforcement actually happens. Read everything below in that light.

Enforcement is not theoretical. LinkedIn sued scraping operators Proxycurl in January 2025 and ProAPIs in October 2025, both cases built around fake accounts at scale, and removed the company pages of several sales-tool vendors, Apollo.io and Seamless.ai in March 2025 among them.

What does LinkedIn document, and what is folklore?

LinkedIn documents triggers and consequences, not quotas. Its invitation restrictions page lists what gets accounts restricted: invitations that recipients ignore, leave pending or mark as spam; many invitations sent in a short time; and too many outstanding invitations, which can mean waiting "up to one month" before sending more. It also states that "all LinkedIn members (Basic and Premium) are subject to invitation limits," and that if LinkedIn suspects "the use of an automation tool," it "may suspend or restrict your account."

How detection works is documented too, at least in outline. LinkedIn's engineering team has described models that look at "the type, order, and frequency" of request paths and the time between them, noting that automation struggles to "simulate the subtle patterns of requests created by normal, healthy, organic member behavior." That is the whole case for warm-up in one sentence: the pattern matters more than the count.

How do you prepare the profile before any outreach?

A new or dormant account that suddenly sends invitations is the easiest pattern to spot. Fix the account first; the checklist is short.

1. A real photo and an accurate identity. LinkedIn's own guidance says a profile photo increases credibility, and a profile recipients do not recognize is the fastest route to "I don't know this person" responses, which count against you.
2. The verification badge. Verify identity or workplace. LinkedIn reports that verified members see 60% more profile views and 50% more engagement. It does not say verification raises limits, but it directly helps acceptance rate, which does matter.
3. Two-factor authentication. It protects the account, and LinkedIn increasingly builds trust features on it: colleague vouching for verification, announced in September 2026, reportedly requires it.
4. A complete profile. Headline that says what you do, current experience, a custom URL. This is observed practice rather than a documented rule, but every source agrees on it.
5. A real network and some history. Guides disagree on a minimum (anywhere from 50 to 300 connections). The point is not the number but that the account has a plausible past: people you actually know, a few posts or comments, normal browsing.

Then use the account manually for a while before automating anything. Recommendations range from one to four weeks; for a new or long-dormant account, take the longer end.

How fast should you ramp up?

Here is the spread across current vendor guides, all observed practice with no LinkedIn source behind any of it:

The outliers at the top of those ranges come from vendors selling volume. The conservative plan costs you perhaps a few dozen invitations a month and keeps you far from the patterns LinkedIn describes.

Three rules matter more than the exact numbers:

• Add one action type at a time. Engagement first (profile views, reactions, comments), then invitations, then messages to people who accepted. A new account that jumps straight to messaging strangers skips every step a person would take.
• Grow in small steps. Roughly 10 to 20% more per week, and only while acceptance holds.
• Spread activity across the working day. Uneven gaps between actions, not a burst at 9:00. "Many invitations within a short amount of time" is one of the documented triggers.

Which signals matter more than volume?

The two metrics LinkedIn's own triggers point at are the ones to watch.

Acceptance rate. Ignored invitations count against you. Practitioners put the healthy floor at 30 to 40% accepted; below about 25%, accounts report throttling. If acceptance drops, the fix is targeting and the invitation itself, not a slower send rate. Invite people with a real reason to know you, and personalize the note where you can.

Pending invitations. An invitation left pending also counts against you, and too many outstanding at once can block sending for up to a month. Withdraw invitations that have sat unanswered for two to three weeks and keep the queue in the low hundreds. Two documented caveats: after withdrawing, you cannot re-invite the same person for up to three weeks, and withdrawing will not lift a restriction that is already in place.

What makes the setup itself look human?

The rest is consistency. None of this is documented by LinkedIn as a rule; all of it follows from how LinkedIn describes its detection.

• One steady IP, in the account owner's region. No VPN hopping, no shared or datacenter IPs, no rotating proxies. A residential IP that matches where the person actually lives and works.
• One steady device profile. Same browser, operating system, language and timezone every session. Several accounts sharing one environment is a cluster signal.
• Human pacing. Varied pauses between actions, from seconds to minutes, weekdays and business hours. Evenly spaced actions are the easiest machine signature to detect.
• Your own login. Never hand a password to a tool. Log in yourself and keep the session revocable.

What should you do if your LinkedIn account gets restricted?

Warning signs arrive before restrictions: CAPTCHAs, forced logouts, "unusual activity" notices, identity verification prompts, or an invitation limit message earlier in the week than usual. Treat the first one as a hard stop, not an obstacle. Pause automation, cut volume by at least a third to a half for a week or more once you resume, and look at acceptance rate and pending invitations before anything else.

If a restriction does land, LinkedIn's restriction guidance is calmer than the fear-marketing suggests: most restrictions are lifted automatically within a week, support cannot shorten them, and withdrawing invitations does not help. For restrictions tied to automation, LinkedIn tells members to stop using the tool and change their password, and its account restrictions page warns that automated inauthentic activity "can result in temporary or permanent restriction of your account." Wait it out, fix the cause, and restart the warm-up from a lower level.

What does a warm-up look like when it is built in?

Most of this playbook is discipline that is hard to keep manually and easy to lose in a tool that optimizes for volume. It is also what we built into Curviate rather than leaving to each user. Every connected account runs in its own isolated browser with a steady home-region IP and a matching device profile; actions pace themselves with uneven, human gaps; every action type has daily, weekly and monthly ceilings you can tighten; and new seats warm up gradually before they reach full pace. You log in yourself, and the password never reaches us. The safety page has the details, and the actions reference shows what an agent can do once an account is ready.

None of that changes the first section of this guide. It keeps an account's behavior inside the patterns LinkedIn describes as normal; what the account sends, and to whom, stays your responsibility.