On this page (9)
- 01What does LinkedIn say about automation?
- 02What does LinkedIn actually detect?
- 03Chrome extension, cloud tool or isolated browser: which is safer?
- 04How many actions per day are safe on LinkedIn?
- 05What makes automated messages look automated?
- 06What should you do if a profile gets restricted?
- 07Should you automate your own profile at all?
- 08How Bazooka handles this
- 09Questions people ask
What does LinkedIn say about automation?
The rule is short. LinkedIn's User Agreement says you agree not to use bots or other unauthorized automated methods to access the service, add contacts or send messages. There is no approved list of automation tools, and no tool can make you exempt from that rule.
In practice, enforcement is about behaviour. LinkedIn doesn't publish its thresholds. It watches for activity that doesn't look like a person, and it responds on a ladder: a warning, a request to verify your identity, a temporary restriction, and for repeat cases, a permanent one.
27%
of 218 automated accounts restricted within 90 days in a Q1 2026 test
Source: Cleverly (vendor test, 2026)
That's a vendor test, so read it as a signal, not a census. It does tell you that “just install it and go” is a coin flip you'll eventually lose.
What does LinkedIn actually detect?
Almost nothing on this list is about one big number. It's about patterns a human doesn't produce:
- Volume jumps. 5 requests a day for a month, then 60 on a Monday.
- Perfect intervals. A request every 90 seconds, all afternoon.
- No browsing. Requests and messages with no scrolling, no profile reading, no reactions in between.
- Identical text. The same note sent to hundreds of people.
- Low acceptance and a pile of pending invitations. Lots of people ignoring you is a signal on its own.
- Odd logins. The same profile showing up from new IPs, devices or countries.
Look at that list again. Volume is one line out of six. The rest is about looking like a script.
Chrome extension, cloud tool or isolated browser: which is safer?
Where your profile runs matters as much as what it sends. Here's how the three common setups compare:
| Setup | Where it runs | Main risk | What to check |
|---|---|---|---|
| Browser extension | Your own Chrome, your own IP | Automation sits inside your live session; stops when the laptop sleeps | Can it run without you logged in and watching? |
| Cloud sequencer | A vendor server, 24/7 | Shared or changing IPs; sends with nothing human around them | Dedicated IP per profile? Same location every time? |
| Isolated browser per profile | A real browser per profile, own IP | Still subject to pacing and message quality | Does it browse, view profiles and pace clicks like a person? |
Browser extension
- Where it runs
- Your own Chrome, your own IP
- Main risk
- Automation sits inside your live session; stops when the laptop sleeps
- What to check
- Can it run without you logged in and watching?
Cloud sequencer
- Where it runs
- A vendor server, 24/7
- Main risk
- Shared or changing IPs; sends with nothing human around them
- What to check
- Dedicated IP per profile? Same location every time?
Isolated browser per profile
- Where it runs
- A real browser per profile, own IP
- Main risk
- Still subject to pacing and message quality
- What to check
- Does it browse, view profiles and pace clicks like a person?
None of these is safe at bad behaviour. An isolated browser sending 80 identical notes a day still gets flagged. But a profile that always comes from the same IP, in its own browser, with real browsing around the sends, gives LinkedIn a lot less to object to.
How many actions per day are safe on LinkedIn?
LinkedIn doesn't publish a number, so everything here is observed, not official.
~100
connection requests per rolling seven days, the ceiling most senders observe
Source: FirstSales, 2026
- Connection requests: around 11 to 14 a working day on a warmed profile.
- Profile views: up to about 70 to 80 a day, spread out.
- New profiles: start low and ramp over a few weeks. A fresh profile at full volume is the fastest way to a restriction.
- Pending invitations: withdraw old ones you no longer expect to be accepted.
Our LinkedIn connection limits guide goes through the numbers and the weekly cap in detail.
What makes automated messages look automated?
Two things: sameness and emptiness. Sameness is the same note to everyone. Emptiness is a note that could be about anyone.
Hi {firstName}, I'm expanding my network with leaders in {industry}. Would love to connect!
Hey Tom, your post on rebuilding the partner program was the most honest take I've read on it. Curious how you're handling the first 90 days.
The second message can't be sent to 500 people unchanged, and that's exactly why it's safer. Writing each one from research on the person is also how you get replies. We cover whether AI-written LinkedIn messages still work, and the short answer is: only when they're built on real research.
What should you do if a profile gets restricted?
Stop all automation on it immediately. Complete any verification LinkedIn asks for. Once access is back, restart at half your old pace and change whatever triggered it. If every note was identical, that was the problem, not the count. The full recovery steps are in LinkedIn account restricted.
Should you automate your own profile at all?
Your main profile carries your network, your content and your reputation. Losing it for a week hurts more than any number of missed requests.
And spread volume. Five profiles at 12 requests a day reach the same people as one profile at 60, with a fraction of the risk.
What it does on LinkedIn, step by step.
Bazooka runs every sending profile in its own isolated browser with its own dedicated IP, with human pacing on every click. Every lead gets a profile view before any outreach, and each profile stays at 11 to 14 connection requests and 70 to 80 profile views a day. Volume grows by adding senders, never by pushing one profile harder.
You can bring your own profiles (AI Starter, $497 a month for 8 senders) or run on warmed, ready-made profiles we provide (AI Accelerator, $1,897 a month with 24, or AI Scaler, $3,297 with 48), so your personal profile stays out of cold outreach. Messages are written from research on each prospect, never one template to everyone. See how it runs LinkedIn, the pricing, or get a demo.
Questions people ask
Can LinkedIn detect automation?
LinkedIn detects patterns: volume jumps, perfectly even timing, identical messages, no browsing, low acceptance and logins from changing IPs. Automation that avoids those patterns is much harder to tell apart from a person.
Will I get banned for using LinkedIn automation?
Most problems start as warnings or temporary restrictions, not permanent bans. Repeat offences after a restriction are what lead to permanent ones.
Is a cloud tool safer than a Chrome extension?
Not automatically. What matters is a stable dedicated IP per profile, human pacing and real browsing around the sends. A cloud tool without those is no safer than an extension.
How many connection requests per week is safe?
Most senders stay under about 100 per rolling 7 days per profile, or roughly 11 to 14 per working day on a warmed profile.
Is it safe to use AI to write LinkedIn messages?
The risk is sameness, not AI. Messages written from real research on each person look personal and get more replies. AI that sends one generic line to everyone looks like a bot.
Should I use my main LinkedIn profile for automation?
Ideally not for cold outreach. Keep it for content, warm replies and closing, and run outreach from separate warmed profiles.
Sources
Get the next playbook in your inbox.
Leave your email and we will send you new guides on LinkedIn and the other channels we run.