Why a 22% LinkedIn Accept Rate Is Better Than It Sounds
A 22% LinkedIn connection accept rate sits in the normal band. Here is what it turns into across 400 invites, and why targeting drives it more than AI copy.
Is a 22% LinkedIn connection accept rate actually good?
Yes. It sits at the bottom of the normal band for a well targeted campaign. Our published range is 20 to 30% accept on connection requests, 12 to 18% reply to the first message after an accept, and 5 to 8% on follow-ups. A 22% accept rate means the list is right and the profile is passable.
Here is what 22% turns into on a single warm account sending 20 invites a day, 20 days a month:
We broke this down further in what a 22% accept rate actually turns into. The number people fixate on is the accept rate. The number that pays rent is the fourth one down.
Did the AI cause the 22%, or did the targeting?
The targeting. Accept rate is decided before your message is read, because the prospect sees your name, your headline, your photo and your mutual connections. Most of them do not open the note. An AI that writes a better connection note is optimising the part of the funnel with the least leverage.
What moves accept rate, in rough order:
We wrote a whole post arguing this: a 22% accept rate is a targeting win.
What does the AI actually change, then?
The replies. Sequence builders have been able to send a connection request, wait three days and fire a follow-up since about 2016. That part was never the bottleneck. The bottleneck is the 15 to 22 humans a month who answer and then need a reply from you within a few hours, in your voice, that handles "what does it cost" differently from "not right now".
Expandi, Salesforge and Overloop all run the sending side well. Overloop bundles prospect sourcing into the campaign build, which suits a team that has no list yet. Salesforge personalises across LinkedIn and email in 21 or more languages with unlimited senders. If your problem is volume across many accounts, those are the right shape of product.
If your problem is that 18 people replied last week and you answered nine of them on Sunday night, more sending capacity makes it worse.
Which tools do people mean by "AI LinkedIn outreach"?
Four different products get called the same thing. Some automate the sending, some source the list, some generate the copy, and a small number hold the conversation afterwards. The split that matters for a buyer is what is left on your desk once the tool is running.
| Tool | What the AI does | What you still do |
|---|---|---|
| Expandi | Cloud-based LinkedIn and email sequences, free for 7 days | Write the copy, answer replies, book meetings |
| Salesforge | Personalisation across LinkedIn and email, 21+ languages, unlimited senders | Answer replies, book meetings |
| Overloop | Sources prospects and builds the campaign | Write, reply, book |
| n8n build | Scrapes Apollo leads with Apify, enriches, sends DMs | Maintain the workflow, write, reply, book |
| Gelee (Custom) | Writes in your voice, answers replies, handles objections, books the meeting | Take the call |
How many invites a day can you send without getting restricted?
20 to 25 a day on a warm account, 5 to 10 a day for the first two weeks on a new one. LinkedIn does not publish a limit and it moves with account age, how many of your invites get ignored, and how many people report you. A 3-year-old account with 2,000 connections gets more room than a 2-month-old one.
At 25 a day across 20 working days that is 500 invites a month from one account. At 22% accept that is 110 accepts. If you need 40 meetings a month, one account will not do it at these rates, and the answer is more sender accounts rather than more invites per account. That is the gap Expandi and Salesforge are built for, and it is a real reason to buy them.
What does running this manually actually cost you?
Sourcing and sending 500 invites, then answering 20 or so conversations, runs about an hour a day. That is 20 hours a month. At $100 an hour of founder time that is $2,000 on top of whatever the tool charges. A $99 sequence tool plus 20 hours is a $2,099 channel.
The arithmetic only flips one way. If you are pre-revenue and your hour is worth $20, 20 hours is $400 and the cheap tool wins outright. If you are a founder taking sales calls at a $15,000 average contract value, the 20 hours is the expensive part of the line item and the subscription price is noise. We worked through the same comparison against cold email in is AI LinkedIn outreach worth automating first.
Where do these campaigns break?
At the reply, and at the personalisation token. Scraped first lines fail loudly: a prospect whose headline says "Head of Growth | ex-Stripe | dog dad" gets a message about their dog, and they screenshot it. The second failure is slower. Replies pile up for four days, the prospect has moved on, and the accept rate still looks great in the dashboard.
Expandi's benchmark set of 13.2M connection requests across 13,302 accounts and sbl.so's 285,000 messages, which put average first-touch reply at 8 to 12% both measure the sending side. Neither measures how many of those replies got answered the same day, because no tool reports it.
Who should not automate LinkedIn outreach?
Four situations. Deal sizes under $2,000, where 15 meetings a month does not cover the cost. No product-market fit, where more conversations produce more polite no. Buyers who are not on LinkedIn, like local trades or most of healthcare ops. And a total addressable list under about 300 people, where you should write 300 manual messages instead.
Gelee is also the wrong buy if you already have two SDRs sitting there with capacity. Buy them Expandi or Salesforge, give them better lists, and spend the difference on the list. We have no self-serve signup, so there is no cheap way to try us for a month and see.
Should you copy the setup in that thread?
Copy the list building and the daily volume. Those produced the 22%. Do not assume the AI is why it worked, and do not scale sending before you can answer replies within a few hours.
The honest sequence of fixes, in order: narrow the list to 300 to 500 real fits, rewrite the headline, send 20 to 25 a day from one warm account, and only then decide whether you need software to hold the conversations. If you get through that list and the replies are what is drowning you, that is the point where an AI SDR earns its place. Pricing is shared on a demo call.