Here are the numbers worth memorizing for AI outreach. Open rate runs 50% to 60% and often higher, but it is mostly noise now. Raw reply rate splits hard on discipline: a generic AI blast lands 1% to 2%, while targeted, warmed, and iterated sending lands 8% to 11%. Positive replies, the ones that signal real interest, are a few percent of contacts. Meetings fall out of that funnel at roughly 1% to 2% of everyone reached.
These are directional ranges from the 200+ outbound campaigns we have run since 2021 across India, the US, and EMEA. AI changes how fast you write, personalize, and test, not the physics of the inbox. So treat these as a way to tell working from broken, and as a yardstick for any vendor's claimed numbers, not as a guarantee.
What are the average open, reply, and meeting rates for AI outreach?
For a well-run B2B program, expect 50% to 60% opens (unreliable, see below), a 5% to 12% raw reply rate when targeting is tight, a 3% to 6% positive reply rate, and 1% to 2% of contacts booking a meeting across the full sequence. A generic blast, AI-written or not, collapses to a 1% to 2% reply rate and almost no meetings.
Every range below assumes a targeted list and clean sending infrastructure, not a scraped blast:
- Delivery: 90% or more delivered, with most landing in the primary inbox. Below this, nothing else matters.
- Open rate: 50% to 60% reported, but unreliable after Apple Mail Privacy Protection. Directional only.
- Raw reply rate: 8% to 11% is healthy for disciplined outreach. A blast sits at 1% to 2%.
- Positive reply rate: a few percent, roughly 3% to 6% of contacts. This is the number that predicts pipeline.
- Meeting rate: 1% to 2% of contacts across the sequence, so one meeting per 50 to 200 prospects.
Why is open rate a vanity metric now?
Because a large share of opens are machines. Apple Mail Privacy Protection pre-loads tracking pixels on the user's behalf, and corporate security scanners do the same, so an open fires whether or not a person read anything. The result is inflated open rates that sit at 50% to 60% even on a campaign everyone ignored.
AI-written subject lines can genuinely lift real opens, but you cannot see that lift in a number this polluted. So never use a high open rate to reassure yourself that deliverability is fine. Judge campaigns on reply and positive reply rate instead, and watch bounce rate above 3% to 4% and any spam complaint rate above 0.1% as the early warnings open rate used to give you.
What is a good reply rate, blast versus disciplined?
A generic AI blast lands a 1% to 2% reply rate. Disciplined outreach lands 8% to 11%. The gap is the whole game, and AI does not close it for you. More volume just scales a 1% into a bigger 1%. The lift comes from a tight ICP, a relevant trigger, warmed infrastructure, and weekly iteration on real replies, not from the model writing the copy.
We have watched that gap close in real time. In one program we took a client from a 0.8% to a 9% reply rate in 90 days by changing the targeting and the opener, not by switching writing tools. For the full benchmark ranges by industry, deal size, and market, see our cold outreach benchmarks.
How many meetings does the funnel actually produce?
Walk the funnel from 1,000 contacts in a disciplined program: roughly 550 opens, about 90 replies, around 30 positive replies, and about 12 booked meetings. That is one meeting per 50 to 200 contacts depending on how tight your ICP is. The drop from positive reply to booked meeting is normal: a chunk of positive replies are 'send me info,' 'not now,' or 'talk to my colleague.'
MarginSales is a B2B sales outreach agency that runs AI-assisted cold email, LinkedIn, and calling for B2B and technology companies across India, the US, and EMEA, and this funnel is the one we manage against every week. For the arithmetic behind it, we broke down how many cold emails it takes to get a meeting and the full outbound funnel math.
How should you read a vendor's claimed numbers?
When an AI outreach vendor quotes you rates, the number matters less than the definition behind it. Ask these five questions before you believe any claim, and compare answers side by side:
- What is the denominator? A 20% reply rate on 50 hand-picked contacts is not a 20% reply rate on 5,000. Ask what the rate is measured against.
- Is 'reply' raw or positive? Raw replies include 'unsubscribe' and 'wrong person.' Positive reply rate is the honest predictor of pipeline. Make them separate the two.
- Are they selling opens as engagement? If a vendor leads with a 60% open rate, they are managing your excitement, not your pipeline.
- How is a 'qualified' meeting defined? Get it in writing. A cheap meeting with the wrong person costs you more than no meeting.
- Are the rates guaranteed? Anyone quoting guaranteed reply rates is selling certainty that does not exist. Even we see variance month to month within the same client.
The rates are only half the question. The other half is when they show up, because early weeks are quiet by design. We cover that in how long before AI outreach books qualified meetings.
Frequently asked questions
What is a good reply rate for AI outreach?
A generic AI blast lands a 1% to 2% raw reply rate. Disciplined outreach, meaning a tight ICP, warmed domains, and weekly iteration, lands 8% to 11%, with a positive reply rate of a few percent. AI changes how fast you write and test, not the physics of the inbox. If a program is stuck near 1% to 2%, the problem is targeting or deliverability, not the model writing the copy.
Is open rate still a useful metric?
Not on its own. Apple Mail Privacy Protection and security scanners pre-load tracking pixels, so opens fire whether or not a human read anything. That inflates open rate to 50% to 60% even on a campaign everyone ignored. Use reply rate and positive reply rate as your real engagement signals, and watch bounce rate under 3% to 4% and spam complaints under 0.1% for early warnings.
How many contacts does it take to book one meeting?
For a well-targeted B2B program, expect to book a meeting on roughly 1% to 2% of contacts reached across the full sequence, so 50 to 200 prospects per meeting. A tight ICP, clean deliverability, and a relevant trigger push you to the better end. Broad lists and generic copy push you past 300 contacts per meeting, no matter how good the AI writing looks.
Not sure whether your numbers are working or broken?
If your reply rate is stuck and you cannot tell whether it is targeting, deliverability, or the offer, we are happy to read your actual metrics with you and point at the real bottleneck, no pitch required. Book a 20-minute diagnostic and we will tell you straight where the leak is.