The state of outbound in 2026: what changed and what works

What changed for email deliverability?

The rules themselves are not new. Gmail and Yahoo set their bulk-sender requirements in February 2024. What changed is that they now bite: failing mail gets bounced back instead of quietly landing in spam.

Three details matter for cold outbound. First, the 0.3% spam ceiling and the authentication basics are Gmail rules for all senders, so low volume does not exempt you; Google recommends staying under 0.1%. Second, bulk-sender status at Gmail is permanent once assigned. Third, spreading volume across many domains keeps each one under the thresholds, but complaints still count per domain, so it does not change the math of a bad list.

What reply rate should you expect from cold email?

It depends on who is counting, and the published numbers for 2025 disagree by about ten times. All four datasets below come from companies that sell outbound services or tools.

• Per email or per person. A four-step sequence to 1,000 people is 4,000 emails, so Hunter's 4.5% per recipient is roughly 1.1% per email on that basis.
• All replies or good replies. A "not interested" counts as a reply in most of these datasets.
• Who is sending. An agency's client campaigns and a tool's whole user base are very different populations.
• The method changed. Belkins' earlier studies quoted 5.8%, measured against opens; it switched to emails sent after dropping the tracking pixel. Many 2026 articles still quote the old figure.

Meetings are the honest metric, and they are rare. Belkins booked "over 1,200 appointments" from 7.5 million emails in 2025, which works out to roughly one meeting per 6,000 emails by our arithmetic. One more caution: the 3.43% figure now circulates on several vendor blogs with shifting attribution. The earliest dated source we found is Instantly's own January 2026 report.

What actually lifts replies?

Where the datasets agree, the advice is consistent: send less, to better-chosen people, with fewer steps. These figures are vendor data from Hunter and Instantly.

• Smaller lists: sequences with 21 to 50 recipients replied at 6.2%, against 2.4% for 500 or more.
• No open tracking: campaigns without it replied at 7.4%, against 4.4% with it. Belkins dropped its pixel because it "was hurting deliverability."
• Three messages beat one: 6.8% against 3.3% total replies, and Hunter found the best sequence length fell from four messages to three.
• The first email does most of the work: 58% of replies in Instantly's data came from step one.
• A real domain: sending from a custom domain replied at 5.2%, against 2.5% from free email addresses.

How has LinkedIn changed?

LinkedIn is where buyers say they want to be reached, and it is also the platform least tolerant of automation. In Hunter's survey of decision makers, LinkedIn was "the preferred outreach channel for 50.5%," against 25% for email (vendor data, sample size not published).

• No official invitation limit. LinkedIn publishes no weekly number; it restricts accounts that send "an excessive number of invitations" and suspects automation, and most restrictions lift within a week. The "safe limits" quoted by tools are their own policies.
• Free members can add a personal note to only three connection requests a month.
• Third-party automation breaks the User Agreement. LinkedIn's help centre says members who use bots or extensions that automate activity "risk having their accounts restricted or shut down."
• Vendors are a target too. LinkedIn sued the data provider Proxycurl, which shut down in July 2025, and on March 25, 2026 it removed the LinkedIn company page of the automation tool HeyReach and restricted its executives' profiles. HeyReach says customer accounts kept working.

The working rule: use LinkedIn at a human pace, from real profiles, with messages a person would actually send.

Where do phone and texting fit?

Calling still earns its place alongside email. In Gong's data, a call "nearly doubles your email reply rate (3.44% vs 1.81%) even if you don't connect live," though the average rep connects with only 5.4% of prospects and the top quartile with 13.3% (vendor data).

Texting is a consent channel. In the US, carriers block business texts from local numbers that are not registered under A2P 10DLC. In January 2025 the 11th Circuit struck down the FCC's "one-to-one consent" rule, but the older rule stands: marketing texts sent with automated systems to mobile numbers still need prior express written consent. In Canada, CASL covers texts like email and carriers filter business texts hard (our CASL guide has the detail). In both countries, cold texts to people who never opted in are the riskiest thing an outbound team can send.

How are buyers reacting to AI outreach?

With fatigue. Gartner's buyer surveys are the most solid data here:

• 73% of B2B buyers "actively avoid suppliers who send irrelevant outreach" (Gartner, June 2025).
• 67% prefer a rep-free experience, and 45% used AI during a recent purchase (Gartner, March 2026).
• Yet 69% prefer to validate AI-generated insights with a sales rep (Gartner, May 2026).
• 69% of decision makers say it bothers them if AI wrote the email, and "too sales-focused" is now the top reason cold emails fail (Hunter, vendor data).
• By 2028 AI agents will outnumber human sellers ten to one, while fewer than 40% of sellers will say agents improved their productivity (Gartner prediction, November 2025).

Read together, buyers do not want to be prospected at, but they still want a person at the moment of decision. That favours outreach that is relevant, short and easy to say no to. The cautionary tale of the year was the AI SDR startup 11x, reported by TechCrunch in March 2025 to have listed customers it did not have. One company's story, but a clear sign that "replace your SDR with AI" was oversold.

What is happening to the tool stack?

• Consolidation at the top: Clari and Salesloft announced a merger in August 2025.
• Repricing: in March 2026 Clay split its pricing into Data Credits and Actions and cut marketplace data prices by "50-90%"; moving to the new model was optional. Some small teams and agencies pushed back on Actions costs and moved to cheaper tables or do-it-yourself setups, while others argued the return still justifies the price.
• Platforms absorbing partners: in June 2026 Clay's changelog said its own lookalike feature replaces Ocean.io's data in Clay. Ocean's founder has since said the split worked out well for Ocean.
• Clay itself is growing by public measures, so the leavers are a segment, not a trend.

One cost of a split stack rarely gets discussed: opt-outs. Each tool keeps its own unsubscribe list, while the deadlines apply to you as the sender.

What about outbound in Canada?

Canada is an opt-in market. CASL requires consent before a commercial email, LinkedIn message or text, there is no general B2B exemption, and the sender carries the burden of proving consent. Enforcement is quiet: the CRTC issued one notice of violation between April and September 2025 while receiving 152,603 complaints, and three out of four investigations start with a complaint. Our full guide covers the consent routes that work for cold B2B outreach.

On buyer behaviour, the data is thin. One agency dataset found Canadian recipients replied at 0.63%, against 0.51% in the US (Belkins, 266,000 emails to Canada). Treat that as a hint, not a finding.

Eight takeaways for small outbound teams

1. Treat authentication and spam rate as table stakes at any volume: SPF, DKIM, DMARC, reverse DNS and TLS, with complaints under 0.1%.
2. Send smaller, tighter campaigns to people whose job owns the problem.
3. Turn off open tracking and measure replies and meetings instead.
4. Write three or four good steps, and put the most care into the first.
5. Add a second channel, and respect its rules: calls alongside email, LinkedIn at human pace, texts only with consent.
6. Use AI for research and drafting, and keep the voice human.
7. Judge benchmarks by their method, then benchmark yourself: meetings per 1,000 contacts on your own data.
8. Keep one opt-out list across every channel, and know which steps in your stack burn metered credits.

Where Duckrow fits

Several of these takeaways are easier when the list, the sequences and the inbox share one database. In Duckrow, an email unsubscribe suppresses the address across the whole workspace, a STOP reply opts the lead out of further texts, and a reply on email, LinkedIn or text stops that lead's next steps. LinkedIn steps go out from your own account, and Duckrow does not add open-tracking pixels to emails. It does not replace judgement about who to contact; it makes the follow-through consistent.