Key Takeaways
- Plan on roughly 18 delivered emails per editorial link. Ahrefs published its own campaign funnel: 515 sent, 473 delivered, 5.71% of delivered emails converted to a link.
- Follow-up moves the number more than copy does. Across 12 million outreach emails, reply rate went from 8.5% on one attempt to 17.1% on three.
- Generic personalization is worse than no personalization. BuzzStream's data on 12,000 campaigns shows real contextual personalization adding 10.7 points and fake personalization costing 1.3.
- Google's Gary Illyes has said on record that links are not a top-three ranking factor and that Google needs very few links to rank a page.
- Ahrefs' study of 75,000 brands found branded web mentions correlate 0.66 to 0.71 with AI-search visibility while backlink counts correlate very weakly, which changes what is worth asking for.
- Outreach at any real volume is a cold email program. It fails on sending infrastructure and reply capacity long before it fails on the pitch.
What link building outreach returns, before you write a single email
Link building outreach is the practice of emailing the people who control other websites and asking them to link to something you published. The largest public dataset on it comes from Backlinko and Pitchbox, who analyzed 12 million outreach emails sent to 3 million websites: the reply rate on a single attempt was 8.5%. Ahrefs published the full funnel from one of its own campaigns in a link building case study: 515 emails sent, 473 delivered, 42 bounced, a 17.55% reply rate, and 5.71% of delivered emails converting into an editorial link.
That last figure is the one worth planning against, because it sets the size of the program before any argument about template quality begins. A 5.71% conversion rate means roughly 18 delivered emails per link. Every page ranking for this term will teach you the same four steps: find prospects, find their email, write a personal pitch, follow up. None of them tell you how many emails stand between you and fifteen links, which is the only question a business owner actually asked.
Follow-up moves the number more than the copy does
The Backlinko and Pitchbox data breaks reply rate out by number of attempts, and the shape of it is the most useful thing in the study. One attempt returned 8.5%. Two attempts returned 14.1%. Three attempts returned 17.1%. Emailing more than one contact at the same website pushed it further, to 16.45% on a single attempt and 22.15% across three.
Read that as a doubling of results from cadence alone, with no change to the words. Most outreach programs I have looked at send once and stop, then conclude the pitch was weak. The pitch may well have been weak. It was also only half sent.
There is a limit to this, and it is a practical one rather than a moral one. Two follow-ups spaced several days apart is a sequence a busy editor reads as persistence. Six is a sequence they report as spam, and a spam complaint costs you the sending reputation you need for every campaign after this one.
What personalization is worth, with a number attached
Every guide on this topic asserts that personalization matters and then declines to say how much. BuzzStream measured it across 12,000 campaigns from 2,000 active accounts, coding more than 30,000 individual emails by the type of personalization used. Median reply rate landed at 12%. The top 5% of senders averaged above 40%, the bottom quartile below 7%.
Inside that spread, a template-only baseline replied at 16%. Genuine contextual personalization, meaning a reference to something the recipient specifically did, added 10.7 points. Referencing a past promotion added a further 2.3.
The number that actually earns its place is the negative one. Fake personalization, the merge-field variety that inserts a first name and a site title into a sentence pretending to be admiration, cost 1.3 points against the plain template. It performed worse than sending nothing personal at all. That result should be unsurprising to anyone who has received one. A recipient who spots the seam has learned two things about you at once, and neither of them helps.
The operational consequence is that personalization is a research cost rather than a writing cost. If the personal line has to be true and specific, someone has to read the target page before the email exists. That is where outreach time actually goes, and it is why a list of 300 prospects is a fortnight of work rather than an afternoon.
It also means the honest ceiling on your program is set by how many pages a person can read carefully in a week, which is a smaller number than any tool's pricing page implies. Software will find you ten thousand prospects. It cannot tell you which of them has a reason to care about your page this month, and that judgment is the entire difference between the 40% senders and the 7% ones in BuzzStream's spread.
Then ask whether a link is still the right thing to request
This is the question the ranking pages step around, and it has moved in the last two years.
Google's own position is on record. At Pubcon Pro in 2023, Gary Illyes said links are not a top-three ranking factor and have not been for some time, as reported by Search Engine Land. In 2024 he went further, saying "we need very few links to rank pages" and that Google has made links less important over the years, documented by Search Engine Journal alongside his own confirmation that he said it. In the same window Google quietly edited the word "important" out of its spam documentation's description of links as a ranking signal.
Meanwhile the thing search visibility now depends on is being retrievable inside AI answers, and there the evidence points somewhere else. Ahrefs studied 75,000 brands and correlated visibility in ChatGPT, AI Mode and AI Overviews against a range of signals. Branded web mentions correlated 0.66 to 0.71. YouTube mentions correlated highest at about 0.737. Backlink counts and URL Rating, the two link-based metrics, showed very weak correlation. Correlation is not causation and Ahrefs says so, but the ordering is hard to ignore: being named across the web tracks with AI visibility considerably better than being linked from it does.
That reframes the ask. Outreach guides treat an unlinked brand mention as a job half finished, something to convert into a proper link with a polite follow-up. On this evidence the mention may already be doing most of the work, and it is a far easier yes for the recipient, who has to change nothing about their page. If you are choosing what to spend a research hour on, earning a citation in a piece someone is already writing beats converting a mention that exists. We wrote more about how that plays out in practice in our piece on digital PR and brand mentions.
The other half of this question is what you are allowed to ask for. Google's spam policies for web search name several practices as link spam: buying and selling links, excessive link exchanges, guest posts published at scale with optimized anchor text, and widget or footer links distributed across sites. Paid and sponsored links are acceptable only when marked with a nofollow or sponsored attribute. Worth noting because the tactics that lift reply rates fastest are exactly the ones on that list. In the Backlinko data, guest post pitches replied at 24.7%, the highest of any category, and Ahrefs found a meaningful share of site owners will simply offer to sell you the link. The high-reply path and the policy-compliant path are not the same path, and choosing the first one knowingly is a risk decision rather than a tactic.
Outreach at volume is a cold email program
This is the part I would most want a business owner to hear, because it comes from running the other kind of outbound for years.
The moment a link building program sends more than a handful of emails a day, it stops being an SEO activity and becomes a sending operation, with all the failure modes that implies. Mail from your primary domain that gets marked as spam damages the deliverability of your invoices. Sending volume that ramps too quickly on a fresh mailbox gets throttled before an editor ever sees the pitch. Authentication records that were never set up correctly send your carefully researched pitch to a junk folder, where its reply rate is zero regardless of how good the personal line was.
None of the six pages ranking for this keyword covers any of that in more than a bullet. It is the thing that actually breaks outreach programs. The mechanics are the same ones we cover for sales outbound, so if you are about to start sending, read the email deliverability checklist and the piece on cold email infrastructure first. A separate sending domain, correct authentication, a slow ramp, and a suppression list that survives your next list refresh are the difference between a program that compounds and one that quietly stops arriving.
There is a capacity limit too, and it is the constraint people size last. Every reply has to be read and answered by someone who can make decisions about your content: agree to an edit, supply a quote, approve a draft. A program producing twenty conversations a week that nobody has time to answer is worse than a smaller one, because the recipients who did say yes now think you were not serious. I ran marketing for a firm that made the Inc. 5000 four years running, from startup through to exit, and the pattern held across every channel we tried. Volume was never the binding constraint. The number of conversations one person could carry properly was.
The arithmetic to run before you commit
Take a target of fifteen editorial links over a quarter, which is a reasonable goal for a company that has not done this before.
At Ahrefs' 5.71% conversion on delivered emails, fifteen links needs about 263 delivered. Their campaign bounced 8% of sends, so call it 286 first touches. Add two follow-ups at the cadence the Backlinko data supports and you are around 850 total sends in the quarter, which is roughly 22 first touches a week. One mailbox handles that comfortably.
Then price the other side. Those 286 targets have to be found, quality checked, and read closely enough to write a true personal line each. Assume a meaningful share of any raw prospect list fails that review, because the page turns out to be dead, the site turns out to sell links, or the topic turns out to be adjacent rather than relevant. Whatever your discard rate is, the list you start from has to be several times larger than the list you send to.
That is the real cost of the channel, and it is measured in research hours rather than sending capacity. A program that fails usually failed here, by skipping the reading and shipping merge fields instead, which the BuzzStream data says performs worse than a plain template.
Run those two numbers against what fifteen links are worth to you now, given that Google says it needs very few of them and AI answers appear to care more about being mentioned. For a lot of B2B companies the honest answer is that outreach earns its place as a brand mention program with links as a welcome side effect, sized small and run properly, while the pages themselves do the ranking work. That is how we sequence it inside a search and AI-search engagement, and the SEO and AEO checklist covers the on-page half that has to be right before any of this outreach is worth sending.
