How we do outreach: the LinkHatch playbook

Most link building advice is written from theory. This is written from files.

LinkHatch took over a body of work that Incify ran for years: 207 labelled email threads from 2020 to 2024, 41 spreadsheets (link trackers, prospect lists, a publisher database), 3,111 addresses that ever replied to the outreach inbox, and 146 contacts we went on to work with. Add the 2026 programme, with its own prospect sheet and live link checks. Everything below comes from those records, and where the records are thin, we say so.

1. How we pick who to contact

Nobody here starts with "who has a high number". Every campaign started with a written test, and the test changed with the client. These are the ones on record.

Where	The test
A niche online retailer	Traffic over 1,000 a month. DR over 30 (Ahrefs). Relevant. Low spam score.
A medical education platform	A block list of 21 competing brands. A High, Medium or Low priority tag on 934 prospect rows.
A US agency's software client	A shortlist of 48 sites, none under DA 30, each held until an editor confirmed.
Our own publisher database	A real "About us" page, flagged bad topics, display ads, country targeting, DR, traffic and a quality tag, per site.

The database is the clearest picture of what "kept" means. It holds 2,333 site rows (2,311 unique domains) with columns for DR, estimated traffic, whether the site has a real About page, whether it carries topics we flag, and whether it runs display ads. Its tags include "low quality" (175 sites), "traffic under 100" (83) and "DA under 30" (42).

We re-ran the retailer's test on that database to show what a filter does to a list. This is our arithmetic on the sheet, not the selection anyone made at the time.

Step	Sites left
All rows in the database	2,333
DR over 30 and traffic over 1,000	757
and has a real About page	524
and no flagged bad topics	439
and no display ads	313

So about 13% of the database survives. That ratio is why we don't hand clients a list of 2,000 "opportunities".

The same shape shows up in the 2026 medical education programme: 1,535 prospect rows, 1,130 contacted all-time, 18 sites with a real or form reply logged, 3 exchanges complete with both links live. Each stage is a different thing, so don't read it as a precise funnel, but the direction is the point. Prospecting is mostly deciding who to leave out.

2. The pitch that gets replies

We went through the replies to see which messages worked. The pattern is the same across 2020, 2021, 2024 and 2026.

It names a specific page of theirs. "I noticed your article on [topic]" with the real topic. In 2026 that opening line is written per prospect, points at a real passage ("especially the explanation of how spaced repetition helps students retain complex information") and a person checks it before it goes.

It names a specific page of ours, with a reason. The pitch says why the two pages belong together for a reader, not why a link would be nice for us.

It offers a way out. "If interlinking is not an option, we can also contribute quality content to your blog." That line is in the 2024 pitches, and it's what the drafted answer to a "we don't do interlinking" reply was built on: three article topics we could write for them.

It's clear about the ask. The 2021 first touch opens by asking whether the site takes guest contributions from writers relevant to its industry, and offers ideas from our writers.

Once someone says yes, we do the design work. The strongest threads are the ones where we wrote the sentence, the anchor and the spot on both pages, so the editor only had to paste it.

Here's the 2024 first touch, with names removed:

Hi [first name],

[Name] from [the platform] here. I noticed your insightful article on [their topic] and felt it perfectly complements our own piece titled [our title] on our blog.

Would you be interested in interlinking our articles? I believe it would provide both our audiences with a more comprehensive view of [shared subject].

We'd love to collaborate for mutual benefit!

If interlinking is not an option, we can also contribute quality content to your blog. Happy to share samples/topics if you're interested.

Best, [Name]

And a typical editor reply, because the replies matter more than the pitch: "We'd be open to linking to your article if we could get a link to another page in return!" That's a yes with a condition, and the condition is reasonable.

3. Follow-up cadence, as it actually ran

These are the timings in the threads, not a recommended schedule.

• Second touch: If the first email gets nothing, send a second as a fresh message, not a reply to nothing. The one documented case (an AI-tools company) got an answer: "Normally we don't do guest posts and backlink trades, but I am willing to listen to what you can propose in return."
• Content to publication: Every hand-off email carries a deadline: publish within 24 to 48 hours, and "if there is no response from you within 48 hours, I'll send a follow-up but even then if there is no response, then I'll have to cancel and send the content to another relevant site for publishing."
• Chasing a late publisher: Ten days after sending the content, then every two days: three short chases, on the 23rd, 25th and 27th. The editor apologised and published.
• Checking your spam folder: One editor followed up after a month of silence from us, because his emails had landed in our spam folder. Check yours.
• Out-of-office replies: Logged with a return date and followed up when the person is back (an April return, a July return).
• Weekly send rate in 2026: 7 to 15 new prospects a week. The sheet's own number, and a deliberate one.
• Nurture: A warm lead that stalls gets a polite hold and a re-touch date. One is set for October 2026.

4. How we verify a link

A link counts when it's checked live: the right page, the right anchor, and indexed. An email that says "it's live" is only a lead on where to look.

In practice that's:

1. Open the page and record the HTTP status.
2. Find every link to the client and record the destination and the anchor.
3. Record a status in the tracker. The vocabulary in use is Live, Unverified, Live (site issue) and Removed, each with a dated note.
4. Re-check. On 7 October 2026 we re-opened 34 placed pages: 21 still showed the client link, 13 didn't confirm. The reasons ran from bot walls and expired certificates to a client link that had been taken out.

Indexing is the one step we can't show you. The files hold no index check, so we describe the rule and don't claim a result.

5. Our standards

Four standards, each tied to a record.

Genuine prospecting

Every list started from a written test, and the test was applied to a database, not guessed. The retailer's four-line test, run over our own database, keeps about 13% of it (757 of 2,333 on the two numeric filters, 313 after three more flags). Each of the medical education programme's 1,535 prospects is a real blog post matched to a client page. Prospecting here is slow, which is the point.

Real one-to-one outreach

The first touch is a short note with the person's name and the specific article. In 2024 it went through an outreach tool. In 2026 every personalised opening line is checked by a person and sent only when that person ticks a box. After a reply, the conversation carries on by hand in the same thread, and the longest labelled thread runs to 109 messages. LinkHatch's rule is one-to-one conversations, never cold blasts.

Strict site qualification

The tests above are written, and the records show sites we said no to and why.

What we turned away	Why
A medical-information site	Closed as not relevant.
A site flagged for brand safety	Closed, logged as rejected for brand safety.
A publisher whose pitch read like a link farm	Closed. It didn't meet our quality bar.
A medical-appraisal site	The client had no article that fit. Polite decline, door left open.
A site with gaming posts on its home page	A client comment, recorded in the tracker against a finance placement.
21 competing brands	On the block list. Excluded.
Pages that already link to a client	An exclusion list, so we don't spend time on a second link.

Traffic: where we recorded it, it was above the bar. The retailer's 33 published links sit on sites with a recorded 1,000 to 62,000 visits a month (median 5,300). For the other accounts the trackers record DA but not traffic, so we don't quote it.

What a LinkHatch link is

• On an independent site. Every link sits on a site that someone else owns and runs. Nothing in the records shows a site owned or run by Incify or LinkHatch.
• Checked, and replaced when it fails. A link counts only once it's checked live, with the right anchor on the right page. If a link is later removed or stops passing value, it's flagged and replaced.
• No link farms. LinkHatch doesn't place links on private blog networks or link farms.
• A written test, and your approval. Every LinkHatch campaign starts from a written site test like the retailer's four lines above, and you approve the sites before anything goes ahead.

What the records can't tell us

• Outcomes. The files hold no rankings, traffic or lead numbers for any client, so these are placement and process records, not results.
• Indexing. No index checks.
• Reply rate. The reply logs only hold threads that got an answer, so there's no denominator for most campaigns.
• Dates on some sheets. Several trackers have month names and no year. We used the dates inside the URLs and email threads.

The short version

1. Write the test before you hunt.
2. Leave most sites out.
3. Name their page, name ours, give a reason, offer a second route.
4. Do the design work for the editor.
5. Chase on a rhythm and check your spam folder.
6. Count a link only after you've opened the page.
7. Say what you can't prove.