How to Build Backlinks: The Complete Method

I'm happier spending a month earning three editorial citations to one useful page than collecting thirty placements I have to explain later. Outreach done properly is slow, personal, and picky about the URL, which suits me after thirteen years in SEO and link acquisition.

My unit of work is the backlink, and the wider practice around it is What Is Link Building? The Definitive Answer. A backlink is a link sitting on a page from another site which points to a page on your site, as Google Search Console Help puts it.

Google explains why the work matters in its starter guide:

Google primarily finds pages through links from other pages it already crawled.

Discovery is the whole reason I bother. Pick the one URL you will put through every check in this method.

Choosing the page you want links to

A product page never gets my outreach, since commercial pages almost never attract citations. I'd rewrite or replace the URL before sending a single email.

Ahrefs states the URL decision bluntly in its linkable asset guide:

Because people rarely link to commercial content, “linkable assets” are almost never product pages or commercial content of any kind.

A page strangers can cite usually takes one of six forms: a study, a tool, a ranking, a guide, a coined term, or a visual explainer. My URL gets matched against that set before I touch a prospect list.

The chosen URL goes through three checks before anything else.

1. Ask whether an existing or intended audience would find the page useful if they came to it directly, using Google's people-first content page as the test.
2. Check the URL against the six citeable forms and reshape it until it fits one.
3. When the URL is a product or pricing page, promote a non-commercial asset instead of asking people to link at the offer.

Ahrefs showed the replacement play in a 2020 statistics page case study. Open a competing statistics page in Site Explorer, read the Anchors report, and open the Backlinks report showing one link per domain.

Record each cited statistic, hunt the original source and its age, and only then propose a replacement. I keep that campaign historical and never treat its old interface names as a current product tour.

Their research sheet recorded the statistics-page URL, linking referring domains for that statistic, the statistic itself, the original source and its age, the newer statistic with its new source and age, plus notes and pitch angles. The campaign team grouped its statistics into categories with jump links in the introduction and wrote custom pitch paragraphs for each cited statistic.

Backlinko and BuzzSumo ran 912.5 million articles published from 1 October 2017 to 1 October 2018 through their content study, released 19 February 2019:

Content longer than 3000 words gets an average of 77.2% more referring domain links than content shorter than 1000 words.

I never turn that line into a word-count target, because a second measurement points the other way. Ahrefs tested around 900 million pages held in Content Explorer for a 14 July 2020 post-length study and found the link gain turns negative past 1,000 words.

That study points one way while the Ahrefs test of around 900 million pages points the other, so I set no word-count target. Write the chosen URL into the research sheet, then keep or replace it with the direct-audience question before you open Site Explorer.

How to find the sites that could link

The Cooking Resources page from UC Santa Barbara Health and Wellness sits open on my screen whenever I explain this step. Its Food Resources section sends readers on to UCSB Food Resources and CalFresh for recipes and support, which is exactly the shape of a working resource list.

Whether my asset belongs on a similar list in my own topic is my call, without assuming the university accepts submissions or will link. That judgment call is the whole step, and the query that finds such pages follows the shape Moz prints: cooking intitle:"resources".

Moz defines intitle: as searching only within page titles and inurl: as matching words inside a URL, in its search operator reference. Google's refine searches page adds the spacing rule: leave no space between the operator and the term.

I run the first route in plain Google before I open any crawler. A short list of pages I have actually opened beats a long export I never inspect.

The prospect list builds from three routes, in this order.

1. Resource pages: run the topic plus intitle:"resources", open one real result, and check the existing destinations the way I checked the UCSB headings.
2. Unlinked mentions: search the brand in Content Explorer, exclude your own site with -site:[own website], keep one page per domain, and enter the domain under Highlight unlinked domains as the Ahrefs academy lesson describes, then confirm on the page that the brand is named without a link.
3. Competitor gap: open Organic competitors in Site Explorer and choose Open in Link Intersect, following the prospecting guide that names both moves.

Moz gives the reason unlinked mentions convert, in the same chapter:

When websites are referring to your brand or a specific piece of content you've published, they will often mention it without linking to it.

A page that names you has done the hard part, and the link is a small extra step. The mention gets confirmed on the live page before I write.

I never lead with intitle:"write for us", since Moz calls that method overused and points instead at pages where known guest bloggers already published, for example seo inurl:author/neil-patel. Moz also warns against low-quality guest posting for the sake of link building, so I treat that route as a slow extra.

Backlinko's outreach hub says to log the target name, your reason for writing, and the relevant URLs in a spreadsheet. One matching page per route goes in before I move on.

Run the resource-page query for your topic and add one matching page to the outreach spreadsheet, keeping the intitle:"resources" shape from route 1.

Qualifying a prospect before you write

Open the prospect URL itself. I trust the live page over any metric, and I fail fast when the page looks wrong.

The row passes only when all five of my page checks hold.

1. Topical or geographic relevance to the asset.
2. A real author or an about page behind the site.
3. More content than display ads on the screen.
4. The page is not an outbound-link dump.
5. Nobody is selling the link.

Fail and remove the row when any single check fails. High DR with zero traffic fails my list on sight, as do posts signed only by an editorial team and single posts stuffed with outbound links.

Ahrefs prints one starting screen in its prospecting guide: Ahrefs Domain Rating (DR) from 40 to 80, monthly organic traffic above 1000, traffic value above $1000, plus niche or audience relevance for the campaign. Those are that guide's initial qualification criteria, never Google requirements.

A second Ahrefs screen, from its outreach lessons, starts at DR 30 to 90 with website traffic 500+, 500+ words, English, one page per domain, homepages and subdomains excluded, live pages only and explicit results filtered. The two screens disagree, DR 40 to 80 with traffic 1000+ against DR 30 to 90 with traffic 500+, and neither is a Google threshold, so I inspect the page anyway.

Ahrefs shows why the visit beats the metric, with a rejected example that passed the numbers:

However, when you visit the site, you can see it is poorly designed with a mass of poor-quality content. And for the most part, there are more display ads on the screen than actual content.

That example met the numerical screens and failed the visit. I strike rows with strong DR for exactly this shape, since ad-heavy thin pages lend nothing to my asset.

When a vendor score tempts me to skip the visit, I remember that Moz's Domain Authority (DA) is a vendor metric, reread What Is Domain Authority? The Definitive Guide, and open the page anyway.

Fail or keep the current spreadsheet row on the 5 page checks above, and write the observed reason into the row when you fail it.

How to write the email

The email I refuse to send is the blast: a merged greeting, three different asks, and no mention of anything on the prospect's page. I send one short request about one citation I can point at, and I settle the follow-up policy before I send.

The send runs in three parts: the address, the components, and the follow-up policy.

1. Find a valid address with an email lookup tool like Hunter on the domain, following the prospecting guide that names that method. Strike the row when no valid address turns up.
2. Cover the components as a checklist: subject, named recipient, the prospect's page URL, the specific missing citation or resource, the destination URL, and one ask. Personalise from the research fields that record the statistic or resource the page already uses, and keep the whole thing short.
3. Choose a follow-up policy before sending, from the disagreement at the end of this section, and record it on the row.

Short means respect for the inbox, so I cut everything that is not the page URL, the missing citation, or the ask.

Backlinko measured a Pitchbox set of 12 million emails mailed through Pitchbox to 3 million websites in 2019, and the personalisation finding from its outreach study is the one that changed my drafts:

Emails with personalized message bodies have a 32.7% better response rate than those that don’t personalize their messages.

A merged first name isn't personalisation, so the body has to mention their page.

The same Pitchbox set recorded its best reply rate at three or more messages, with a single extra follow-up lifting replies by 65.8%. That metric counts replies, not links won, so don't treat a reply bump as links earned.

Backlinko's outreach hub takes the other side:

Either don’t follow up or only do it once. Journalists list “ too many follow-ups” as one of the top three most annoying things they have to deal with from email outreachers.

The source leaves a stray space before the word too inside its quotation marks, and the meaning stays unchanged.

Pitchbox data rewards a second and third message while Backlinko caps follow-ups at one, so I pick a policy before sending and write it on the row. When a prospect wants money or a reciprocal followed link, I ignore them, which is the outreach choice Ahrefs recommends in its email outreach lessons.

Write one request for a valid contact from the 6-part checklist above and the citation in the research fields, holding it to a single ask.

Journalist requests, and when they are worth answering

Open today's Help a Reporter Out (HARO) edition. I work the free emails first and answer only where my expertise fits, skipping anything that reads as a pitch.

I handle one request in five moves.

1. Match the request to real expertise and skip the rest.
2. Put the quote in the email body.
3. Send no attachments.
4. Keep self-promotion out of it.
5. Say how to credit you, and send before the deadline.

Speed matters less than fit, so a fast reply to the wrong request still loses.

The credit and format moves follow the Help a B2B Writer (HaB2BW) how-to page, published 19 January 2022. Its homepage now carries a launching-soon banner for MentionMatch, which I read as a separate observation rather than a price.

Featured, which acquired HARO from Cision in April 2025 and runs morning, afternoon, and evening editions, states the pricing position plainly on its Journalist Requests page:

HARO remains free for journalists and sources.

That free inbox is where I start, and it's never a paid desk. HARO's own page offers the same free daily queries.

The same Featured page sets the pass mark for a reply:

Strong responses are helpful, authentic, relevant, and on time: they answer the actual question, arrive before deadline, and skip the sales pitch.

I test every draft against that line and skip anything that reads as self-promotion. Connectively is the searchable real-time version of those requests, revived on 26 May 2026 according to the revival post, while HARO itself stays email.

Pick one live request that matches your expertise and draft it against the 5-move checklist above, sending before the stated deadline.

How to inspect the live link

Google illustrates the reference pattern with Appenzeller, a cheese anchor pointing at https://cheese.example.com/Appenzeller_cheese. The reference tag sits in its outbound link guide:

<a rel="sponsored" href="https://cheese.example.com/Appenzeller_cheese">Appenzeller</a>

That tag is Google's own illustration, never a real publisher placement, and it is the shape I hold every live tag against.

I record four things from the live page.

1. Open the referring URL and find the link.
2. Right-click the link, choose Inspect, and read the a tag in the Elements panel, following the Chrome DevTools docs.
3. Record the anchor text, the href, and the rel value.
4. Click the link and track the redirects, which is the check Google Search Console Help describes for confirming the destination.

Google documents sponsored for paid placements, ugc for user-generated content (UGC) such as comments and forum posts, and nofollow when the others do not apply, with several values allowed as a space- or comma-separated list. A tag carrying none of those markers counts as followed in the HTML.

The same guide draws the policy line for paid placements:

Mark links that are advertisements or paid placements (commonly called paid links) with the sponsored value.

A paid placement without sponsored or nofollow sits on the wrong side of that line, which matches the link spam case in Google's spam policy. I treat a nofollow or sponsored result as advertising, never as the editorial citation I pitched.

For my own target URL I use URL Inspection in the currently opened Google Search Console property, following the URL Inspection help, and I never inspect a publisher page through my own account. When the click lands somewhere unexpected, I record the final destination, since the target URL can reflect a redirect.

Open the referring page in Chrome and record destination and rel from the Elements panel, matching the live tag against the Appenzeller pattern above.

Measuring the links you actually won

The Links report in Google Search Console (GSC) is a sample, never a census of every link pointing at the site.

Google Search Console Help warns, on the Links report page:

This report isn't a comprehensive list of every link on your site. It shows a sample of internal and external links

That line ends my arguments with missing rows, so I don't chase rows Google never promised. My campaign sheet stays the primary record, and Google's sample sits next to it.

I compare three records side by side.

1. The spreadsheet of asks I sent.
2. The inspected live placements with their anchors and rel values.
3. Google's sample for the target page.

The Links report path runs through External links to Top linked pages. I expand with MORE and filter to the page URL and the linking site, then read Top linking text alongside it.

Google Search Console Help names the artefact I file next to the campaign sheet:

You can export your link data as a CSV file or a Google Sheet.

Single-table exports run to 1,000 rows, while Latest links and More sample links on the landing page each run to 100,000 rows. A new referring domain in a crawler is never a ranking win on its own. Only inspected placements go in the won column.

Export the target page's link data and mark which inspected placements appear in the 100,000-row sample.

When to hire help, and what to ask for

Hire when this method stops being the right use of your time. The usual trigger is an asset that needs media relationships you do not have.

The table below lists tactics that appear as sold services across the 70 provider files we score, counted as distinct files, from our catalogue dated 2026-09-06. A provider counts as selling a tactic only when a source key, title, or url in its file names the tactic as a service.

Scroll the table sideways for the remaining columns.

Guest posts at 12 of 70 and link insertions at 4 of 70 are paid-placement labels, HARO at 6 of 70 is a journalist-request service, and digital PR at 15 of 70 is the earned label. Niche edits, resource pages, broken link building, skyscraper, infographics, data studies, and unlinked mentions appear as sold services in none of the 70 files. Paid-placement labels read as exactly that: labels for placements money bought, never for citations earned.

Compare providers on Best Link Building Services, which applies our M1.3 Provider Buyer Score described in the methodology with evidence cutoff 2026-08-30. I am Sigrid Vahter, Head of Provider Research in Tallinn, with thirteen years in SEO and link acquisition that began in iGaming, several hundred publisher relationships from a Baltic agency, and current work on independent audits of bought-link portfolios.

Google's spam policy opens its link spam examples with this line and its first case:

Buying or selling links for ranking purposes. This includes: Exchanging money for links, or posts that contain links

A followed paid placement fails that test, so I would never pay for a followed guest post or link insertion.

If you hire, ask for one sample referring URL and inspect it with the Elements panel method, testing it against the 2-line buying test quoted above.