How to Build Backlinks: The Complete Method

You have one page you want links to, and a week you can actually spend on it. That is the whole setup for this guide, and everything below assumes it. A backlink, in Google's own plain wording, is "a link on a page from another site that links to a page on your site." Google finds new pages through links on pages it already crawled — so the unit of this work is one citation, on one live page, pointing at your one page. Follow the method for that page first; only afterwards decide whether it scales to a second one.

The method runs in five moves: pick a page worth citing, find the people who cite things like it, check each of them on the live page, send one short specific request, then inspect whatever link lands. Most guides invert this and start with the prospect list, which is why most outreach fails before the first email goes out. In my experience the page decides whether outreach is possible at all, and no amount of personalisation rescues a page nobody has a reason to cite.

Start with the page, not the prospect list

Open your target page and ask the question Google's people-first guidance asks of all content: would someone who arrived without Google, an existing or intended audience who came directly to you, find this page useful? If the honest answer is no, stop here, because a stranger with a publishing queue will answer the same way. I treat this as the qualifying round for the entire campaign, and I would rather spend two days fixing the page than two weeks pitching a page that cannot win.

The pattern is consistent enough to state flat. Product pages, pricing pages, category pages and homepage-style overviews almost never attract citations, because they answer "buy from us" and nobody cites that sentence. What gets cited instead is reference material: original studies and survey figures, free tools and calculators, ranked lists, thorough guides, a coined term people need to define, or a visual explanation that makes a hard idea fast to grasp. If you want the evidence behind that claim, the old Backlinko and BuzzSumo study of 912.5 million articles found that "why", "what" and infographic formats earned about a quarter more links than video and how-to posts. Against that sits Ahrefs' later crawl of roughly 900 million pages, where the length-to-links correlation turned negative past 1,000 words. The two datasets disagree, so I chase neither word counts nor formats as rules.

Read that disagreement as permission to think instead of counting. The test is whether your page gives a writer something their draft is missing: a number, a definition, a worked example, a checkable fact. When I audit bought-link portfolios I see the same failure from the other side, which is placements pointing at commercial pages with anchor text doing all the persuasive work the page itself never does. That is a favour being asked, not a citation being earned, and editors can tell the difference in about four seconds.

The one-page decision

So make the decision before you build any list. If your commercial page is the thing that must rank, you have two honest options, and I have run both. Either build a citable asset next to it and let internal links carry the value across, or accept that every request you send is a favour — guest post, insertion fee, exchange — and price it accordingly. What you cannot do is pitch a pricing page as though it were research and then wonder why the reply rate sits at zero.

The asset version works like this in practice. Take a payroll features page that no writer on earth would cite: the pattern I reach for is a small free calculator published beside it — enter headcount and pay frequency, get the employer's true cost per run. The features page never earns a link directly. The calculator earns the citations and the internal link carries the value across. The cost is a designer and a day of a developer's time, against a month of outreach spent asking for favours.

Buying the favours instead is the alternative I do not do, and the reason is practical rather than moral. Google's spam policies name buying or selling links for ranking purposes, excessive link exchanges and links with optimised anchor text in distributed articles as link spam. Paid links qualified with rel="sponsored" or nofollow are permitted. Read that carefully and the bought guest-post market collapses into advertising that passes no citation value, which is a fine thing to buy if you want traffic but a strange thing to buy as SEO. If someone must buy, I want them to know which of the two they are buying.

The statistics-page play

The cleanest example of "give them a reason" is also one of the oldest, and it still works because stale numbers never stop embarrassing writers. Find a statistics page in your niche that already holds links, read which statistic people actually cite, trace it to the original source and check the date, then publish a newer figure and pitch the people citing the stale one. You are not asking for a favour here; you are offering a correction, and corrections get answered.

Ahrefs ran this exact campaign in 2020 and published the numbers, which is why I keep citing it: they reviewed about 200 prospects, emailed 111 of them, heard back from 41% and won 19 links. Their tracking sheet held the statistics page URL, the referring domains citing it, the statistic itself, the original source and its age, the newer statistic and its source, plus notes and a pitch angle per row. Steal that sheet layout rather than inventing your own, because every column in it answers a question the pitch has to answer anyway.

Start on the concrete report. In Ahrefs' Content Explorer you search the topic, filter to the statistics-style pages with the most referring domains, and open the top one to see which figure the citations cluster around. Then run the unlinked-mention variant in reverse: search the brand or figure, exclude your own site, and look at who mentions the topic without linking anywhere current. The typical opening is a niche where every roundup still cites a 2019 survey — small-business failure rates ran exactly this shape for years — and a fresh cut of the same question, even with a smaller sample honestly labelled, gives every one of those writers a correction to make. Old data is a liability to the writer citing it, and your pitch is the fix.

Finding prospects with Google before paying for anything

You do not need a paid tool to build the first fifty prospects; you need three operators and the discipline to read the results. For resource pages, Moz's worked query is still the template: cooking intitle:"resources", swapped for your topic, where intitle: searches page titles only and there is no space after the colon. The UC Santa Barbara "Cooking Resources" page this kind of query surfaces is a good specimen of the species: a plain list of food and cooking links on a wellness subdomain. It shows what you are hunting — a maintained list with an owner, not a link dump. I take "maintained" seriously here and check the page dates and whether dead entries have been pruned; a list nobody has touched in four years will not add yours either.

For guest-post openings, skip the query every spammer runs. Moz is right that intitle:"write for us" is exhausted — every spammer runs it — and I agree with their alternative: find where known guest bloggers in your niche already publish, with something like seo inurl:author/neil-patel, since inurl: matches words in the URL and author slugs are wonderfully regular. One real byline on a real site beats twenty "write for us" pages, because it proves the site publishes outsiders and tells you who edits them. When the query returns a name, I open the author archive, read the two newest posts, and note the editor's name from the author box or the accompanying about page before anything else happens.

A note on budget, since it splits the method early. The free slow way is these operators plus a spreadsheet, and it finds the first fifty prospects in an afternoon. The paid fast way is a crawler subscription — Ahrefs Lite at $129 a month, Semrush's SEO plan at $139, Moz Pro Standard at $99, each with a free or $29 entry tier — and it finds the next five hundred. If this is your first campaign, I would run the free way first; the constraint forces you to read pages, and reading pages is the skill the rest of this guide depends on.

Finding prospects with a crawler

Three reports earn their keep once you pay, and each answers a different question about who already links to things like yours. Unlinked brand mentions answer "who named us and forgot the link": in Ahrefs' Content Explorer you search the brand, exclude your own site with -site:yoursite.com, switch on one page per domain, and tick "Highlight unlinked domains". Competitor gap answers "who links to two rivals but not me": Ahrefs Link Intersect and Moz Link Intersect both do this, and Semrush's Backlink Gap runs the same comparison across up to five sites. The one-page-per-domain filter matters in all of them because a single enthusiastic forum or a site-wide footer can otherwise flood the list with fifty rows that are really one decision-maker.

Run the gap report on the concrete pair first. Put your two closest content rivals in, not the giants of the niche — a small SaaS tool should intersect two direct competitors, not a publisher ten times its size whose links it will never replicate. Sort the output by referring domains to the linking page rather than by the linking domain's authority, because a link from a strong page on a modest site beats a link from a forgotten page on a strong site. Then de-duplicate against everything you already found with Google before counting; in my experience a third of the crawler's output repeats the operators' output, and counting it twice is how lists look bigger than they are.

If you want the broader context on what these scores predict and where they mislead, our What Is Domain Authority? The Definitive Guide walks through it. The short version for this step: export the report, keep the URL, the linking page title, the rival it links to, and one line on why your page is a fit. Anything you cannot write that line for leaves the list now rather than after you have emailed it.

How big a list is big enough

Small and read beats big and scraped, and I keep lists deliberately short. My working minimum is 100 qualified prospects per target page, built over two or three sittings, because below that number one bad week of replies ends the campaign. The published numbers back the shape of this: Ahrefs' statistics campaign reviewed about 200 prospects, emailed 111 after filtering, and won 19 links — roughly a one-in-six hit rate on emailed prospects and one-in-ten on reviewed ones. Plan your list backwards from that arithmetic: if you want ten links, expect to review two hundred pages.

That 19-from-111 figure also sets the patience horizon. Outreach compounds across follow-ups and across months, not across list size alone, and a list of 500 uninspected domains is a list, not a campaign. I would rather see a reader finish this guide with 60 prospects they have actually opened than 600 they have only filtered. The qualifying step below is where the list earns its size, and it is the step most people skip.

Qualifying a prospect: the five-minute inspection

This is the part of the method I care about most, and it happens on the live page with no tool open. Five checks, about a minute each, and a prospect that fails any of them leaves the list regardless of its metrics. First, relevance you can state in one sentence: the page's topic genuinely touches yours, not via three hops of category logic. Second, a real author or about page behind the content — a byline, an author box, a masthead with names — because anonymous content with no owner cannot be pitched and should not be trusted. Third, more content than ads on the visible screen. Ahrefs' own rejected example put it perfectly: a site thrown out because "for the most part, there are more display ads on the screen than actual content." Fourth, the page is not an outbound-link dump: if the article links to forty vendors with keyword-rich anchors, your link joins a sold inventory, not an editorial choice. Fifth, nobody is selling the link — no "advertise", "sponsored posts from $50" or rate card anywhere on the path from the page to the contact form.

I run this routine because it beats a Domain Rating floor, and I say that as someone who uses DR daily as a starting filter. The published screens prove the point about what those numbers are. One Ahrefs guide starts at DR 40 to 80 with 1,000+ monthly organic visits and $1,000+ traffic value. Another walkthrough starts at DR 30 to 90 with 500+ traffic, 500+ words, one page per domain and live pages only. Those two ranges cannot both be Google's rule, because neither is — they are author filters, guesses at where the good sites cluster, and the DR component measures the strength of a site's own backlinks rather than the editorial standard of any single page. A DR 65 site with a sold link inventory fails my fourth check in seconds, while a DR 22 niche blog with a named author and a careful reading list passes all five. If you want the full argument on why vendor scores stop at the filter stage, the authority guide linked above makes it at length; here the rule is short and I enforce it strictly.

The email

Every email I send has four parts and nothing else: their page named exactly, the specific thing missing or wrong on it, my page as the fix, and one ask. A resource-page pitch reads something like: your packing checklist links to a 2019 baggage-fee table that Delta has since changed — I maintain the current table with the 2025 figures here — would you consider swapping it in? That is the whole message plus a subject line under fifty characters. Backlinko's outreach guide is right to say the logging happens before sending: target name, why you are writing and the relevant URLs go in the spreadsheet first. The email stays short because the thinking already happened elsewhere.

Personalisation means one observed sentence, never a merge field. The Backlinko and Pitchbox study of 12 million emails found personalised subject lines earned 30.5% more replies and personalised bodies 32.7% more, with 8.5% of all emails getting any reply and 36-to-50-character subjects doing best. My reading of that data is narrow: the lift comes from proof you opened the page — the missing-stat sentence or the dead link you name — not the recipient's first name in the greeting. I write that sentence from the inspection notes, which is why prospects without inspection notes never get emailed.

Follow-ups are where I have to pick a policy because the data genuinely disagrees with itself. The same 12-million-email dataset found one extra follow-up lifting replies by 65.8%, with sequences of three or more doing best. Against that, Backlinko's own guide says to follow up once or not at all, journalists rank "too many follow-ups" among their top three annoyances, and Brian Dean says he emails once. My policy splits the difference by prospect type, and I hold it: one polite follow-up after five to seven working days for site owners and bloggers, zero follow-ups for journalists on deadline. Eighteen follow-ups is not persistence; it is the behaviour the spam folder was built for, and I have watched it burn domains that had good pages.

What not to do

The shortest section, because Google already wrote most of it. Do not buy links for ranking purposes, exchange links at scale, run automated link programs, or place keyword-stuffed anchors in distributed guest posts or press releases. The spam policies name all four almost verbatim. And do not send AI-blasted pitches with attachments to a thousand strangers — that one fails on its own economics long before policy matters. Scaled guest posting deserves its own warning since it still gets sold as safe: a guest post written for the host's audience with one natural citation is a real thing, while fifty near-identical posts on fifty invented blogs are a link network with an editorial disguise. I audit portfolios where the only difference between those two cases is visible in the author boxes, and most buyers never looked.

Swaps and mass exchanges sit in the same bucket under a friendlier name. One reciprocal link between two sites that genuinely reference each other is normal web behaviour; a spreadsheet of thirty arranged swaps is an exchange scheme, and the pattern is trivially visible to anyone who pulls both backlink profiles. Attachments, for completeness: never attach anything to a first email, since filters eat them and strangers do not open them. Link the page and let the page do the work — which returns, as everything in this guide does, to whether the page is worth opening.

Journalist requests: an hour a day, answered properly

Journalist-request services invert the model: instead of pitching your page, you answer a reporter's stated question and the citation follows. The lineup as the services stated it in September 2026 runs from free to serious money. HARO at helpareporter.com is free with three editions a day under Featured's ownership, and Connectively relaunched that May as the searchable real-time version. Featured itself runs a free tier of two to three opportunities a week with Lite at $29 and Pro at $79 a month on annual billing. Qwoted gives two delayed pitches a month free and 35 pitches at $149 Pro, SourceBottle offers free alerts with $25 assisted pitches or $65 a month unlimited, and Help a B2B Writer is mid-rebrand to MentionMatch. Prices move, so confirm before budgeting; the shape of the market is what matters here.

Whether it is worth an hour a day depends on whether you hold genuine expertise the reporter cannot Google. Featured's own answering advice is the whole playbook in three lines: match your expertise, meet the deadline, skip the sales pitch. I would add two working rules from watching what gets picked up. Answer only where you are the practitioner — a payroll founder on payroll questions, not a payroll founder on interior design — and file before the deadline with the credential in the first line, because reporters triage by "can I quote this person" in seconds. Anything with a pitch for your product goes in the bin, and rightly so. If your week has five hours for links and no citable asset yet, I would spend four building the asset and one answering requests; requests without expertise behind them convert at lottery rates.

Inspecting the link that lands

A link nobody has inspected does not count, so this step is not optional. When someone says your link is live, open the page, right-click the link, choose Inspect, and read the actual <a> tag: the anchor text, the href, and the rel attribute. Google defines three rel values for outbound links: sponsored for paid placements, ugc for comments and forum posts, and nofollow for everything else you want to disqualify. They are written as a space- or comma-separated list, and Google states plainly that links carrying them "will generally not be followed." Follow the href through any redirects as well, since a chain through two tracking hops to a different destination is not the citation you pitched.

Treat the result honestly. A nofollow or sponsored link is advertising or moderation, not the editorial citation you asked for: log it, thank them, and do not count it toward the campaign. A followed link with honest anchor text on a page that passed inspection is the win, so check the anchor reads naturally — branded or descriptive, not the keyword-stuffed string from a link seller's menu.

And the uncomfortable case I see in audits constantly: a paid placement with no sponsored marking. That is Google's problem in the sense that policy forbids it, and yours in the sense that your site sits at the end of the unmarked paid link. I flag these to the buyer in writing and keep them out of the win column; counting them is how portfolios quietly accumulate the exact risk they paid to avoid.

Measuring what you won

My spreadsheet has four columns after the prospect's name: sent, replied, live, inspected. Only the fourth column scores, because a link that is live but uninspected is a rumour and a link that is inspected but nofollow is a different product. Against that sheet I set Search Console's Links report, read with its stated caveats kept alongside. Google says the report "isn't a comprehensive list of every link" and includes links "found over time" which "may have since been removed." It never tells you whether a link is nofollow. It is a sample of what Google has recorded, which makes it the right second opinion rather than the scoreboard.

Two timing notes save needless worry. URL Inspection only works on URLs inside the property you own and shows the last indexed version rather than the live web, so use it for your own target page and the Elements panel for the publisher's. And crawlers take days to find new links. Ahrefs puts it plainly in its help pages: "it can take a few days up to some weeks before we crawl your new backlink." I check a new placement after two to three weeks, not the next morning. A link missing from every tool in week one is normal operations rather than failure. If you need the definitional companion to this section, our What Is Link Building? The Definitive Answer sets out what counts as what before anything gets measured.

When to hire help, and what to ask for

Hire when the asset needs relationships you do not have, which in practice means digital PR. A data study that deserves national-press coverage needs a practitioner with a media book, a newsroom sense of timing and follow-up capacity across dozens of desks; none of that is learnable in the week this guide assumes. What you are buying is earned coverage on editorial merit, which is also why it costs more and converts less predictably than anything with a price per placement. Our ranking of link building services compares providers on that basis, and our methodology shows how the scoring works — start there rather than with whoever emailed you first.

Never pay for a followed guest post or a followed link insertion, under whatever euphemism the menu uses. Across the 70 providers we score (catalogue dated 6 September 2026, evidence cutoff 30 August 2026), the named-service counts run as follows:

Scroll the table sideways for the remaining columns.

Guest posts and link insertions are paid-placement labels and digital PR is the earned label; niche edits, resource-page placements, broken link building, skyscraper, infographics, data studies and unlinked-mention campaigns appear as sold services in none of the 70 files. Read any provider's menu against that split before believing the word "outreach" in it. And whoever you shortlist, ask for one sample referring URL and run the five-minute inspection from this guide on it — author, ads, outbound-link pattern, rate card. A provider whose sample fails inspection has told you everything about the inventory, and no case study undoes what the live page shows.