Types of Link Building: The 4 Categories That Still Work

Kirill Sajaev

SEO & Founder

 ·  8 min read

Table of contents

Link building is the practice of earning or buying links from other websites to your own. Google treats links as credibility signals, and AI assistants rely heavily on mentions, linked or not, when deciding which brands belong in an answer. Tactics fall into four groups: editorial, outreach, technical, and paid.

I have been doing this since 2010, when my job was leaving comments on forums and blog posts. It worked. It stopped working around 2012, and I still meet agencies selling it.

Key takeaways

  • Four categories cover every tactic: editorial, outreach, technical, and paid. Everything sold as a new method is a variation on one of them.
  • The majority of this industry likely uses grey hat tactics for link acquisition, us included: the question worth asking is what was bought.
  • Most link vendors share the same inventory: reseller marketplaces and dealers draw from a pool of a few thousand sites, so three vendors will often sell you placements on the same domains.
  • Buy at the page level: we look for individual pages that already rank, then check they rank for terms related to the client’s category before buying anything.
  • Mentions now matter as much as links: for DeepInfra we generated 411 brand mentions and over 300,000 Reddit impressions across 42 buyer communities, almost none of it in link form.
  • Listicles are the single best format for AI visibility: they are what assistants reach for when someone asks which tools are best.

Every tactic falls into one of four categories, separated by how the link is acquired rather than by what it looks like on the page.

1. Editorial links

Editorial links are links someone gives you without being asked, usually because you published something worth referencing.

Original research, proprietary data, free tools, and benchmark studies earn these. Very little else reliably does.

This is the category most SaaS teams underinvest in, because it requires having something worth saying.

2. Outreach-driven links

Outreach links are links you ask for: guest posts, digital PR placements, expert quotes, podcast appearances, and resource page inclusions.

It works, and it is legitimate work when the placement is relevant.

It stops working the moment relevance is traded for volume. A link from a general business tips blog does very little for a technical SaaS product.

3. Technical and product links

Technical links come from integrations, partner directories, API documentation, changelogs, open-source repositories, and marketplace listings.

These are the most undervalued links in B2B SaaS. They are relevant by definition, they rarely disappear, and your product team can often create them without a content budget.

If you have an integrations page and your partners do not link back from theirs, that is free authority sitting unclaimed.

4. Paid links

Paid links are placements bought directly from a site owner or through a marketplace. This is the largest category by spend in the industry, and the one it is least open about.

It gets its own section below, because the interesting question is what to buy.

We buy links. Judging by what gets offered to us, and by the inventory that turns up in competitor backlink profiles, the majority of this industry likely engages in some form of grey hat link acquisition too, whatever their websites say about it.

Grey hat sits between tactics search engines clearly endorse and the ones they clearly penalise. Paying for a placement on a relevant page that already ranks is grey. Buying a thousand links from a network built for the purpose is black hat with an invoice.

I would rather say that plainly than publish another page pretending the whole category runs on charm and press releases.

The real problem with bought links is inventory.

Reseller marketplaces and link dealers mostly work from the same pool of a few thousand websites. Buy from three different vendors and you will frequently be offered placements on the same domains, sitting a few paragraphs from your competitors, on sites whose only real business is selling links.

Those sites have a domain rating and almost nothing else. No audience, no rankings of their own, no reason for anyone to read the page your link sits on.

How we choose placements instead

We buy at the page level rather than the domain level. The process is simple enough to copy.

  • Find pages that already rank: use Ahrefs or Semrush to pull individual URLs holding positions for terms in the client’s category, rather than filtering a list of domains by authority score.
  • Test for strength: does the page hold real positions and receive real traffic? A page ranking in the top ten for terms people actually search has demonstrated that search engines trust it.
  • Test for relevance: do the keywords it ranks for relate to what the client sells? A high-traffic page about kitchen renovation is worthless to an AI inference platform, whatever its domain rating says.
  • Buy the placement on that specific page: when a page passes both tests, we buy a placement on it.

What you get is a link on a page with existing traffic and demonstrated topical authority, which is a different asset from a link on a domain with a good score.

For AKOOL the deliverable was roughly ten vetted links per month from relevant, authoritative sources, with no spam and no private blog networks. Ten placements chosen this way outperform a hundred bought off a list.

To be fair to the marketplaces, they are fast and cheap, and speed sometimes matters. But you are buying volume from a shared pool, and most likely so are the competitors you are trying to outrank.

A mention is your brand name appearing in text, with or without a link attached. For classic SEO, an unlinked mention did comparatively little. For AI visibility it can matter more than a link.

The reason is how language models build associations.

Models learn from how often words appear together. When a model has repeatedly encountered your product named alongside its category and its competitors, it treats you as part of that set. That association is what decides whether you appear when someone asks which tools are best in your space.

Retrieval works on the same principle at answer time. When an assistant fetches sources to answer a question, it summarises what those pages say.

If your brand is named in a retrieved page, you are in the answer. Whether the mention carried a link never enters into it.

Which is why I stopped filtering prospect lists by follow status a long time ago, and why we now track mentions as a deliverable in their own right.

For DeepInfra, an AI inference platform, we mapped the 42 subreddits where ML engineers and startup CTOs actually compare providers, then built a presence in them over time. That produced 411 brand mentions and more than 300,000 impressions, with zero callouts and zero account flags.

Almost none of that is a link. It is still the most valuable authority work we did on that account.

For fraud.net, content and authority work together reached 1,230 AI Overview citations and 5.6% AI share of voice by April 2026, against 937,000 monthly impressions.

Why listicles win

Listicles do more for mentions than any other format, because they are what assistants reach for when someone asks a comparison question.

Ask any assistant for the best tools in a category and look at what it cites. Roundups, comparison posts, and best-of lists, because those pages already contain a structured set of named options with descriptions attached.

A model answering “best fraud detection platforms” reaches for a page that already names ten of them. Your homepage names one.

That makes inclusion in ranking listicles the most direct path to being named in AI answers.

Listicle strategies that work

  • Get into the lists that already rank: run the same page-level analysis described above, filtered to best-of and comparison pages in your category. Then get added, by outreach where the publisher is open to it and by paid placement where they are not. This is the same play as buying a link, except you also gain a mention inside a source assistants quote.
  • Publish your own comparisons, including competitors: a roundup that names only your product will not be cited. One that fairly compares the category will be, and you control how your entry reads.
  • Keep them updated: best-of queries reward recency, and both search engines and assistants favour lists that look current. A 2024 roundup gets skipped.
  • Make each entry extractable: a clear product name, a one-sentence description, and an explicit category label. If a model has to infer what you do from a paragraph of positioning, it often gets it wrong.
  • Audit the lists you are missing from: pull the pages that rank for “best [your category]” and mark which ones name your competitors and not you. That gap list is the highest-value outreach target you will build this quarter.

The overlap here is the part I find most useful. A paid placement in a ranking listicle delivers a link, a brand mention, and presence inside a page assistants cite, from one transaction.

There is no universal number. The only figure that matters is how many referring domains the pages currently ranking for your target query have, which you can check in any backlink tool in about five minutes.

For a low-competition long-tail query in a technical niche, that number is often under ten. For a broad category term it runs into the thousands.

I get asked for a target number constantly and I understand why, because a number is easier to brief than a judgment. Anyone who gives you one without looking at your SERP is guessing.

A backlink is a single link. A referring domain is a unique website linking to you, regardless of how many links it sends.

Referring domains is the more useful metric. Fifty links from one site is one referring domain and behaves, roughly, like one vote.

When a tool reports a large backlink number next to a small referring domain count, that usually means sitewide footer or sidebar links. It looks impressive in a report and does very little.

Yes, indirectly. A nofollow link passes no direct ranking equity, but it still drives referral traffic, helps search engines discover pages, and contributes a brand mention.

Since 2019 Google has treated nofollow as a hint rather than a directive, which means it may choose to count the link anyway.

For AI visibility the distinction collapses completely, for the reasons in the mentions section above.

Forum backlinks are links placed in forum posts, signatures, or profile pages. Built manually at scale, they were a dominant tactic of the early 2010s and are now close to worthless for ranking.

I know, because that was my first job in this industry.

What replaced them is genuinely different. Participating in a community where your buyers already research, under a real identity, produces the mentions that both people and models read. The DeepInfra numbers above came from exactly that.

Digital PR earns links by giving journalists something worth covering, usually data. HARO and its successors earn them by answering reporter queries with real expertise.

Both work. Both are slower and less predictable than vendors admit.

Reciprocal links are fine in small numbers between genuinely related businesses and obvious spam at scale. Google has named excessive link exchanges as a violation for over a decade.

My rule: if you would not make the link without the return favour, do not make it.

Yes, and the return has widened. Links still influence rankings, and the mentions that come with them now influence which brands assistants name, which is where a growing share of B2B discovery happens.

For Microblink, an aggressive B2B link acquisition programme helped grow the value of their organic traffic to $33,000 per month.

For Vizard.ai, we retooled the same approach specifically for LLM discovery and increased their page one keywords by 3,000.

Link building still works. Its payoff now reaches well beyond rankings.

What I would actually do first

  • Claim your technical links: integration directories, partner sites, marketplace listings, documentation. Fastest wins, lowest effort, permanently relevant.
  • Build your missing-from-listicles gap list: every best-of page ranking in your category that names competitors and not you. Work that list first.
  • Buy at the page level: if you are going to pay, pay for placements on pages that already rank for terms related to what you sell.
  • Track mentions alongside links: mentions drive AI visibility, and a report that only counts referring domains misses them.
  • Build one genuinely linkable asset: original data beats another roundup. You only need one to change what people have to cite.

Final thoughts

Most link building advice is written as though the goal is a number in a dashboard. The goal is for the sources your buyers trust, and the sources machines quote, to mention you accurately and often.

I spent my first two years in this industry building links that Google spent the next two years learning to ignore.

What survived that is the work where the link was a by-product of being worth mentioning, and the buying was guided by judgment about individual pages.

Read more

Common questions about link building and mentions

  • Does AUQ buy links?

    Yes. The majority of this industry likely engages in some form of grey hat link acquisition, and we would rather say so than pretend otherwise. We buy placements on individual pages that already rank for terms related to a client’s category. We do not buy from reseller lists or private blog networks, because that inventory is shared across every vendor and every competitor.

  • What counts as grey hat link building?

    Grey hat sits between tactics search engines clearly endorse and the ones they clearly penalise. Paying for a placement on a relevant page that already ranks is grey. Buying a thousand links from a network built for that purpose is black hat with an invoice.

  • How do you pick which pages to buy a link on?

    Two tests. Strength: does the page hold real positions and get real traffic? Relevance: do the keywords it ranks for relate to what the client sells? A page has to pass both. Domain rating on its own tells you very little.

  • Do mentions without links help?

    For AI visibility, a great deal. Assistants summarise the sources they retrieve, so a brand named in a cited page appears in the answer whether or not the mention carried a link. Repeated mentions alongside your category also build the association that decides when you get named at all.

  • Why do listicles matter so much for AI search?

    Because they are what assistants cite for comparison questions. A model answering which tools are best in a category reaches for pages that already name a set of options with descriptions. Being inside those pages is the most direct route into the answer.

  • How many backlinks do I need to rank?

    There is no fixed number. Check how many referring domains the pages currently ranking for your target query have, and use that as the benchmark. For a long-tail technical query it is often under ten. For a broad category term it can run into the thousands.

  • Do nofollow links help SEO?

    Indirectly. They pass no direct ranking equity, but they drive referral traffic, aid discovery, and create a brand mention. Since 2019 Google has treated nofollow as a hint rather than a directive, so it may count the link anyway.