Does link building still work?

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

The short answer

Yes. Links still influence how search engines rank pages, and there is no serious evidence that they have stopped. Every practitioner running controlled tests, every large-scale correlation study, and Google's own public statements point the same way: links remain one of the signals, and on competitive queries they are one of the more consequential ones.

The caveats are large enough that the one-word answer is nearly useless on its own.

• The effect size varies enormously by query type. On some queries links are close to decisive; on others they are close to irrelevant.
• The tactics that produced links a decade ago mostly do not work, and several are actively harmful.
• The cost of a link that actually helps has risen sharply, because the cheap methods stopped working while the expensive ones did not.
• Links have acquired a second job — supporting brand presence in generative answers — that is measured differently and understood far less well.

So the accurate answer is: link building works, the returns are concentrated, and most of what is sold as link building is not the version that works. Both halves of that matter. People who say links are dead are wrong, and people who sell packages of a hundred links a month are selling something that will not help you.

What has not changed

Start with the durable part, because it explains why the practice survives at all.

The underlying mechanism is intact. A hyperlink is one of the few machine-readable signals that one independent party considered another worth referencing. Search engines still use links to discover pages, still use them to understand what a page is about, and still treat an editorial link from a relevant, trusted source as evidence. No replacement signal with the same properties has emerged.

Relevance still beats authority. A link from a modest site that genuinely covers your subject outperforms a link from a strong site that does not. This has been true throughout and is more true now than it was, as topical modeling has improved.

Editorial placement still matters. A link inside a paragraph a human wrote, in context, pointing at something worth pointing at, is the pattern that has held its value through every algorithm change since PageRank.

Concentration still describes real profiles. Good links remain rare and hard-won. On a twenty-year profile I analyzed in depth, only 446 referring domains — 2.0% of 22,260 — ever reached Trust Flow 41 or above. That is roughly 22 genuinely strong domains a year from a site that publishes constantly. Scarcity is the reason the signal still means something.

What has changed

Five things, in rough order of how much they matter to a working practitioner.

1. Volume stopped working, and then started hurting. Penguin was the turning point: the era where a thousand cheap links moved a page ended, and patterns that look built became a liability rather than a slow-burning advantage.
2. Discounting replaced penalizing for most junk. Search engines got much better at ignoring low-quality links quietly. The practical effect is that most bad links do nothing at all — neither helping nor harming — which changes the entire risk calculation around what arrives unsolicited.
3. Anchor text became dangerous at scale. Exact-match commercial anchors across unrelated domains is one of the clearest built-link signatures there is. The natural distribution is dominated by brand names and bare URLs.
4. Attribution got harder. Search results changed shape — more features, more zero-click answers, more personalization — so the clean before-and-after ranking chart that once justified a link campaign is much harder to produce.
5. Links acquired a second function. Generative answer systems draw on the same corpus of citations and mentions. Whether a link is followed or nofollowed matters less to that function than whether the mention exists at all.

Underneath all five is a single trend: the marginal value of a link that anyone can get has fallen toward zero, and the marginal value of a link that is hard to get has held or risen. That is a widening gap, not a decline.

Where the effect is largest

The honest version of "does it work" is "for what, and against whom". Links matter most in these conditions:

• Competitive commercial queries. Where several sites have comparable content, comparable technical health and comparable topical coverage, links are one of the remaining differentiators. This is where link building budget earns its keep.
• Established, crowded categories. Insurance, finance, legal services, software comparison — categories where every competitor has already done the on-page work.
• New sites in a category with entrenched incumbents. Not because links alone will break in, but because the absence of any independent citation is a hard ceiling.
• Queries where trust is part of the judgment. Health, money, safety. Independent corroboration from recognized sources plausibly carries more weight where the cost of a bad answer is higher.
• Domains with no history. A site nobody has ever referenced has a specific problem that content alone does not fix.

The common feature: links matter most where everything else is already equal. That is a genuinely useful diagnostic. If your technical health, content depth and topical coverage are behind your competitors', links are not your constraint and buying them will not help.

Where the effect is smallest

Equally important, and much less often said by people selling links.

• Low-competition queries. If nobody is competing, a relevant page with adequate content will rank. Spending on links here buys nothing.
• Navigational and branded queries. You already win these or you do not, for reasons links do not touch.
• Sites with an underlying problem. Thin content, a broken crawl, a bad site architecture, a trust problem. Links do not compensate. I have watched more link budgets wasted on this than on any other mistake — the site could not have ranked with any link profile.
• Long-tail informational queries where the winner is usually whoever answered the specific question best.
• Categories where the search result is dominated by platforms. If the results are all marketplaces, directories and video, no volume of links to your page changes the shape of the page.

There is a further consideration that rarely appears in this discussion: links decay. The median referring domain on that twenty-year profile stopped linking after 1,080 days, and 49.6% of all links recorded were eventually lost. Any program has to run at some rate simply to stand still. If the effect size in your category is small, that maintenance cost may exceed the benefit — which is a legitimate reason to decide against link building entirely.

What to conclude

I will state my position without hedging, and then mark the parts that are genuinely uncertain.

What I am confident about: links still count; relevance and editorial context are what make them count; cheap links at volume do not work and can hurt; the effect is concentrated in competitive commercial contexts; and no site should start link building before its content and technical foundation are competitive.

What is contested: how large the effect is relative to other signals, how much nofollow attribution matters now, how quickly discounting happens, and whether unlinked mentions carry weight in classical ranking as opposed to generative retrieval. Anyone giving you a confident number on those is guessing.

What nobody outside Google knows: which specific links are counted, how they are weighted, and how much of any observed change is attributable to links at all. Treat any claim that depends on knowing these as marketing.

The practical test I apply before recommending link building to anyone: would this link exist if search engines did not? If yes — it is real coverage, a real citation, a real recommendation — it is worth pursuing and it will keep working. If no, you are buying an entry in a database that is getting better at ignoring you.