After this lesson you should be able to
- Separate the discovery effect of links from the ranking effect
- Explain why link-ranking correlation studies overstate their case
- Estimate how much links are likely to matter for a given query
- State which claims about links are genuinely contested
Three jobs, routinely confused
When someone says "links improve rankings", they are usually compressing three different effects into one claim. Pulling them apart is the single most useful thing you can do before reading any study.
- Discovery. Links get pages crawled and indexed. This effect is uncontroversial, mechanical, and observable — you can watch a new page get indexed after an external link appears. It is also the effect people undervalue most, because it does not feel like ranking.
- Ranking contribution. Links feed systems that assess a page's importance and standing, which affects its position for queries it is relevant to. This effect is real but indirect, delayed, and impossible to isolate for any individual link.
- Referral and brand. Links send people. Those people convert, remember you, search for you by name, and sometimes cite you themselves. This effect has nothing to do with the algorithm and is the only one you can measure directly.
Most arguments about whether link building "still works" are two people talking about different items on this list.
The correlation problem
Every year somebody publishes a study finding that pages ranking in the top positions have more referring domains than pages ranking below them. The correlation is consistently positive and often strong. It is also close to useless as evidence of causation, for reasons that are not subtle.
Reverse causation
Pages that rank well get seen more, so they get linked more. A page sitting at position one for a popular query accumulates citations precisely because it is easy to find. The study cannot distinguish links causing the ranking from the ranking causing the links, and in the real world both happen at once.
Confounding by quality
Genuinely good pages attract links and rank well, without one causing the other. Better content, better brands, larger companies with PR budgets, and older sites all produce both outcomes. A correlation between links and rankings is partly a correlation between "being a substantial organization" and "both of these things".
Selection on the outcome
Studying only pages that already rank in the top ten throws away every page that had links and did not rank. That is exactly the group you would need to see to know how often links fail to produce rankings.
Aggregation across query types
Averaging a local plumber query and a national insurance query produces a number that describes neither. The effect of links varies far more between query types than the headline correlation suggests.
None of this means links do not work. It means the studies routinely cited as proof are not proof, and you should stop quoting them as if they were.
What we can say with reasonable confidence
Strip out the weak evidence and a defensible set of claims remains.
- Links cause discovery. Directly observable, no inference needed.
- Google says links are a signal. Google representatives have described links as an important ranking signal consistently over many years, while also saying they are not the top signal and that their weight has decreased over time. This is testimony rather than measurement, but it is testimony from the only party with the data.
- Competitive commercial queries are dominated by well-linked sites. The exceptions are rare enough to be notable. Whatever the mechanism, you will not find many pages ranking for high-value commercial terms with no external citations at all.
- Removing links can hurt. This is the most persuasive line of evidence available to practitioners, because it runs the causal arrow the right way. Sites that lose large numbers of links — through a network being deindexed, a partner site closing, or a cleanup — frequently lose rankings. That is the closest thing to a natural experiment this field offers.
- Relevance modulates everything. A link from a page about your subject behaves differently from a link on an unrelated page with an identical metric profile. Nobody serious disputes this, and every practitioner has watched it happen.
Why the effect size varies so much
This is the part that generic advice gets wrong, and it matters more than anything else in this lesson. How much links matter depends almost entirely on what you are trying to rank for.
At one end: a long-tail informational query where three pages on the whole web address the topic properly. Here relevance and basic technical health do nearly all the work. Links contribute mostly by getting the page discovered. I have seen pages with no external links at all rank first for queries like this and stay there for years.
At the other end: a national query where money is on the table and every competitor has a budget, an agency, and a decade of history. Content quality is table stakes — everyone has it — so the differentiators are the signals that are hard to manufacture. Links carry much more weight here, not because the algorithm treats them differently but because everything else is equalized.
Between those poles, the practical test is a comparison, not an absolute. Look at who ranks for your target query and pull their referring domain counts. If the top ten have between ten and forty referring domains each, links are a modest and achievable factor. If they have thousands, links are the game, and a first campaign will not change your position — it will only start the clock.
Query intent shifts things too. Results where Google appears to favor established sources — health, finance, legal, anything where bad information causes harm — reward the kind of standing that citations from recognized institutions provide, and reward it more than a comparable hobbyist query would.
What is genuinely contested
Honest reporting means naming the parts nobody has settled. These are live disagreements among competent practitioners, not questions with a hidden right answer.
- How much weight links carry now relative to a decade ago. Google has said the weight has come down. Practitioners disagree about how far, and there is no way to measure it from the outside.
- Whether nofollow links contribute anything. Google described nofollow as a hint rather than a directive when it introduced the sponsored and UGC attributes, which leaves the door open. Many practitioners report ranking movement from heavy nofollow coverage. Others attribute that movement to the traffic and brand searches the coverage produces. Both explanations fit the data.
- Whether link velocity matters in itself. The idea that acquiring links too quickly triggers scrutiny is widely believed and has never been demonstrated cleanly. Real events — a product launch, a viral study — produce sharp spikes routinely without consequence.
- How much a single high-authority link is worth against many modest ones. Everyone agrees the distribution is skewed. Nobody agrees on the exponent.
- Whether toxicity scores predict anything. Tool vendors sell them. I have never seen evidence that they correlate with actual algorithmic outcomes, and I treat them as a sorting aid at best.
If a course, a vendor, or an agency gives you confident answers on all five, that confidence is a sales technique.
Questions
How many links do I need to rank?
There is no universal number, and anyone quoting one is guessing. The answerable version is comparative: count the referring domains of the pages currently ranking for your target query and treat that range as the entry price. For many local and long-tail queries that number is in the low double digits. For national commercial terms it can be in the thousands.
Can a page rank with zero backlinks?
Yes, routinely, when the query is not competitive and the page is genuinely the best available answer. Internal links from a site that already has standing will handle discovery. What almost never happens is a page with no external citations ranking for a high-value commercial query where every competitor has invested for years.
Do links still work as well as they used to?
No, and Google has said as much — the weight given to links has come down as other signals improved. But "less than before" is not "irrelevant". Links remain one of the few inputs that cannot be self-declared, which is exactly why they retain value in competitive results where every on-page factor is already optimized.
Why did my rankings not move after I built links?
Most often because the links were not the constraint. If your page is thin, slow to be crawled, poorly matched to the query intent, or competing against sites with a hundred times your history, links will not fix it. Second most common reason: not enough time. Crawling, indexing and reassessment of the linking pages takes weeks, sometimes months.