By Rahul Saini, Author at Search Counsel Co. Last updated [July] 2026.

Most “what makes a good backlink” advice is guesswork dressed up as fact, because Google keeps its ranking systems private. The 2024 documentation leak is the closest thing we’ve had to a look under the hood. It doesn’t reveal weightings or prove any feature is live in ranking, and Google was quick to caution against over-reading it. But it names the link features Google’s systems track, and those names line up with what SEOs have measured for years. This guide walks through how Google appears to evaluate a link: PageRank and seed sites, the tier and traffic of the source page, homepage trust, relevance and anchors, and the spam controls. Then it turns all of that into what you should actually do.

1) What the leak actually was

In late May 2024, about 2,500 pages of internal documentation for Google’s Content Warehouse API surfaced publicly, first surfaced by Rand Fishkin and Mike King. Google later confirmed the documents were authentic. What they contain is a catalog of “features,” the data attributes Google’s systems can store about a page, a link, or a site, with short internal descriptions. What they don’t contain is the recipe: no weightings, no confirmation of what’s used in live ranking. So the leak is a vocabulary list, not an answer key. That still matters, because for the first time we can see which link signals Google bothers to track, and the list quietly contradicts years of official “we don’t use that” statements. For the confirmed side of the picture, see our guide to Google ranking factors.

2) PageRank-NearestSeeds and seed sites

PageRank never died. It evolved. The leaked docs reference a feature called pageRankNS, described as “PageRank-NearestSeeds,” and note it as the production PageRank value teams should use, while the original PageRank is flagged as long-deprecated. That single line reframes link building, and it’s a useful companion to how search engines work more generally.

Here’s the concept, and it isn’t complicated. Instead of calculating authority across the whole messy web, Google appears to start from a set of trusted “seed” sites and measure how far every other page sits from them. The closer your linking page is to a seed, in link hops, the more value it can pass. Google has only ever publicly named two seeds, The New York Times and the now-defunct Google Directory, and the concept traces back to a Google patent analyzed at length by the late Bill Slawski. In practice, seed sites are a topic’s most authoritative, well-connected publications. A link from one of them, or from a page one hop away, is worth far more than a link from a site buried 40 hops out in the link graph.

• Proximity is the point. A relevant link from a trusted publication beats a pile of links from small, isolated sites that no seed site would ever link to.
• This is why digital PR wins. Earning a mention in a real publication puts you close to a seed. Directory submissions and forum drops do not. See our guide to digital PR for link building.
• Small isn’t disqualifying. A tiny but topically perfect site can still pass value if it’s itself close to trusted sources. Proximity plus relevance, not size alone.

3) Source quality, tier and freshness

The docs include a feature called sourceType that records the quality of a link’s source page, tied to the index tier the page lives in. Analysts read the tiers as roughly high, medium, and low quality (TYPE_HIGH_QUALITY, TYPE_MEDIUM_QUALITY, TYPE_LOW_QUALITY), and the description notes it “can be used as an importance indicator” for the link. Plainly put: a link from a page in Google’s top tier is worth more than the same link from a low-tier page.

Freshness gets special treatment. The docs mark freshly published content (TYPE_FRESHDOCS) as equivalent to high quality for the purpose of link importance. That’s a real strategic nudge: a link from a brand-new, high-quality article can carry more weight than one on a page that’s sat untouched for a decade. It also explains why an ongoing stream of new links tends to outperform a one-time burst that then goes stale, and why a content refresh cycle helps on the receiving end too.

4) Why zero-traffic links are worthless

This is the finding that changes how you vet a link. Per the analysis of the leak, Google appears to use clicks to a page to decide which quality tier that page belongs in. A page that earns real clicks lands in a higher tier; a page that gets none tends to fall into the low-quality tier. And links from low-tier pages appear to pass little or no PageRank. In other words, a link on a page nobody visits may do nothing for you at all.

Be careful with certainty here, because this is interpretation, not a confirmed rule, and correlation isn’t causation. But it lines up with hard evidence from two directions. First, Ahrefs’ 2025 study of links from pages with traffic found the strongest single correlate of a link’s value was the referring page’s own link strength (URL Rating), which tracks closely with whether a page is alive and well-linked. Second, when Ahrefs disavowed the links to three of its own pages, rankings and traffic fell, then recovered once the disavow was lifted, confirming that real links carry real weight. Put together, the practical rule is blunt: a link is only as good as the page it lives on, and a dead page is a dead link. Our guide to what makes a good backlink turns that into a five-minute check.

5) Homepage trust and site authority

The leak shows Google tracking PageRank and trust at the homepage level (homepagePagerankNs and homepage-trust signals), and separately a feature literally named siteAuthority inside a quality-signals module. Two takeaways follow. First, site-wide authority is a real thing Google measures, even if it isn’t computed the way Moz’s DA or Ahrefs’ DR are, so those third-party scores are rough proxies at best. Second, trust appears to flow from the homepage inward: strengthening your homepage’s authority can lift the inner pages it links to, which is a strong argument for deliberate internal linking and a clean site architecture. It’s also why a link to your homepage from a trusted source has value beyond the page it points at, it raises the trust that trickles across your whole site.

6) Relevance, anchors and mismatch

Relevance shows up on both sides of the link. The docs reference topical weighting on links and a demotion for anchor mismatch, where the anchor text and the destination page don’t line up. That confirms two long-held rules. A link from a topically related page is worth more, and your anchor text should honestly describe where the link goes. It also flags a risk: stuffing your profile with exact-match keyword anchors is a pattern Google can detect and discount, and at scale it correlates with penalties. Natural, varied, descriptive anchors are the goal. We cover the safe distribution in anchor text best practices.

One more useful detail from the docs: there’s no “dofollow” attribute, only the absence of a nofollow tag, and the documentation suggests that if a page links to you more than once and any single instance is followed, the link can be treated as followed. It’s a small point, but it’s another reason not to obsess over the follow tag on any single mention. For the full picture of what separates a strong link from a weak one, start with what makes a good backlink.

7) How Google neutralizes spam links

The docs are full of spam controls, which tells you Google expects manipulation and plans for it. Features reference link velocity (acquiring many links too fast can be flagged), anchor-spam measures, and Penguin timestamps. The takeaway isn’t fear. It’s that most junk links are simply ignored rather than punished, so a few bad links won’t sink you. The danger is a deliberate pattern: a sudden spike of paid links with matching exact-match anchors, or a profile dominated by low-tier spam, the sort of tactics covered in white hat vs black hat SEO. Google’s spam systems (Penguin, and the machine-learning SpamBrain) are built to discount those automatically. That’s also why the disavow tool is a last resort, not routine hygiene, a point we cover in toxic backlinks and the disavow tool.

8) What to do about all this

Strip away the feature names and the leak points to one strategy, the same one good SEOs already followed, now with evidence behind it.

The single highest-leverage move is to become the kind of source seed sites cite: original data, real research, genuinely useful tools. That earns links close to the seeds and, increasingly, citations in AI answers built on the same trust signals. We break that down in original research that earns links and citations, in off-page SEO for AI, and in our guide to getting cited in AI search.

9) Sources used for this guide

This is a data-led piece, so it leans on primary analyses of the leak and on named link experiments rather than opinion.

FAQ: how Google evaluates links

Conclusion: build the links Google’s own systems reward

The 2024 leak didn’t give us the algorithm, but it settled a lot of arguments. Links are scored by proximity to trusted seed sites, by the quality, freshness, and traffic of the page they sit on, and by relevance. A link on a page nobody reads is close to worthless, and no domain score changes that. So stop counting links and start earning the few that sit near the sources Google already trusts.

For the practical playbook, read how to get backlinks that actually work, and to judge the links you already have, run a backlink audit.