PageRank

TL;DR — PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. is Google’s original way of scoring how important a page is, based on who links to it. Links are votes — and a vote from a popular page counts more than a vote from a nobody. Google killed the public PageRank score in 2016, but it still uses the algorithm behind the scenes.

What PageRank is

The original PageRank paper describes a link-analysis method based on the web’s link graph. Evidence for this claim Official or primary documentation supporting the adjacent article claim, with scope limited to the source's published description. Scope: No ranking guarantee or undisclosed system mechanics are inferred beyond the cited source. Confidence: high · Verified: 2026-07-14 Stanford: PageRank paper Google says link analysis systems, including PageRank, remain part of its core ranking systems alongside many other signals. Evidence for this claim Official or primary documentation supporting the adjacent article claim, with scope limited to the source's published description. Scope: No ranking guarantee or undisclosed system mechanics are inferred beyond the cited source. Confidence: high · Verified: 2026-07-14 Google: Ranking systems guide

Back in the late 1990s, two Stanford students — Larry Page and Sergey Brin — had an idea: you could figure out which web pages are important by looking at the pattern of which pages link to them — a page linked to by important pages is itself important, and that importance flows recursively through the whole link graph. They called it PageRank (named after Larry Page, with a pun on web “pages”).

The simple, non-technical way to say that: links are votes. If lots of pages link to you, that’s a sign your page is worth something. But not all votes count the same — a link from a hugely important page is worth far more than a link from an unknown one. And in the model as published, there’s a catch: each page only has so much voting power to give, and it gets split across every link on the page. Link to 100 things and each link passes a tiny sliver; link to 3 things and each gets a bigger share.

”I thought Google got rid of PageRank?”

For years, you could see a page’s PageRank as a 0-to-10 score in the Google Toolbar. Google stopped updating that score in 2013 and removed it completely in 2016. A lot of people took that to mean PageRank was dead.

It isn’t. Google removed the public display — not the algorithm. Google still lists PageRank as one of its core ranking systems, and a big 2024 leak of internal Google documents showed several versions of PageRank still in use.

What this means for you

• Earn links from good, relevant pages. A few links from genuinely authoritative, on-topic sites beat a pile of links from junk.
• Where a link sits matters. A link in the main body of an article is worth more than one buried in the footer (more on this in the Advanced version).
• Internal linksAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. matter too. How you link between your own pages decides where your PageRank flows inside your site.
• Don’t obsess over a number. There’s no public score anymore. Tools like Ahrefs’ URL Rating estimate it, but no third-party number is the real PageRank.

Want the formula, the Reasonable Surfer model, the full history, and what the 2024 leak actually revealed? Switch to the Advanced tab.

TL;DR — PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. is a recursive graph model: a page’s score is the probability a “random surfer” lands on it by following links, and that score depends on the scores of the pages linking to it. In the published model, each page splits its PageRank across its outbound links, and the calculation runs iteratively until it converges. The original 1998 algorithm was replaced internally around 2006, the Reasonable Surfer patent describes a model where not all links pass equal value (a patent claim, not a confirmed current weighting), and the public Toolbar score died in 2016. Google still lists PageRank as a core ranking system; leak-based reporting on Google’s 2024 API leak described at least seven internal PageRank-related fields, and practitioners (this author included) believe PageRank also factors into crawl budgetThe number of URLs an engine will crawl in a timeframe. and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. — though Google hasn’t documented that link directly. Nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting has been dead since 2009; Google says 30x redirectsA redirect sends browsers and crawlers from a requested URL to a different one. An HTTP redirect specifically is a 3xx status code paired with a Location header; meta refresh and JavaScript redirects achieve a similar navigation without being a 3xx response themselves. Permanent redirects (301/308) are Google's signal the target should be canonical; temporary ones (302/303/307) aren't. no longer lose PageRank.

What PageRank actually is

PageRank is a mathematical link-analysis model, not a public toolbar score or a complete account of modern ranking. Evidence for this claim Official or primary documentation supporting the adjacent article claim, with scope limited to the source's published description. Scope: No ranking guarantee or undisclosed system mechanics are inferred beyond the cited source. Confidence: high · Verified: 2026-07-14 Stanford: PageRank paper Google’s current documentation confirms continued use without disclosing live weights or per-page values. Evidence for this claim Official or primary documentation supporting the adjacent article claim, with scope limited to the source's published description. Scope: No ranking guarantee or undisclosed system mechanics are inferred beyond the cited source. Confidence: high · Verified: 2026-07-14 Google: Ranking systems guide

PageRank is a link-analysis algorithm Larry Page and Sergey Brin developed at Stanford between 1996 and 1998. The name is a double meaning — it’s named after Larry Page, and it ranks web pages. It treats the web as a graph and inbound hyperlinks as votes, then assigns every page a score representing the probability that a random surfer — someone clicking links at random, occasionally jumping to a random page — ends up on that page.

What follows is the published, historical model from the 1999 Stanford paper — it’s the model Google has confirmed evolved since (see “Does Google still use PageRank?” below), so treat the formula and the worked example as how the original public algorithm worked, not a live description of Google’s current internal computation.

The formula from the original paper:

PR(A) = (1-d)/N + d * (PR(B)/L(B) + PR(C)/L(C) + PR(D)/L(D) + …)

• d — the damping factor, typically 0.85: the probability the surfer keeps clicking rather than jumping to a random page.
• N — total number of pages.
• L(x) — number of outbound links on page x.
• PR(x) — the PageRank of page x.

A few things fall out of that formula. Your PageRank depends on the PageRank of the pages linking to you — which is why it’s recursive and has to be computed iteratively until the numbers converge. Each linking page divides its PageRank across all its outbound links, so every extra outbound link dilutes the share each link passes. And the damping factor — 0.85 — models a 15% chance at each step that the surfer abandons the chain and starts fresh, which is what keeps the math from collapsing into “rank sinks” (pages with no outbound links).

A simplified worked example (illustrative only — a toy graph to show the mechanics, not a real site or a Google computation): say pages B, C, and D each link only to page A, and each of B, C, and D has a PageRank of 1 with one outbound link apiece (so each passes its full score to A). With N = 4 and d = 0.85:

PR(A) = (1 - 0.85)/4 + 0.85 * (1/1 + 1/1 + 1/1) = 0.0375 + 2.55 = 2.5875

That single pass shows the mechanics — every linking page’s score, divided by its outbound-link count, feeds A’s score — but a real computation iterates this across every page on the web until the scores stop changing, which is what “converges” means below.

One nuance almost nobody mentions, and that I dug into in The Evolution Of Google PageRank: the formula as published in the original paper contains a mathematical error. The leading constant is written as (1-d) when it should be (1-d)/N — divided by the total number of pages — for the probabilities to actually sum to 1. It’s a small thing, but it’s a fun reminder that even foundational papers ship with bugs.

Not all links are equal — the Reasonable Surfer

The original model assumed a random surfer clicking every link with equal probability. The Reasonable Surfer patent (filed 2004, granted 2010, updated 2016) describes a different approach: a surfer who is more likely to click some links than others, so those links would pass more PageRank. A patent describes what Google claimed as a method it could use — it isn’t proof of the exact weights or logic running in production today, and Google hasn’t published a current, itemized list of what “counts.” Treat the features below as the patent’s described possibilities, not verified production weights:

• Link position — main content beats sidebar beats footer.
• Anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. — longer, more informative, more relevant anchors.
• Visual prominence — font size, styling, above- vs. below-the-fold.
• Link target — topical alignment between source and destination.
• Actual click data — how often users really click that link.

The practical takeaway, based on what the patent describes: a link in the body of a high-authority page is plausibly worth more than the same link stuffed in that page’s footer. Pursue editorial, in-content links — the kind a real person would actually click — as a reasonable hedge even without confirmed current weights.

A short history

• 1996–1998 — developed at Stanford. (Worth noting for accuracy: Robin Li, later Baidu’s founder, patented a similar link-ranking idea, RankDex, in 1996.)
• 1998 — patent filed (January 9, 1998); the paper follows.
• ~2000 — the Google Toolbar exposes a public 0–10 score (logarithmic scale).
• ~2006 — the original algorithm is replaced internally with a faster, less resource-intensive version.
• 2009 — Google changes how nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. works, killing PageRank sculpting (see Myths).
• 2013 — last public Toolbar PageRank update (December).
• 2016 — Toolbar PageRank removed entirely; Gary Illyes also confirms 30x redirects no longer lose PageRank.
• 2017 — Illyes confirms it’s still in use: “after 18 years we’re still using PageRank.”
• 2019 — the original patent expires (September 24, 2019).
• 2024 — the Google API leak reveals at least seven PageRank variants; the updated SEO Starter Guide confirms PageRank is still active.

Does Google still use PageRank? Yes.

This is the question people actually ask, and the evidence is unambiguous. Google’s ranking systems guide says PageRank is “one of our core ranking systems used when Google first launched” and that “it continues to be part of our core ranking systems.” Removing the Toolbar score in 2016 removed a display, not the algorithm.

In May 2024, an anonymous source leaked thousands of internal Google Search API documentation files; SparkToro published the original report and Search Engine Land followed with further analysis. Google never confirmed or explained the leaked field names — what follows is the industry’s interpretation of what those fields are likely to be, not a Google-published specification:

The same leak coverage described a link quality tier system (low / medium / high) that reportedly determines whether PageRank and anchor signals flow through a link at all, with click data (NavBoost) feeding into which tier a link lands in — again, this is leak-based reporting, not something Google has documented directly. If accurate, a low-tier link may pass nothing. For my own read on the canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. side of this, see my canonicalization guide — that piece is practitioner analysis, not a Google specification either.

Beyond ranking — PageRank’s other jobs (practitioner analysis)

PageRank probably isn’t only a ranking input, but I want to be precise about what’s confirmed and what isn’t: Google’s own documentation doesn’t explicitly name PageRank as a crawl-budget or canonicalization input. What follows is practitioner analysis — mine, in The Evolution Of Google PageRank — inferred from Google’s crawl-demand and canonicalization documentation plus general statements about page importance, not a direct Google confirmation:

• A likely crawl-budget factor. Google’s crawl-demand documentation ties crawlingCrawling is how search engines use automated bots (like Googlebot and Bingbot) to discover URLs and download pages. A page has to be crawlable to be indexed, but crawling on its own isn't a ranking factor. frequency to a page’s perceived importance/popularity in its indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., without naming PageRank specifically; the common practitioner reading is that more important pages — higher PageRank — get crawled more frequently. This connects to crawl budgetThe number of URLs an engine will crawl in a timeframe. and crawl frequencyCrawl frequency is how often a search engine comes back to re-fetch a page it already knows about. Popular pages that change often get refreshed many times a day; stable pages can go weeks or months between crawls — and you influence it indirectly, not by setting a dial. elsewhere in this cluster.
• A plausible canonicalization signal. When Google picks a canonical from a cluster of duplicates, it weighs roughly 40 signals it hasn’t fully enumerated publicly; link authority is understood to be one input, and the practitioner inference is that higher-PageRank pages are more likely to be chosen as the canonical. That inference isn’t independently confirmed by Google.

”Links are less important now” — the nuance

In 2023, Gary Illyes said links are not in his top three ranking factors and “hasn’t been for some time.” At SERPconf 2024 he went further — “We need very few links to rank pages” — and then, as I noted from the room, immediately walked it back (“I definitely shouldn’t have said that”). John Mueller has similarly said the weight on links will drop off over time and that you shouldn’t over-focus on link counts.

The honest read: links matter less than they used to for well-established sites, but they’re still foundational — for establishing authority, for crawl priority, and for canonicalization. “Less decisive” is not “irrelevant.”

What this means for your SEO

• Earn links from genuinely authoritative, on-topic pages. The recursive math means one link from a strong, relevant page beats many from weak ones.
• Chase editorial, in-content links. Reasonable Surfer logic: pursue links a real user would actually click, in the body, not the footer.
• Treat internal linkingAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. as PageRank plumbing. Your architecture decides how PageRank flows to your money pages — keep important pages well-linked and shallow.
• Don’t sculpt with nofollow. It’s been broken since 2009 — nofollowed PageRank evaporates rather than redistributing.
• Don’t fear 30x redirects. No PageRank loss since 2016.
• Avoid link schemes. Low-quality links can be tiered “low” and pass no signal at all — wasted effort at best.

Where this fits

PageRank is why crawlingCrawling is how search engines use automated bots (like Googlebot and Bingbot) to discover URLs and download pages. A page has to be crawlable to be indexed, but crawling on its own isn't a ranking factor. and linking matter in the first place, which is why it sits in the crawling sub-cluster. For the discovery and fetch side, see crawling, crawl budget, and crawl frequencyCrawl frequency is how often a search engine comes back to re-fetch a page it already knows about. Popular pages that change often get refreshed many times a day; stable pages can go weeks or months between crawls — and you influence it indirectly, not by setting a dial.; for how it shapes which URL Google keeps, see canonicalization. For the broader picture, see How Search WorksSearch works in three stages — crawling, indexing, and serving (ranking). A page has to clear each one to appear in results: getting crawled doesn't mean you're indexed, and getting indexed doesn't mean you rank..

AI summary

A condensed take on the Advanced version:

• PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. = a recursive link-graph algorithm (Page & Brin, Stanford, 1996–1998). Named after Larry Page. A page’s score is the probability a “random surfer” lands on it, and it depends on the scores of the pages linking to it — “links are votes” is the simplified version of that.
• The published, historical formula: PR(A) = (1-d)/N + d * Σ(PR(x)/L(x)). Damping factor d ≈ 0.85. In this model, each page splits its PageRank across its outbound links, so more outbound links = less per link. It’s recursive and computed iteratively — this describes the original public algorithm, not a confirmed description of Google’s current internal computation.
• The published paper has a math error — the constant should be (1-d)/N, not (1-d).
• Reasonable Surfer patent (2004/2010): describes a model where links could be weighted by position, anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page., visual prominence, topic, and real click data — body links plausibly outrank footer links. It’s a patent claim, not confirmed production weights.
• History: Toolbar 0–10 score launched ~2000, original algorithm replaced ~2006, last public update 2013, Toolbar removed 2016, patent expired 2019.
• Still active: Google’s ranking systems guide lists it as a core ranking system that’s evolved since launch. Leak-based reporting on the 2024 API leak described ≥7 internal fields (rawPagerank, pageRank2, pageRank_NS, firstCoveragePageRank, ToolBarPageRank) plus a low/medium/high link-quality tier system — reporters’ interpretation of leaked names, not a Google-confirmed spec.
• Practitioner analysis links it to crawl budgetThe number of URLs an engine will crawl in a timeframe. (higher PR → crawled more, plausibly) and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. (one of ~40 signals) — inferred from Google’s documentation, not stated by Google directly.
• Practical: earn authoritative, on-topic, in-content links; treat internal linkingAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. as PageRank plumbing; nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting is dead (2009); Google says 30x redirectsA redirect sends browsers and crawlers from a requested URL to a different one. An HTTP redirect specifically is a 3xx status code paired with a Location header; meta refresh and JavaScript redirects achieve a similar navigation without being a 3xx response themselves. Permanent redirects (301/308) are Google's signal the target should be canonical; temporary ones (302/303/307) aren't. don’t lose PageRank (2016); no public score — third-party metrics like Ahrefs URL Rating are proxies, not PageRank, and guarantee nothing.

Official documentation

Primary-source documentation and original references.

Google

• Ranking Systems Guide — lists PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. as a “core ranking system” that “continues to be part of our core ranking systems.”
• How Search Works — Ranking results — links from prominent sites as a quality/trust signal.
• Optimize your crawl budget — the crawl-capacity + crawl-demand model PageRank feeds into.
• Matt Cutts — PageRank sculpting — why nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting doesn’t work after the 2009 change.
• Matt Cutts — More info on PageRank.

The original research

• PageRank — Wikipedia — the formula, history, and references to “The PageRank Citation Ranking: Bringing Order to the Web” (Page, Brin, Motwani, Winograd, 1999).
• Stanford CS54N handout — the Google PageRank algorithm — a clean walkthrough of the math.

The Reasonable Surfer patent

• Google’s Reasonable Surfer (SEO by the Sea) — Bill Slawski’s analysis of US 7,716,225.

Quotes from the source

On-the-record statements from Google. Each link jumps to the quoted passage where the source page supports it.

Google — PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. is still a core ranking system

• “Among these is PageRank, one of our core ranking systems used when Google first launched… How PageRank works has evolved a lot since then, and it continues to be part of our core ranking systems.” — Google Search Central, Ranking Systems Guide. Jump to quote

Google — links as a quality signal

• “One of the factors used to determine quality is understanding if other prominent websites link or refer to the content. This is generally a good sign that the information is trustworthy.” — Google, How Search WorksSearch works in three stages — crawling, indexing, and serving (ranking). A page has to clear each one to appear in results: getting crawled doesn't mean you're indexed, and getting indexed doesn't mean you rank.. Jump to quote

Gary Illyes, Google — still in use after 18 years

• “DYK that after 18 years we’re still using PageRank (and 100s of other signals) in ranking?” Read the post (tweet renders via JavaScript and resists automated checking; reproduced from Search Engine Land’s verbatim coverage — confirm against the live post.)

Gary Illyes, Google — 30x redirectsA redirect sends browsers and crawlers from a requested URL to a different one. An HTTP redirect specifically is a 3xx status code paired with a Location header; meta refresh and JavaScript redirects achieve a similar navigation without being a 3xx response themselves. Permanent redirects (301/308) are Google's signal the target should be canonical; temporary ones (302/303/307) aren't.

• “30x redirects don’t lose PageRank anymore.” Read the coverage (relayed via Search Engine Land’s reproduction of Illyes’ 2016 statement.)

Gary Illyes, Google — links not in the top three

• “I think they [links] are important, but I think people overestimate the importance of links… I don’t agree it’s in the top three. It hasn’t been for some time.” Read the coverage (relayed via Search Engine Land.)

Matt Cutts, Google — let PageRank flow

• “I wouldn’t recommend it, because it isn’t the most effective way to utilize your PageRank… I would let PageRank flow freely within your site.” Read the post

John Mueller, Google — don’t fixate on link counts

• “My recommendation would be not to focus so much on the absolute count of links… over-focusing on links will often result in wasting time on things that don’t improve your site.” Read the coverage (relayed via Search Engine Land.)

Internal link-equity audit

You can’t control who links to you, but you fully control how PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. flows inside your own site. Run this pass whenever you launch, restructure, or reprioritize pages:

• Every priority (“money”) page is reachable from somewhere crawlable — no orphan pages with zero internal linksAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. pointing to them.
• Priority pages sit shallow in the architecture (a handful of clicks from the homepage, not buried ten levels deep in a category tree).
• Priority pages get links from genuinely high-authority pages on your own site — not just a single footer or sidebar mention.
• Links to priority pages appear in main body content, not only in nav chrome — Reasonable Surfer weights in-content links higher.
• Anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. on internal linksAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. is descriptive, not generic (“click here”, “read more”).
• nofollow isn’t being used to try to “sculpt” internal PageRank onto other links — that’s been broken since 2009; use it only for genuinely untrusted or paid links.
• Internal redirect chainsA → B → C instead of A → C. Each hop loses link equity and adds latency. are flattened to a single hop (each extra hop is another place for signal to get lost or delayed).
• Canonical tagsA rel=\"canonical\" annotation — in the HTML <head> or an HTTP Link header — that tells search engines which URL is the preferred version of duplicate or near-duplicate content. on near-duplicateThe same or very similar primary content reachable at more than one URL. There's no general duplicate content penalty — the real costs are possible signal dilution, the wrong URL getting chosen, and less-efficient crawling. pages point to the version you actually want to accumulate authority — don’t split link equity across near-duplicates with no canonical signal.
• XML sitemapAn XML sitemap is a UTF-8 file listing the canonical URLs on your site (with optional lastmod) so search engines can discover and prioritize them. It's a discovery and diagnostic aid, not a guarantee of indexing — and Google ignores its priority and changefreq tags. and internal nav don’t route equity toward low-value pages (internal search results, filtered/faceted URLs, thin tag pages) at the expense of pages you want to rank.

The mental models

1. Links are weighted votes — and the pie is finite. A page’s PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. is split across its outbound links. A link from a strong page passes a big slice; that same page linking to 200 things passes thin slivers. Quality of the linking page and how many links it has both matter.

2. It’s recursive, so authority compounds. Your PageRank depends on the PageRank of the pages linking to you, which depends on the pages linking to them. This is why a single link from a genuinely authoritative, on-topic page can outweigh dozens of weak ones.

3. The Reasonable Surfer — not all links on a page are equal. Weight by likely clicks: main-content > sidebar > footer; informative anchors > generic; prominent > buried. Earn the links a real person would actually click.

4. PageRank probably does more than one job. Ranking is confirmed by Google; crawl budgetThe number of URLs an engine will crawl in a timeframe. (higher PR → crawled more) and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. (higher-PR pages favored as canonical) are practitioner inferences from Google’s other documentation, not things Google states about PageRank directly. Improving a page’s authority plausibly moves all three.

5. Internal linkingAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. is PageRank plumbing. You can’t change who links to you overnight, but you fully control how PageRank flows inside your site. Keep important pages shallow and well-linked; don’t strand them.

PageRank cheat sheet

The formula

PR(A) = (1-d)/N + d * ( PR(B)/L(B) + PR(C)/L(C) + … ) — d (damping) ≈ 0.85, N = total pages, L(x) = outbound links on page x. (The original paper printed the constant as (1-d); it should be (1-d)/N.)

The 2024 leak — PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. variants

(At least seven variants total were referenced.)

Fast facts

• Named after Larry Page (pun on web pages).
• Public Toolbar score: 0–10, logarithmic — last updated Dec 2013, removed 2016.
• Original algorithm replaced internally ~2006.
• Original patent expired Sept 24, 2019.
• Still a core ranking system per Google’s ranking systems guide.
• 30x redirectsA redirect sends browsers and crawlers from a requested URL to a different one. An HTTP redirect specifically is a 3xx status code paired with a Location header; meta refresh and JavaScript redirects achieve a similar navigation without being a 3xx response themselves. Permanent redirects (301/308) are Google's signal the target should be canonical; temporary ones (302/303/307) aren't. don’t lose PageRank (2016).
• Nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting is dead (2009) — nofollowed PageRank evaporates.

No public score — third-party proxies: Ahrefs URL Rating (page-level, designed to approximate PageRank) and Domain Rating (site-level), Moz Domain Authority, Majestic Trust Flow / Citation Flow. These are vendor models built on each vendor’s own link indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., not measurements of Google’s actual PageRank, and none are used by Google or guarantee any crawlingCrawling is how search engines use automated bots (like Googlebot and Bingbot) to discover URLs and download pages. A page has to be crawlable to be indexed, but crawling on its own isn't a ranking factor., indexingStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., ranking, traffic, or AI-citation outcome.

PageRank myths, debunked

Myth: “Google no longer uses PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems..” Google removed the public display in 2016, not the algorithm. Its ranking systems guide still lists PageRank as a core ranking system, and leak-based reporting on the 2024 API leak described at least seven internal PageRank-related fields — though Google itself hasn’t confirmed or explained those fields.

Myth: “Nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting conserves link equity.” Dead since 2009. Nofollowed links don’t redirectA redirect sends browsers and crawlers from a requested URL to a different one. An HTTP redirect specifically is a 3xx status code paired with a Location header; meta refresh and JavaScript redirects achieve a similar navigation without being a 3xx response themselves. Permanent redirects (301/308) are Google's signal the target should be canonical; temporary ones (302/303/307) aren't. their PageRank to other links — it evaporates. Use nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. for untrusted/paid/UGC links, not as an internal PageRank lever.

Myth: “301 redirectsA 301 redirect is the HTTP status code for a permanent move: it tells browsers and search engines a URL has moved for good, and it's the strongest signal for consolidating a page's ranking signals onto the new URL. Google says permanent redirects don't cause a loss in PageRank. lose PageRank.” Not anymore. Gary Illyes confirmed in 2016 that 30x redirects don’t lose PageRank.

Myth: “Every link on a page passes equal value.” The Reasonable Surfer model weights links by position, anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page., prominence, and click data. A body link beats a footer link.

Myth: “More links always means more PageRank passed.” Reporting on the 2024 leak described link quality tiers (low/medium/high) — not confirmed by Google. If accurate, low-tier links may pass nothing, with click data influencing the tier.

Myth: “PageRank is the same as Domain Rating / Domain Authority.” Those are third-party proxies built because no public PageRank exists. Ahrefs URL Rating is designed to approximate PageRank, but none of these metrics are used by Google.

Myth: “PageRank only matters for ranking.” Practitioner analysis (mine included) also links it to crawl budgetThe number of URLs an engine will crawl in a timeframe. (higher PR → crawled more often, plausibly) and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. (higher-PR pages plausibly favored as canonical) — reasonable inferences from Google’s crawl and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. documentation, but not something Google states about PageRank directly.

Myth: “PageRank is a 0–10 score.” That was the public Toolbar (logarithmic). Internal PageRank is a probability distribution; even the leaked pageRank_NS uses its own scale.

Tools that approximate PageRank

There’s no public PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. score, so these are proxies — useful for relative comparison, not literal PageRank:

• Ahrefs URL Rating (UR) — page-level link strength on a 0–100 scale, designed to approximate page-level PageRank using Ahrefs’ own link indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed..
• Ahrefs Domain Rating (DR) — site-level link strength.
• Moz Domain Authority (DA) — Moz’s site-level link-based score.
• Majestic Trust Flow / Citation Flow — quality (TF) and quantity (CF) of links.
• Internal-link tools (Ahrefs Site Audit, Screaming Frog) — map how link equity flows inside your site so you can spot under-linked money pages and orphan pagesAn orphan page is a page on your site that no other page links to internally. Because crawlers discover pages by following links, an orphan page is effectively invisible to search engines unless it's in an XML sitemap or linked from an external site..

Remember: none of these are Google’s PageRank, none are used by Google, and none of them guarantee crawlingCrawling is how search engines use automated bots (like Googlebot and Bingbot) to discover URLs and download pages. A page has to be crawlable to be indexed, but crawling on its own isn't a ranking factor., indexingStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., rankings, traffic, or AI-search citations. They’re vendor models built on each vendor’s own link index — useful for relative, within-tool comparison, not a stand-in for Google’s actual, unpublished PageRank value.

The standing KPI for internal link equity

There’s no way to measure Google’s actual PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. — no public score, no API. So the honest standing KPI here isn’t “PageRank” itself; it’s how well your internal linking is distributing the authority you do have toward the pages you want to rank. Track the distribution, not a fake absolute number.

Orphan and under-linked priority pages

• Metric — Count of priority pages with zero, or very few, internal linksAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. pointing to them.
• What it tells you — An orphan or thinly-linked priority page can’t receive much internal PageRank no matter how strong your site’s overall authority is — it’s structurally starved.
• How to pull it — Crawl the site with Screaming Frog or Ahrefs Site Audit and cross-reference the crawl against your XML sitemapAn XML sitemap is a UTF-8 file listing the canonical URLs on your site (with optional lastmod) so search engines can discover and prioritize them. It's a discovery and diagnostic aid, not a guarantee of indexing — and Google ignores its priority and changefreq tags. and page list; both tools flag orphan pagesAn orphan page is a page on your site that no other page links to internally. Because crawlers discover pages by following links, an orphan page is effectively invisible to search engines unless it's in an XML sitemap or linked from an external site. and report an “Inlinks” count per URL.
• Benchmark / realistic range — No universal target — it depends entirely on your site size and architecture. Establish your own baseline (how many internal linksAn internal link is a hyperlink from one page on a website to another page on the same website. Internal links help search engines discover your pages and pass ranking signals (PageRank and anchor-text context) between them. your current top-ranking pages get) and treat zero or near-zero as the failure state to fix first.
• Cadence — After every migration, launch, or nav/template change; quarterly otherwise.

Click depth to priority pages

• Metric — How many clicks a priority page sits from the homepage in your site’s internal link graph.
• What it tells you — Click depthCrawl depth usually means click depth — how many clicks it takes to reach a page from the homepage by following internal links. It can also mean a crawler setting that limits how many levels deep a crawl goes before it stops. is a practical proxy for PageRank flow — pages closer to the homepage and other high-authority hubs generally receive more of the authority those pages pass along.
• How to pull it — Screaming Frog’s crawl-depth report, or Ahrefs Site Audit’s “Page Depth” distribution.
• Benchmark / realistic range — Situational, not a fixed number — shallower is generally better, but the right depth depends on your site’s size and category structure. Compare priority pages against your site’s own median depth rather than an invented industry figure.
• Cadence — Quarterly, and immediately after any restructuring of nav or categories.

Internal link-equity proxy trend (URL Rating)

• Metric — Ahrefs URL Rating (UR) for your priority pages over time — a third-party proxy designed to approximate page-level PageRank (not a measurement of it), and one that responds to internal linkingLinks between pages on the same site. changes as well as external links.
• What it tells you — A rising UR on a priority page, with no new external links, is a signal that internal link changes are successfully routing more authority to it. A flat or falling UR despite internal-link work suggests the changes aren’t reaching that page.
• How to pull it — Ahrefs Site Explorer or Site Audit, tracked per-URL.
• Benchmark / realistic range — No fixed target — UR is a relative, site-specific proxy, not Google’s real PageRank. Track your own priority pages’ trend line rather than chasing a round number.
• Cadence — Monthly is enough; UR doesn’t move fast enough to justify checking more often.

Test yourself: PageRank

Five quick questions on how PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. works and where it stands today. Pick an answer for each, then check.

Resources worth your time

My PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. writing

• The Evolution Of Google PageRank — my definitive piece: the formula error, the ~2006 replacement, PageRank as a crawl-budget and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. signal, and URL Rating as the modern proxy.
• Canonicalization — where PageRank shows up as one of ~40 canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. signals.

My speaking

• I shared the PageRank deep dive on LinkedIn — the current state of Google PageRank.

From around the industry

• The PageRank Citation Ranking: Bringing Order to the Web — Wikipedia overview — the original Page/Brin/Motwani/Winograd paper and the formula.
• Google’s Reasonable Surfer (SEO by the Sea) — Bill Slawski on which link features change a link’s value.
• An anonymous source shared thousands of leaked Google Search API documents (SparkToro) — Rand Fishkin’s original report on the 2024 leak.
• Unpacking Google’s massive search documentation leak (Search Engine Land) — the PageRank variants and link-tier findings.
• Matt Cutts — PageRank sculpting — why nofollowrel=\"nofollow\" is a value of the HTML link rel attribute that tells search engines you don't vouch for a linked page and don't want to pass ranking signals to it. Since 2019–2020 Google treats it as a hint, not a directive — and it does not reliably block crawling or indexing. sculpting stopped working after 2009.
• Are backlinks still important for SEO? (Search Engine Land) — the Illyes/Mueller statements on links being overestimated.
• Stanford CS54N — the Google PageRank algorithm (PDF) — a clean walkthrough of the math.

PageRank

PageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems.

Related: Crawling, Canonicalization, Crawl Budget

PageRank

PageRank is the recursive link-graph algorithm Larry Page and Sergey Brin invented at Stanford in 1996–1998 — named after Larry Page, with a deliberate pun on “web pages.” It scores every page on the web by modeling a “random surfer” who clicks links at random and occasionally jumps to a random page; a page’s score is the probability that surfer lands on it, and that score depends recursively on the scores of the pages linking to it. In plain terms: links are votes, votes from important pages count more, and in the published model each page splits its voting power across all the links on it.

The original public face of PageRank — the 0–10 Toolbar score — stopped updating in December 2013 and was removed entirely in 2016, which led many people to assume the algorithm was gone. It isn’t. Google lists PageRank in its ranking systems guide as “one of our core ranking systems” that has “evolved a lot” and “continues to be part of our core ranking systems.” Leak-based reporting on the 2024 Google API leak described at least seven internal PageRank-related fields — Google itself hasn’t confirmed or explained what those fields do.

Beyond ranking, practitioner analysis (drawing on Google’s crawl and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. documentation, not a direct Google statement about PageRank) links it to crawl budgetThe number of URLs an engine will crawl in a timeframe. (higher-PageRank pages plausibly get crawled more often) and canonicalizationHow search engines pick one canonical URL among duplicates and consolidate signals onto it. (higher-PageRank pages plausibly more likely to be chosen as the canonical in a duplicate cluster). The Reasonable Surfer patent describes a model where links could be weighted by position, anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page., and likely clicks — a patent claim, not a confirmed current weighting. There’s no public PageRank score anymore — third-party proxies like Ahrefs’ URL Rating approximate it but are not used by Google and guarantee no outcome.

Related: Crawling, Canonicalization, Crawl Budget

Changelog