Journal · Search

PageRank: what it is, and what it is costing your business

Google stopped publishing PageRank in 2016, but something by that name is still in its ranking system. The scores that replaced it are vendors' estimates, and what a weak position costs you is counted in clicks, not points.

Highland Willow · Boone, North Carolina · 31 October 2026

Ask an AI assistant how domain rank works and you get a confident answer with a tier chart. Zero to ten is a new site, eleven to thirty a small local business, somewhere past fifty is strong and the top tier is Wikipedia. We collected two such answers while researching this post and got two different charts. One said strong starts at 51; the other said 50, stopped at 70, and left 71 to 79 out altogether. Where they cited anything for the tiers, it was a YouTube video, a Reddit thread or an agency's blog.

Nobody official drew those lines. The companies that publish these scores say they are relative: Ahrefs calls its Domain Rating relative by definition, and Semrush says its Authority Score is for comparing domains, not for grading one as good or bad. That small muddle is the whole subject in miniature. The number people ask about is not the number Google uses. The number Google uses cannot be seen. And the cost of a weak position is real and countable, while the score attached to it is a vendor's estimate. This post takes them in that order.

Two things called page rank

Your page's rank is its position in the results: third, eleventh, nowhere. PageRank, one word, is a score Google's founders published in 1998 for how much of the rest of the web points at a page. The second is one input to the first, among hundreds. That is how PageRank costs you money: a weak link profile is one reason a page sits lower, and where a page sits decides who gets the click. The position is also the one you can measure for free.

What PageRank is

Sergey Brin and Larry Page described it in the 1998 paper that introduced Google. Picture a surfer who lands on a random page and keeps clicking links, never pressing back, until they get bored and jump to another random page. A page's PageRank is the share of time that surfer spends on it. A link from a page the surfer visits often is worth more than a link from a page they rarely reach, and every page's vote is split between all the pages it links to. The formula carries a damping factor the paper usually sets at 0.85, which in effect means the surfer follows a link 85% of the time and jumps somewhere random the other 15%. (One sentence of the paper describes the factor the other way round, as the chance of getting bored; the formula only works as we have put it, which is how it is normally read.)

The idea is simple and it was a good one. Instead of believing what a page says about itself, believe what other pages say about it, weighted by how much those pages are believed.

Does Google still use it?

Yes, and the evidence is better than the usual "a Googler once said".

For years Google showed a 0 to 10 PageRank in its browser toolbar. It last updated that score in December 2013 and switched the feed off in April 2016, which is when most people concluded PageRank was dead. Only the public display died. In February 2017 Google's Gary Illyes wrote that after eighteen years Google was still using PageRank, alongside hundreds of other signals.

The most recent and most specific description is in a court record. In United States v. Google, the Justice Department published notes of a February 2025 call with a long-serving Google search engineer as trial exhibit PXR0356. They describe PageRank as a single signal relating to "distance from a known good source", used as an input to Google's page quality score, which the same notes call hugely important and largely a property of the site rather than the query. They are notes of a conversation, not a specification, and they spend one line on PageRank. Read them for what they are: confirmation that something Google calls PageRank was still in the system in 2025. Whether it is computed the way the 1998 paper describes, they do not say.

The scores you can actually see

Because Google's number is private, SEO software companies built their own. They are useful, and none of them is PageRank. All three run up to 100.

Three things about these scores matter more than which tier you land in.

Google does not read them. Google's John Mueller has said so about Moz's Domain Authority repeatedly since 2016, on one occasion adding that it is a metric from an SEO company. They correlate with rankings because they are built largely from the same raw material, links, not because Google looks them up. Google does keep its own judgement of sites, the quality score in that court exhibit. It is not Moz's number, and you cannot see it either.

They are relative. Ahrefs notes that your Domain Rating can fall without you losing a single link, because other sites gained some. A DR of 30 tells you nothing until you know the DR of the sites ranking above you for the searches you care about.

They can be bought. A score built from links can be inflated with links, and there is an industry that sells exactly that. Google's spam policies name buying or selling links for ranking purposes as link spam. Paid links are allowed if they carry a nofollow or sponsored tag, which marks them as paid, the way the policy asks.

One good link or three cheap ones

The question underneath all of this is whether a link from somewhere trusted beats several from nowhere. We ran the 1998 formula on a 12-page toy web to find out, and the answer depends on which version of PageRank you run.

Your site, A, has one link, from a trade association's member directory that five member sites link to. The directory also links to one other member, C. A competitor, B, has three links, each from a page that nothing links to: the kind of throwaway page a link seller makes in bulk.

Each figure is the share of the random surfer's time spent on that page. The twelve scores in each version add up to 100%.

Under the plain formula the three cheap links win. Two are not enough; three are. That is the weakness link sellers lived on for years. Every page gets a little PageRank simply for existing, so a page that exists only to link passes that little on, and a thousand of them add up.

The same 1998 paper offered a fix: let the surfer restart only from a chosen page or group of pages, which the authors said could make it nearly impossible to mislead the ranking on purpose. Seed the surfer at the directory and the throwaway pages are never visited, so they pass nothing on. B scores zero with three of them, and still zero with thirty or three hundred.

Two choices of ours decide that result, and you should know them. We seeded the surfer at the directory itself, the very page that links to you, which is the kindest case for your site. And when the surfer reaches a page with no links of its own, we sent it back to the seed, a detail the paper does not settle. Send it anywhere at random from those dead ends instead, and the throwaway pages get visited again: with 6 of them, the competitor overtakes you even in the seeded version. So the toy shows a direction, not a guarantee. Starting from trusted pages makes cheap links much weaker; it does not make them worthless.

The 2025 court exhibit's one line points the same way, describing PageRank as distance from a known good source. It does not say how Google measures that distance, and a distance need not be a random surfer at all.

Look at what else scores zero when seeded: the five honest member sites, because nothing trusted links to them. That is the practical lesson for a small business. What counts is less how many links you have than whether there is a path to you from somewhere already trusted, and how short it is.

The core of the calculation is the function below; the toy web and the choice of seed are a few lines more. The figures on this page are imported from that script rather than typed, so they cannot drift from the arithmetic.

D = 0.85


def pagerank(links, seeds=None, d=D, dead_ends="restart", tol=1e-13):
    """links: {page: [pages it links to]} -> {page: score}, scores sum to 1.

    seeds: pages the surfer restarts on; None means every page equally.
    dead_ends: where the surfer goes from a page with no links. "restart"
    sends it where it restarts; "anywhere" sends it to any page at random.
    The 1998 paper does not say, and in the seeded web the choice decides
    whether throwaway pages can ever win.
    """
    pages = sorted(links)
    start = seeds or pages
    jump = {p: (1 / len(start) if p in start else 0.0) for p in pages}
    pr = dict(jump)
    while True:
        dangling = sum(pr[p] for p in pages if not links[p])
        if dead_ends == "restart":
            new = {p: ((1 - d) + d * dangling) * jump[p] for p in pages}
        else:
            new = {p: (1 - d) * jump[p] + d * dangling / len(pages) for p in pages}
        for p in pages:
            for q in links[p]:
                new[q] += d * pr[p] / len(links[p])
        if sum(abs(new[p] - pr[p]) for p in pages) < tol:
            return new
        pr = new

It is the published algorithm on a toy web, not Google's system. Google's actual seeds, weights and inputs are not public.

What a low ranking costs

Here is where the money is, and none of it depends on knowing a score. A business does not lose money because its Domain Rating is 12. It loses money because the people searching for what it sells click on somebody else.

The clicks are front-loaded. Backlinko analysed click-through rates from several companies' Google Search Console accounts, supplied by Semrush, and found the first organic result was clicked on an average 27.6% of the searches that showed it, about ten times the tenth result. Only 0.63% of searchers clicked anything at all on page two. Backlinko does not say when its click data was collected.

So the cost of a position can be written down. For one search term:

monthly cost =
  searches
  × the click rate you could have,
    minus the one you have now
  × share of visitors who buy
  × average order

Here it is with invented round numbers that describe nobody's business: 1,000 searches a month, 2% of visitors buying, and a $150 average order.

All figures are per month. Page two's are a ceiling: those 6.3 clicks are everything page two gets, shared among ten results.

On those figures, page two against first place is roughly $810 a month, about $9,700 a year, for a single search term. Most businesses have several terms their customers use. Put your own numbers in and the answer changes; the shape does not.

The AI Overview row is the new part. Ahrefs re-ran its study in February 2026 on 300,000 keywords, comparing December 2023 with December 2025. Where Google showed an AI Overview, the top result's click rate came out 58% below the trend Ahrefs forecast for it. Its first run, in April 2025, had put the cut at 34.5%, so the effect has been growing. Pew Research Center, which tracked the browsing of 900 American adults during March 2025, found people clicked an ordinary result on 8% of visits to a results page with an AI summary, against 15% without one. First place now buys less than it did on the searches Google answers itself. That is a reason to choose which searches to compete for, not a reason to stop.

When we pick the main search term for a page, this one included, we check Serpstat's record of whether Google shows an AI Overview for it. On 31 October 2026 it showed none for pagerank, though it did show a featured snippet, which answers part of the question on the results page too. Its two-word spelling, page rank, carried an AI Overview.

The floor is low, too. Ahrefs estimated, across about 14 billion pages in its index in 2023, that 96.55% got no traffic from Google at all, and found a clear relationship between how many sites link to a page and how much search traffic it gets.

The other way to price the gap is to ask what it would cost to rent. Multiply the clicks you are missing by what you pay per click in Google Ads. Rented clicks stop the day you stop paying; earned ones fade slowly, which is the trade our SEO ROI calculator works through month by month.

What to do with this

1. Write down the five to ten searches a customer types before they buy from you. Not your company name; the thing they need.
2. Find where you rank for them. Google Search Console is free and reports your real average position and clicks. It is the only source in this post that measures your site rather than estimating it.
3. Compare against the sites above you, not a tier chart. Look up their DR or DA. As a rough guide, if they all sit at 60 and you sit at 15, links are the gap. If they sit near you, the gap is more likely the page. We tried that comparison on three real searches in our post on domain authority.
4. Earn links along short paths from trusted places. Suppliers, trade associations, local press, the chamber of commerce, a university you work with. In the toy web, one of those was worth more than a pile of throwaway pages once the surfer started from somewhere trusted.
5. Do not buy links that pass ranking credit. It breaks Google's spam policies, which say sites that break them may rank lower or not appear at all.

What this does not establish

• The toy web shows how the published formula behaves under the two choices stated above. It does not show how much PageRank weighs in Google's ranking today. The 2025 exhibit says it feeds a quality score; it does not say how much.
• The cost table borrows from studies measured in different ways: Backlinko's click rates (dates and devices not stated), Ahrefs's AI Overview cut (desktop, against its own forecast) and Backlinko's share of searchers who reach page two. They are not on one base, so the table shows the shape of the cost, not a forecast.
• Click rates are averages across very different searches, and studies disagree. Backlinko puts first place at 27.6%; the same page cites a First Page Sage figure of 39.8%. Your own Search Console numbers beat both.
• Ahrefs measured the AI Overview effect against informational searches, on desktop data. A search with clear buying intent may behave differently, and we have not measured that.
• The 2% and $150 in the cost table are invented. They show the arithmetic; they describe no business.
• We cannot tell you your PageRank. Nobody outside Google can, including anyone who offers to.

Want help growing your online presence?

Everything above is the work we do for small businesses that want more of the right people to find them. Our site and page optimization work starts by finding exactly what is wrong with your website, ranked by what it costs you, before anybody spends money changing it. Search and content care then tracks your positions against named competitors month by month, checks term by term whether an AI answer sits above you, and reviews the sites that link to you.

We do not promise rankings, because nobody honestly can. We will show you the arithmetic, the way this post does, with your own numbers in it. If that sounds useful, tell us what you sell and who you want to reach.

Questions people ask

Does Google still use PageRank?

Yes. Google said so in 2017, and notes of a 2025 interview with a Google engineer, published as a trial exhibit in United States v. Google, describe PageRank as a signal relating to distance from a known good source that feeds Google's page quality score. What Google retired, in 2016, was the public 0 to 10 score in its toolbar.

Can I check my PageRank?

No. Google last updated the public toolbar score in December 2013 and switched it off in 2016. Since Google does not publish the number, any PageRank checker is showing you its own estimate under Google's old name.

What is a good domain authority?

One that is close to or above the sites ranking ahead of you for the searches you care about. The companies that publish these scores describe them as relative, so a fixed tier chart cannot answer the question.

Is Domain Rating a Google ranking factor?

No. Domain Rating is Ahrefs's measure of a site's link profile, and Domain Authority is Moz's. They correlate with rankings because links matter to Google, but Google does not read either score.

Is page rank the same as PageRank?

No. Your page's rank is its position in the results. PageRank is one of Google's link signals, and one input among hundreds into that position. PageRank affects the position, and the position is what costs or earns you money.

How can a small business improve its online presence?

Start with the searches your customers type before they buy, and find where you rank for each in Google Search Console, which is free. Fix the page before chasing links: say plainly, in the searcher's own words, what you sell and where. Then earn links from places related to your trade that are already trusted, such as suppliers, associations and local press. Done in that order, every step can be measured, and nobody can honestly promise you a ranking at the end of it.

Sources

1. Brin, S. and Page, L. (1998). The anatomy of a large-scale hypertextual Web search engine. Computer Networks and ISDN Systems 30, 107 to 117. Section 2.1. http://infolab.stanford.edu/~backrub/google.html
2. U.S. Department of Justice. United States v. Google LLC, 1:20-cv-03010-APM, trial exhibit PXR0356: notes of a February 18, 2025 call with a Google engineer, redacted public version. https://www.justice.gov/atr/media/1398871/dl
3. Search Engine Roundtable. Google: We're Still Using PageRank After 18 Years. 10 February 2017. https://www.seroundtable.com/google-still-using-pagerank-23392.html
4. Search Engine Roundtable. Google Fully Turns Off Feed To Toolbar PageRank. 18 April 2016. https://www.seroundtable.com/google-fully-turns-off-toolbar-pagerank-feed-21946.html
5. Ahrefs Help Center. What is Domain Rating (DR)?. https://help.ahrefs.com/en/articles/1409408-what-is-domain-rating-dr
6. Semrush Knowledge Base. Authority Score. https://www.semrush.com/kb/747-authority-score
7. Search Engine Journal. Domain Authority: Is It A Google Ranking Factor? (Googlers' statements, 2015 to 2022). https://www.searchenginejournal.com/ranking-factors/domain-authority/
8. Google Search Central. Spam policies for Google web search: link spam. https://developers.google.com/search/docs/essentials/spam-policies
9. Backlinko. Google organic click-through rate statistics. Updated 16 April 2025. https://backlinko.com/google-ctr-stats
10. Ahrefs. AI Overviews Reduce Clicks by 58%. 4 February 2026, re-running the study below. https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/
11. Ahrefs. AI Overviews Reduce Clicks by 34.5%. 17 April 2025. https://ahrefs.com/blog/ai-overviews-reduce-clicks/
12. Pew Research Center. Google users are less likely to click on links when an AI summary appears in the results. 22 July 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links/
13. Ahrefs. 96.55% of Content Gets No Traffic From Google. 2023. https://ahrefs.com/blog/search-traffic-study/