The most-linked websites in 2026 (MrKeyword web graph)

What are the most-linked websites in 2026?

The most-linked website in MrKeyword's web graph (release October 2026) is googleapis.com, linked from 17,769,679 different domains, followed by facebook.com (17,696,362), google.com (15,904,056), instagram.com (13,197,285) and googletagmanager.com (11,357,422). “Linking domains” means distinct websites with at least one link to the domain, the figure SEO tools call referring domains.

• Google and Meta own the top. Five of the top ten are Google domains (googleapis.com, google.com, googletagmanager.com, youtube.com, gstatic.com), and two more are Meta's (facebook.com and instagram.com).
• Much of the top 100 is plumbing. 36 of the 100 are domains that pages mostly load code, fonts, images, tags or widgets from, or reference in their markup, such as googleapis.com, googletagmanager.com, gstatic.com, gmpg.org, gravatar.com. gmpg.org at #8 is most likely there because many WordPress themes put a profile link to it in the page head.
• Links are concentrated. The 100th domain has 361,419 linking domains, the 1,000th 37,520 and the 10,000th 7,179. Only 5,862 domains (0.012% of our 50 million) have 10,000 or more.
• Public institutions are linked far more than their numbers suggest. .gov domains are 0.04% of the domains in our table and 5.8% of the top 1,000; .edu is 0.02% against 3.6%. nih.gov is the most-linked .gov (673,479) and harvard.edu the most-linked .edu (268,680).
• .com dominates, .de under-performs. .com is 44.52% of all domains and 54.6% of the top 1,000. .de is the second-largest ending in the table (6.06%) but only 2.1% of the top 1,000.
• Raw counts can be gamed. A band of .cn domains with thousands of linking domains each, but weak PageRank, shows why a link count alone is a poor measure of authority (details below).

The 100 most-linked websites

These are the 100 registered domains with the most linking domains in the release. PageRank position is the web graph's own ranking of the same domain by PageRank (1 = strongest), included so you can see where the two measures disagree.

What kinds of websites are in the top 100?

36 of the top 100 are mostly not destinations but services that web pages load in the background or reference in their code: script libraries, fonts, CDNs, analytics and tag managers, site-builder asset hosts, and a few markup and banner references. We labelled each of the 100 by what the domain mainly is. The labels are our judgement, so treat the counts as a guide rather than a measurement.

Two things stand out. First, website builders and blog platforms (wordpress.org, godaddy.com, shopify.com, wordpress.com and others) rank high because sites built on them often carry a link back or load files from them, and each site on its own domain counts separately. Platforms that give each user a subdomain are counted as a single domain, though: blogspot.com groups 3,957,021 hosts under one entry. Second, short-link domains such as goo.gl, youtu.be, bit.ly are in the list because people paste short links into pages.

Which domain endings dominate the top 1,000?

.com holds 546 of the top 1,000 places (54.6%), well above its 44.52% share of all the domains in our table. The table compares each ending's share of the top 1,000 with its share of all 50,000,000 domains in our table, the baseline you would expect if links were spread evenly.

The gap is widest for .gov and .edu. Both are restricted endings with few domains, and government and university sites are cited across the web, so a small number of them collect a large share of links. Four of the five country endings in the table (.uk, .de, .ru, .fr) sit below their baseline, while .jp sits above it. A likely reason is that the global top 1,000 is led by services used in every country.

The most-linked sites for each domain ending

Each list below ranks the domains ending in one suffix by their linking domains. “Overall rank” is the domain's position among all 50,000,000 domains, so you can compare across lists. A suffix group matches the end of the registered domain: .au includes .com.au, .gov.au and .edu.au, and .in includes .co.in and .gov.in.

US government: .gov

The most-linked .gov domain is nih.gov, with 673,479 linking domains (overall rank 67). The 20th, census.gov, has 82,107.

Education: .edu (mainly US universities)

The most-linked .edu domain is harvard.edu, with 268,680 linking domains (overall rank 127). The 20th, brookings.edu, has 51,456.

Organisations: .org

The most-linked .org domain is gmpg.org, with 8,223,115 linking domains (overall rank 8). The 20th, hbr.org, has 152,780.

Canada: .ca

The most-linked .ca domain is canada.ca, with 116,009 linking domains (overall rank 307). The 20th, gov.on.ca, has 20,148.

United Kingdom: .co.uk

The most-linked .co.uk domain is bbc.co.uk, with 255,769 linking domains (overall rank 134). The 20th, gamstop.co.uk, has 26,529.

India: .in

The most-linked .in domain is lnkd.in, with 78,223 linking domains (overall rank 463). The 20th, india.gov.in, has 8,396.

Germany: .de

The most-linked .de domain is google.de, with 400,087 linking domains (overall rank 89). The 20th, nrw.de, has 39,470.

Australia: .au

The most-linked .au domain is abc.net.au, with 84,181 linking domains (overall rank 426). The 20th, theaustralian.com.au, has 17,444.

For a closer look at well-known sites from this list, see the domain authority of popular websites: one profile per site with its authority score, linking domains and web-graph position. Domains in the top 100 table that have a profile are linked to it.

Why do so many .cn domains have thousands of linking domains?

Below the top 1,000, a large group of .cn domains shares a narrow band of linking-domain counts while ranking poorly on PageRank, which fits a network of sites linking to each other rather than links from across the web. Only 7 .cn domains are in the top 1,000, led by miit.gov.cn at #62, which is most likely there because Chinese websites display their ICP registration number with a link to the ministry's registration site. In the top 10,000 there are 2,486.

The count of domains should fall steadily as the number of linking domains rises. Instead it jumps in the 5,000–5,999 band, and 91% of that band ends in .cn. Their PageRank tells the same story: the median PageRank position of the .cn domains in the top 10,000 is 4,883,364, against 6,197 for the other domains in the top 10,000. Many linking domains with weak PageRank means the links come from sites that are themselves weakly linked, which is what a link network looks like.

We have not inspected these sites one by one, so we do not name them or say who runs them. The lesson for SEO is general: a raw count of linking domains is easy to inflate, which is why authority scores weight links by where they come from and why Google's spam policies treat link schemes as spam. Our guide what are backlinks covers what makes a link worth having.

Is the most-linked site also the strongest?

Mostly, but not always: 68 of the top 100 by linking domains are also in the top 100 by PageRank position. PageRank weights each link by the strength of the domain it comes from, so a few links from very strong domains can count for more than thousands from weak ones.

The clearest case in this release is internalfb.com: PageRank position 88, with only 69 linking domains. The reverse is the .cn band above: thousands of linking domains each, and a median PageRank position of 4,883,364. This is why authority scores, including Moz Domain Authority and Ahrefs Domain Rating, do not simply count links.

How was this study done?

We counted, for every domain in MrKeyword's domain-level web graph (release October 2026) that we keep, how many distinct domains link to it, and ranked the domains by that count.

What MrKeyword's web graph is

MrKeyword's web graph is a domain-level map of which websites link to which, built from public web crawl data. This release, from October 2026, combines three crawls made in August–October 2026. It has 134,979,706 domains, and the release notes of the crawl data put the links between them at 1.8 billion. Hosts are grouped into registered (pay-level) domains using the public suffix list, so blog.example.com and shop.example.com both count as example.com.

What “linking domains” (in-degree) means

In the domain-level graph, an edge from A to B means at least one crawled page on domain A links to domain B. A domain's in-degree is the number of distinct domains with such an edge to it, which is the same idea as referring domains. One link and ten thousand links from the same domain both count once. Links from a domain to itself are excluded.

How we computed it

• The release's ranks file lists all 134,979,706 domains with a PageRank and harmonic-centrality position. We keep the 50,000,000 with the best PageRank positions.
• We stream the release's edges file once and count, for each kept domain, the edges that point to it: 1,682,338,593 in all.
• Domains outside the top 50 million by PageRank are not counted. Every figure on this page, including the counts of domains with 1,000 or more linking domains and the “share of all domains” baselines, covers those 50 million domains, not the whole graph. The weak-PageRank .cn domains discussed above have a median position of 4,883,364, well inside the cut-off, but we have not checked whether any domain beyond it has 1,000 or more linking domains.
• The study query ran read-only on 31 October 2026. Endings are the last part of the registered domain. The script that reproduces it is scripts/build-data-study.mjs in our codebase.

Limits

• Crawl coverage. Only pages that were crawled can contribute links. The crawl data's release notes say 90.1 million (66.7%) of the domains are dangling, with no outgoing links in the graph, for example because they were linked to but not crawled. That leaves about 44.9 million domains whose links are counted, so googleapis.com's 17.8 million linking domains are roughly 40% of them.
• Not only clickable links. The link extraction behind the graph keeps <a href> links, <link> elements such as stylesheets, <script src>, some <meta> URLs and HTTP Link headers. That is why script and font hosts rank so high.
• No nofollow weighting. The link extraction does not read the rel attribute, so nofollow, sponsored and ugc links count the same as any other link.
• No spam filtering. We did not remove link networks or parked-domain links; the .cn band above is left in so you can see it.
• Not traffic, not rankings. Linking domains measure how widely a site is referenced, not how many people visit it or where it ranks on Google.
• Domain level. All subdomains of a registered domain count as one, so a platform with millions of user subdomains appears once.
• One snapshot. The graph reflects three crawls made in August–October 2026. We plan to refresh the study with each new release of MrKeyword's web graph.

What does this mean for your website's links?

• Count domains, not links. Ten links from one site are one linking domain. That is the figure worth watching in any backlink tool.
• Weigh where links come from. The .cn band shows that thousands of linking domains can still mean little. One link from a site that is itself well linked can be worth more.
• Earn citations from institutions. .gov, .edu and major news sites are among the most-linked on the web, so a link from one carries the weight of a well-linked site in PageRank-style scores, and is correspondingly hard to get.
• Check your own profile. Our free backlink checker lists who links to any domain, and our guide to what domain authority is explains how scores built on these links differ.

How to cite this study

Cite as: MrKeyword analysis of the MrKeyword web graph, release October 2026, queried 31 October 2026. Please link to this page. A full reference:

MrKeyword (2026). The most-linked websites in 2026: an analysis of the MrKeyword web graph, release October 2026. Queried 31 October 2026. https://mrkeyword.io/blog/most-linked-websites-2026

You are welcome to quote the tables; please name MrKeyword's web graph as the source and link to this page.

Questions and answers