Key takeaways
The Hilltop algorithm, created in 1999 by Krishna Bharat and George A. Mihaila, ranks a web page by the number and quality of links it receives from "expert" pages: documents that link out to many independent resources on a single topic. A page becomes an "authority" only when several non-affiliated experts on the query topic point to it. Google hired Bharat in 1999 and the algorithm's ideas are widely reported to have informed Google News around 2003, making Hilltop a direct ancestor of today's topical-authority thinking.
- Built by Krishna Bharat (then at DEC/Compaq Systems Research Center) and George A. Mihaila (University of Toronto), with the foundational patent filed October 15, 1999.
- Ranks a target page by links from non-affiliated expert pages that are themselves about the query topic, not by raw link count alone.
- An expert page must link out to at least 5 distinct non-affiliated hosts; a target needs links from at least 2 non-affiliated experts to be scored.
- Google hired co-creator Krishna Bharat in 1999; Hilltop's ideas are widely reported to have informed Google News around 2003, though the patent (US7346604B1) was never owned by Google.
- Hilltop introduced topical, query-dependent link authority and is the conceptual root of modern topical authority and E-E-A-T.
What the Hilltop algorithm is
Definition
The Hilltop algorithm is a link-analysis ranking method that measures the topical authority of a web page rather than its raw popularity. It works from a special index of so-called expert documents: pages that link out to many independent resources on a specific subject. When several of these topic experts point at the same page, Hilltop treats that page as an authority on the topic and ranks it higher.
The core principle, stated in the original work, is blunt: a page is an authority on a query topic if and only if some of the best experts on that topic link to it. This made link authority query-dependent and topic-aware, a sharp contrast to the query-independent popularity score that PageRank computes once for the whole web. Hilltop is one strand in the wider story of Google's ranking algorithms, which our pillar guide ties together end to end.
Hilltop at a glance
- Created by
- Krishna Bharat and George A. Mihaila
- Origin
- DEC/Compaq lab and University of Toronto, 1999
- Patent priority
- October 15, 1999
- Patent number
- US7346604B1, granted 2008, assigned to DEC/Compaq/HP (not Google)
- Expert threshold
- At least 5 distinct non-affiliated hosts
- Target threshold
- At least 2 non-affiliated experts
- Google connection
- Bharat joined Google in 1999; ideas reported in Google News around 2003
- Core idea
- Topical authority from independent experts
Who created Hilltop and when
Hilltop was developed by Krishna Bharat and George A. Mihaila. Bharat was working at the DEC Systems Research Center (DEC was acquired by Compaq in 1998, so sources also cite the Compaq Systems Research Center) and Mihaila was at the University of Toronto. The foundational patent carries a priority date of October 15, 1999.
The pair published the work in two well-known papers: Hilltop: A Search Engine based on Expert Documents, circulated as a University of Toronto technical report and presented around the WWW9 conference (Amsterdam, May 2000), and When Experts Agree: Using Non-Affiliated Experts to Rank Popular Topics at WWW10 in 2001. Bharat joined Google in 1999 and later created Google News in the aftermath of the September 11, 2001 attacks; Google News launched in 2002.
How Hilltop actually works
Hilltop runs in two phases.
- Phase 1 - Expert lookup. The algorithm queries a pre-built index of expert documents and finds the experts whose titles, headings, and anchor text best match the query. Each matching expert receives a relevance score.
- Phase 2 - Target ranking. The pages those top experts link to are the candidate target documents. Each target's score is the sum of the edge scores from the experts pointing at it, where an edge score reflects the expert's own relevance multiplied by how strongly the query keywords appear in the qualifying anchor text and surrounding phrases.
Two thresholds keep the method honest. To qualify as an expert, a page must link out to at least 5 distinct non-affiliated hosts on the topic. To be ranked at all, a target must be pointed to by at least 2 non-affiliated experts. If too few independent experts agree, Hilltop returns no result and the engine can fall back to its general ranking.
The non-affiliation rule (why independence matters)
The single most important idea in Hilltop is non-affiliation. Links only count as independent endorsements if they come from genuinely separate sources. Two hosts are treated as affiliated (and their links discounted to a single vote) when they either share the same first three octets of their IP address, or their rightmost non-generic hostname token matches (for example, ibm.com and ibm.co.mx). Affiliation is transitive, so a whole network of related sites collapses into one voice.
This rule is what made Hilltop resistant to the link-farming and self-referential linking that could inflate raw PageRank. It anticipated, by years, Google's later war on link schemes that the Penguin and SpamBrain systems now enforce, and the modern emphasis on diverse, independent referring domains.
Hilltop versus PageRank and HITS
Hilltop sits alongside two other foundational link algorithms, and the differences explain why it mattered.
- PageRank (Larry Page and Sergey Brin) computes one global, query-independent importance score per page from the entire link graph. It measures popularity, not topical fit.
- HITS (Jon Kleinberg's hubs and authorities) is query-dependent like Hilltop but computes scores at query time over a subgraph, which is slower and prone to topic drift and nepotistic linking.
- Hilltop restricts the link graph to vetted, non-affiliated expert pages, making the authority signal both topical and computationally cheaper while filtering out self-promotion.
In short, PageRank asked "how popular is this page across the whole web?" Hilltop asked "do independent experts on this exact topic vouch for this page?" Our PageRank guide covers the popularity side of that contrast in full.
From Hilltop to modern topical authority
Hilltop's path to Google runs through its co-creator, not through a patent purchase. Google hired Krishna Bharat in 1999, and the expert-document approach is widely reported to have informed Google News, the product Bharat went on to build, around 2003. The patent itself (granted as US7346604B1 in 2008) was assigned to Digital Equipment Corporation and later to Compaq and Hewlett-Packard, and Google was never listed as its owner. Many SEOs also associate Hilltop's ideas with the November 2003 Florida update, which shifted ranking away from purely keyword-independent signals toward more query- and topic-sensitive evaluation.
You can draw a clean line from Hilltop to the systems that followed: Panda (2011) rewarded topically focused quality sites, Hummingbird (2013) and the Knowledge Graph pushed entity and semantic understanding picked up later by RankBrain and the BERT and MUM language models, and E-E-A-T formalized expertise, experience, authoritativeness, and trust through the Helpful Content System. Today's advice to build interlinked topic clusters and earn links from independent, on-topic sources is Hilltop's expert-and-authority logic restated for the modern web.
History of Hilltop: a timeline
Hilltop evolved from a late-1990s research idea about expert documents into a patent and a body of work whose ideas reached Google through its co-creator and shaped topic-sensitive ranking.
-
1998
DEC becomes Compaq
Compaq acquires the Digital Equipment Corporation (DEC), so Krishna Bharat's lab is cited as both the DEC and the Compaq Systems Research Center.
-
1999
Hilltop patent filed
Krishna Bharat and George A. Mihaila file the foundational patent (priority date October 15, 1999) covering ranking by hyperlinks from expert documents. Bharat joins Google the same year.
-
2000
Hilltop paper presented
Hilltop: A Search Engine based on Expert Documents is published as a University of Toronto technical report and presented around the WWW9 conference in Amsterdam.
-
2001
When Experts Agree
Bharat and Mihaila present the follow-up paper, When Experts Agree: Using Non-Affiliated Experts to Rank Popular Topics, at the WWW10 conference.
-
2002
Google News launches
Bharat builds Google News during his 20% time after the September 11 attacks; the service launches in 2002 and relies heavily on source authority.
-
2003
Hilltop linked to Google News
Hilltop's expert-document approach is widely reported to have informed Google News by around February 2003, via co-creator Krishna Bharat. The patent itself was never owned by Google. SEOs widely link Hilltop's ideas to the November 2003 Florida update.
-
2008
Patent granted
US7346604B1, Method for ranking hypertext search results by analysis of hyperlinks from expert documents and keyword scope, is granted on March 18, 2008.
The core signals Hilltop uses
Hilltop is built from a handful of clearly defined moving parts. Understanding each one shows why independent, on-topic links carry so much weight in the model.
| Signal | What it does |
|---|---|
| Expert documents | Pages that link out to many independent resources on one topic. A page qualifies as an expert only when it links to at least 5 distinct non-affiliated hosts. |
| Authority (target) pages | The pages that experts point to. A target is scored, and ranked, only when at least 2 non-affiliated experts on the query topic link to it. |
| Non-affiliation | Links count as independent votes only when hosts differ in their first three IP octets and in their rightmost non-generic hostname token. Affiliation is transitive, so related sites count once. |
| Anchor text and qualifying phrases | Edge scores depend on how strongly the query keywords appear in the anchor text and surrounding text of the expert's link, making descriptive, on-topic anchors central. |
| Query-dependent topical relevance | Unlike PageRank's single global score, Hilltop computes authority relative to the specific query topic, so a page can be authoritative for one subject and not another. |
The practical takeaway is that a few links from genuinely independent topic experts move the needle far more than a pile of links from one network or from off-topic pages.
How to optimize for Hilltop-style authority
To benefit from Hilltop's expert-and-authority logic, earn links from independent on-topic experts, diversify your referring domains, use descriptive anchor text, and build deep topic clusters that read as subject expertise.
-
Earn links from independent, on-topic resource and expert pages
Hilltop only counts links from non-affiliated experts, so a few links from genuinely separate topic authorities outweigh many links from one network.
-
Diversify referring domains rather than stacking links from sites you control
Affiliated hosts (shared IP ranges or root domains) collapse into a single vote, so a wide spread of distinct domains maximizes counted endorsements.
-
Use descriptive, topic-relevant anchor text when you earn or place links
Hilltop's edge score weights query keywords appearing in the linking page's anchor text and surrounding phrases, so vague anchors carry less signal.
-
Build deep, interlinked topic clusters so your site reads as a subject expert
Becoming the kind of focused, link-rich resource Hilltop classifies as an expert makes your outbound endorsements and inbound authority both more credible.
-
Pursue links from established niche hubs and resource lists, not generic high-traffic pages
Pages that curate many independent resources on one topic are exactly the expert documents Hilltop trusts most when distributing authority.
Hilltop myths vs. reality
Hilltop is old enough that a lot of half-truths circulate about it. Here are the most common myths and what is actually true.
Myth Google still runs the original Hilltop algorithm as a standalone ranking factor.
Reality Hilltop's ideas were folded into Google's broader systems (notably Google News and topic-sensitive ranking). Google does not confirm running the 1999 method verbatim; its influence is conceptual, not a switch you can target directly.
Myth Hilltop just counts backlinks like an early version of PageRank.
Reality Hilltop counts only links from non-affiliated expert pages relevant to the query, and discounts links from affiliated hosts. It is a topical, query-dependent signal, not a raw popularity count.
Myth Any high-authority site linking to you helps under Hilltop.
Reality A link helps only if the linking page is an expert on the query topic and is non-affiliated with you and the other experts. An off-topic or affiliated link contributes little or nothing.
Myth Hilltop and PageRank are the same algorithm.
Reality PageRank is a query-independent, global popularity score. Hilltop is a query-dependent, topic-scoped authority signal that filters the link graph down to independent experts first.
Frequently asked questions
Hilltop is a link-analysis algorithm that judges a page by how many independent expert pages on the same topic link to it. Instead of measuring raw popularity, it measures topical authority, ranking a page highly only when several unrelated topic experts vouch for it through their links.
Krishna Bharat and George A. Mihaila created it. Bharat was at the DEC (later Compaq) Systems Research Center and Mihaila was at the University of Toronto. The foundational patent carries a priority date of October 15, 1999, and the work was published in papers around 2000 and 2001.
An expert document is a page that links out to many independent resources on a single topic, like a curated resource list. To qualify in the original method, the page must link to at least 5 distinct non-affiliated hosts about that topic, marking it as a credible topical reference point.
PageRank computes one global importance score per page from the whole link graph, independent of any query. Hilltop is query-dependent and topical: it first finds expert pages on the specific topic, then ranks the pages those non-affiliated experts link to, so authority is measured per subject.
Indirectly. Google hired Hilltop co-creator Krishna Bharat in 1999, and Hilltop's expert-document ideas are widely reported to have informed Google News, which Bharat built and which launched in 2002. The Hilltop patent itself was never owned by Google; it stayed with Digital Equipment Corporation and its successors Compaq and Hewlett-Packard. Many SEOs also associate Hilltop's thinking with the November 2003 Florida update that made ranking more topic-sensitive.
Two hosts are affiliated if they share the same first three IP-address octets or their rightmost non-generic hostname token matches, such as ibm.com and ibm.co.mx. Affiliation is transitive. Links from affiliated hosts count as one vote, so only genuinely independent experts add authority.
Its specifics are dated, but its logic underpins modern topical authority and E-E-A-T. Earning links from independent, on-topic expert pages, diversifying referring domains, and building deep topic clusters all reflect Hilltop's principle that independent expert endorsement signals authority.
US7346604B1, titled Method for ranking hypertext search results by analysis of hyperlinks from expert documents and keyword scope, lists Bharat and Mihaila as inventors. It was filed with a priority date of October 15, 1999, and granted on March 18, 2008. It was assigned to Digital Equipment Corporation and later to Compaq and Hewlett-Packard, not to Google.
The bottom line
Bottom line
Hilltop reframed link authority as a topical question: not how popular a page is across the whole web, but whether independent experts on a specific subject vouch for it. The literal 1999 mechanism is long obsolete, yet its logic survives everywhere modern SEO talks about topical authority, diverse referring domains, and E-E-A-T. Earn links from genuinely independent on-topic experts, keep your anchor text descriptive, and build clusters deep enough that the rest of the web treats you as the expert.
References
- Hilltop algorithm - Wikipedia
- When Experts Agree: Using Non-Affiliated Experts to Rank Popular Topics (WWW10, Bharat and Mihaila)
- Hilltop: A Search Engine based on Expert Documents (University of Toronto technical report)
- US7346604B1 - Method for ranking hypertext search results by analysis of hyperlinks from expert documents and keyword scope - Google Patents
- Krishna Bharat - Wikipedia
- Authoritative Sources in a Hyperlinked Environment (HITS, Jon Kleinberg)
- Topical SEO: 7 Concepts of Link Relevance and Google Rankings - Moz