Where the topical map came from: topic maps to SEO

The short answer: seven older ideas under a 2021 name

The topical map is a 2021 SEO name for ideas built up over 145 years. Library classification gave it one place per subject in 1876. Thesauri gave it one term per concept in 1959. Concept maps gave it labeled links in 1972. Ontologies, from 1993, gave it typed relations. The ISO topic maps standard gave it topics, associations and occurrences in 2000.

Link research then made authority depend on the topic, between 1997 and 2002. Content marketing named the topic cluster in 2017. Koray Tuğberk Gübür built the whole thing into an SEO plan in 2021.

You do not need any of this history to build a map. You need it to build a good one, because each ancestor solved a problem your map still has. The family tree below shows which part of what a topical map is came from where.

The best evidence that this is a real lineage, not a story told afterwards, is the founding SEO text itself. On 11 February 2021, Koray published "Importance of Topical Authority: A Semantic SEO Case Study" on Oncrawl. Step one reads: "Create a Topical Map Before Starting Publishing the First Article". The next words are "Remember ontology and taxonomy?"

The person who made the topical map an SEO standard pointed straight back at library and information science in his first sentence about it. When we pulled the US results for "history of topical maps" on 12 November 2026, none of the ten was about SEO. The rest of this lesson follows his pointer back, one ancestor at a time.

Library classification gave the map one home per subject

Library classification gave the topical map its first rule: every subject has one place, decided before the books arrive. Melvil Dewey published the scheme in Amherst, Massachusetts, in 1876, after the Amherst College Library adopted it in 1873.

The 1876 text, A Classification and Subject Index for Cataloguing and Arranging the Books and Pamphlets of a Library, says "the classification is mainly made by subjects or content". Its numbers already form a tree: 500 is science in general, 510 is mathematics and 513 is geometry. That is a pillar, a cluster and a page, 145 years early.

The Library of Congress started its own list of subjects soon after. Its Library of Congress Subject Headings "has been actively maintained since 1898". A list of subjects that one organization keeps up to date for over a century is what a topical map is for one website.

Ranganathan and the first cannibalization complaint

S. R. Ranganathan, a mathematician who became the first trained librarian of Madras University, found a flaw in Dewey. M. P. Satija's entry on Colon Classification in the ISKO Encyclopedia of Knowledge Organization gives his example. A book on "Anatomy of flowering plants" could take the number for plant anatomy or the number for the botany of flowering plants.

One book, two homes. In Ranganathan's view that "defeated the purpose of classification itself". Swap books for pages and shelves for search results. Now you have keyword cannibalization: two pages that answer the same search, so neither is clearly the one to show.

His fix came from a toy. In 1924, at Selfridge's department store in London, he watched a salesman build different toys from one Meccano kit. He conceived the Colon Classification from 1924 to 1928 and first published it in 1933. Instead of one ready-made slot per book, it built each subject from parts called facets.

In the fourth edition, in 1952, he cut the facets down to five fundamental categories: personality, matter, energy, space and time, known as PMEST. Two of his rules matter for a map:

• One basic subject: "every subject belongs to one and only one basic subject". Every page in a map belongs to one cluster first, even when it links to others.
• The wall before the picture: his Wall-Picture principle orders facets by dependency, because "there cannot be any mural without a wall". In a map, the page others depend on goes live first, which is the logic behind a publishing order.

Facets also reached shops. Birger Hjørland's ISKO entry on facet analysis cites the Circuit City website's camera filters, resolution, camera style, brand and price, as faceted classification on the web. Every store filter is a facet. Which filters deserve their own indexable page is covered in topical maps for ecommerce stores.

Thesauri gave the map one term per concept

Thesauri gave the topical map its rule that each concept gets one preferred term, with every variant pointed at it. The first full-scale, operational in-house retrieval thesaurus was built at E.I. du Pont de Nemours in Wilmington in 1959, per Stella G. Dextre Clarke's ISKO entry on the thesaurus, citing Roberts (1984).

A retrieval thesaurus is not the book of synonyms on a writer's desk. It is a controlled vocabulary for finding documents. Calvin Mooers first suggested one for retrieval in 1947. After Du Pont came the Thesaurus of Engineering and Scientific Terms in 1967, and the first ISO thesaurus standard, ISO 2788, in 1974.

The current standard, ISO 25964-1 from 2011, states the aim in one line: a thesaurus guides "the indexer and the searcher to choose the same term for the same concept". In the United States the same rules sit in ANSI/NISO Z39.19-2005 (R2010), the guidelines for controlled vocabularies.

Thesauri record three kinds of relation, each with a tag that has not changed in decades. Each one is a decision your map makes too.

The equivalence rule is the one most maps break. A list of 400 keywords often holds 40 concepts, each spelled ten ways. A thesaurus would collapse them into 40 entries with their variants listed under each. Deciding which searches are variants of one concept and which deserve their own page is the subject of when a search deserves its own page.

Concept maps gave the map labeled links and a parking lot

Concept maps gave the topical map two habits: start from one question, and write the relation on every line. Joseph Novak developed them in 1972 at Cornell, in a research program that followed how children's knowledge of science changed, built on the learning psychology of David Ausubel.

The source is Novak and Alberto Cañas's 2008 technical report for the Florida Institute for Human and Machine Cognition, The Theory Underlying Concept Maps and How to Construct and Use Them. Three parts of their method carry straight into a topical map.

• Linking words. "Words on the line, referred to as linking words or linking phrases, specify the relationship between the two concepts." Two concepts and their linking words form a proposition, a statement that can be true or false. A link on your map should be one too: "the burr grinder page links to the grinder cleaning page because burr grinders need cleaning every few weeks".
• A short list first. "Usually 15 to 25 concepts will suffice" for one map, ranked from the most general to the most specific. A topical map is bigger, but the habit holds: rank before you link.
• The parking lot. Novak and Cañas call the list of concepts waiting to be placed "a parking lot". Some concepts stay there when the map is finished, "if the mapmaker sees no good connection for these with other concepts in the map".

The parking lot is the most useful idea in this lesson for anyone with an older site. A page that connects to nothing else in your map does not belong in the map. My rule for buyers is the same one: a page without a purpose does not deserve to be on the website. Before removing one, I check whether it can merge into a page that is more relevant.

Novak and Cañas also warn about the opposite mistake. Once people learn to write good linking words, "they can see that every concept could be related to every other concept", and so "they must choose to identify the most prominent and most useful cross-links". That is the case for linking with restraint, covered in internal links inside a topical map.

The ISO topic maps standard gave the map its parts

Topic maps, published as ISO/IEC 13250 in 2000, gave the topical map its three parts: topics, the typed associations between them, and occurrences, the resources about each topic. They started in the early 1990s as a way to merge back-of-book indexes.

Who made topic maps, and when

Steve Pepper, founder of the topic map software company Ontopia, told the story in The TAO of Topic Maps, first published in June 2000. In the early 1990s, a group later known as the Davenport Group was working on how to exchange computer documentation; the same group went on to create DocBook. One of its problems was merging the indexes of different sets of documentation.

Pepper quotes the insight, which he credits to Steven R. Newcomb, one of the original developers: indexes "conform to models of the structure of the knowledge available in the materials that they index. But the models are implicit, and they are nowhere to be found!" Topic maps were the attempt to write those models down.

• 2000: ISO 13250:2000, the first standard, defined HyTM, a way of writing topic maps in SGML, the document markup language that came before XML. Its editors were Michel Biezunski, Steven R. Newcomb and Martin Bryan, per the committee's 2002 guide to the topic map standards.
• 2001: TopicMaps.Org, an independent group with Pepper as editor, wrote XTM, an XML syntax for the web. It was accepted into ISO 13250 in October 2001.
• 2006: ISO/IEC 13250-2:2006 set out the data model, edited by Lars Marius Garshol and Graham Moore. ISO's summary says it defines "the rules for merging topic maps".

Six parts of a topic map, and what each means for yours

What is a topical map shows the three parts everyone knows. Pepper's paper describes more. These six are the ones that matter to a person planning pages.

The last row is the one I would print out. Pepper writes that the goal is a one-to-one relationship between topics and subjects, "in order to ensure that all knowledge about a particular subject can be accessed via a single topic". When two topics turn out to share a subject, they are merged.

That is the oldest anti-cannibalization rule there is, written for merging indexes in the 1990s. It works the same way on a website. Two pages, one subject: keep one URL, move the best content of the other into it, and redirect the old address.

One part of the standard maps onto something SEO usually leaves unwritten: scope. Pepper's example is the word "tosca", which means Puccini's opera to his readers and "quite other and sweeter connotations" to an audience of bakers. A topic map records the context a name is valid in. A topical map does the same when it states the site's market and purpose, the source context.

Taxonomies, ontologies and SKOS took the idea onto the web

Taxonomies sort subjects into a tree, ontologies add typed relations and properties, and SKOS writes thesauri and taxonomies in a form the web can read. A good topical map sits between a thesaurus and an ontology on that line.

The word ontology came into computing with a definition that still fits a map. Thomas Gruber wrote in 1993, in A translation approach to portable ontology specifications: "An ontology is an explicit specification of a conceptualization." A topical map is that for one business, written down so somebody else can check it.

The web got its own taxonomies quickly. Maria Teresa Biagetti's ISKO entry on ontologies notes that Yahoo! ran a directory of sites by subject categories from 1994 to 2002, and that the volunteer directory DMOZ kept a taxonomy of the web from 1998 to 2017.

In May 2001, Tim Berners-Lee, James Hendler and Ora Lassila described The Semantic Web in Scientific American, a web where machines read meaning, not only text. The W3C then published SKOS, the Simple Knowledge Organization System, as a Recommendation on 18 August 2009. SKOS gives thesauri and taxonomies a shared format: preferred and alternative labels, broader, narrower and related.

Google took the same step in search on 16 May 2012. Its post Introducing the Knowledge Graph: things, not strings said the graph held "more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects". Search had moved from matching words to knowing things, which is where semantic SEO in plain words picks up the story.

Where a topical map sits on the line

Knowledge organization researchers arrange these tools by semantic richness, the number of kinds of relation they can hold. The ISKO entry cites Marcia Zeng's 2008 spectrum and Olensky's 2010 "semantic staircase", rising from simple word lists to ontologies.

Search engines made authority a property of a topic

Link research made authority depend on the topic. Jon Kleinberg's hubs and authorities, first circulated as an IBM research report in May 1997, found the best pages on a broad topic through the hub pages that link to them.

His paper, Authoritative Sources in a Hyperlinked Environment, puts it in one line: "a good hub is a page that points to many good authorities; a good authority is a page that is pointed to by many good hubs." The two make each other stronger.

In 1998, Sergey Brin and Lawrence Page described Google's first design and its PageRank in The Anatomy of a Large-Scale Hypertextual Web Search Engine. Haveliwala's 2002 abstract describes that original PageRank this way: "a single PageRank vector is computed" for the whole web. In 2002 Taher Haveliwala of Stanford proposed Topic-Sensitive PageRank: instead of one score, "a set of PageRank vectors, biased using a set of representative topics". Importance became something a page has for a topic.

Papers show what researchers designed, not what Google runs today. But they explain two habits of a good map. A pillar page is a hub inside one site, which is the subject of pillar pages. And if importance is counted per topic, a cluster whose links wander off its topic spends its strength in the wrong place.

Marketing named the cluster in 2017, and Koray named the map in 2021

The SEO words arrived last. HubSpot launched a topic cluster tool in September 2017, and Koray Tuğberk Gübür published the case study that taught SEO to plan with a topical map on 11 February 2021.

On 7 September 2017 the trade press reported HubSpot's new Content Strategy tool and its "topic-based model on creating content, called topic clustering". HubSpot's reason: "For years, the best way to solidify a spot at the top of the results was to focus on creating multiple pieces of content focused on niche, long-tail keywords." The pillar and cluster spread through content marketing from there. How a cluster differs from a full map is in topical map vs topic cluster.

Koray did not invent the words. His own article shows a "Topical Map (Topical Graph) example" from InLinks, an SEO company. What his February 2021 case study on Oncrawl did was make the map the first step of a published method, with results attached. Its page carries the publish date 11 February 2021. The longer version on his own site was first published the same day and updated in 2022. Jason Barnard, writing in Search Engine Land in 2026, says Koray "coined 'topical map' as a standard SEO deliverable".

I learned this method from Koray, and the vocabulary in this library is mostly his. His full framework, in plain words, is in Koray Tuğberk Gübür's topical map framework.

What SEO added that the libraries never had

Koray's step one does not stop at ontology and taxonomy. It goes on: "You should check Google's Knowledge Graph because for Google, things can connect to each other in a different way than according to a Dictionary or Encyclopedia." That sentence marks the break with the ancestors. Four things are new.

• Demand. A library classifies every book it owns. A map plans pages people search for, so search volume decides what gets a page.
• Money. A thesaurus has no sales page. A map puts the pages that earn money at its core and builds the rest around them.
• Order. A library catalog is complete on the day it opens. A map is published over months, so the order pages go live in is part of the plan.
• A judge. In a topic map, the author decides which relations exist. In SEO, the search engine decides which relations count, which is why Koray tells you to check how Google connects things.

Why Google still reads topical map as a land map

Google still treats "topical map" as a topographic map. When we pulled the US results on 12 November 2026, all ten were about topographic maps, such as USGS topoView and CalTopo. On the same date, every result for "history of topical maps" was about historical topographic maps.

The top results for that history search included the USGS Historical Topographic Map Collection and Esri's historical topo maps. Google's related searches offered "Topographic map". The USGS explains what those pages are about: "The distinctive characteristic of a topographic map is the use of elevation contour lines to show the shape of the Earth's surface."

That is the scope problem from the topic maps standard, live in Google today. The word means one thing to SEO people and another to everyone else, and Google follows everyone else. Pepper's "tosca" would understand.

The practical lesson is for every page on your map, not only this one. Search each target term before you build a page on it. If the results are about another meaning of the word, pick the term your buyers and Google share. Or add the word that tells them apart ("topical map SEO", "espresso grinder" rather than "grinder"). Koray's advice from 2021 says the same thing: check the results "to see which entity has been connected to which and in what way for which queries".

The borrowed ideas checklist: nine checks from 150 years of practice

Run these nine checks on your map before you publish it. Each comes from one ancestor in this lesson, and each catches a mistake that is older than SEO.

Download the borrowed ideas checklistCSV, nine rows: check, borrowed from, passes when, and an empty column for your notes

If your map fails a check, the fix is usually small: merge two rows, write a missing sentence for a link, or move a page to the parking lot. If you would rather have the whole map built and checked for you, that is the growth plan.