What is co-occurrence?
Co-occurrence is the pattern of two terms appearing near each other, again and again, across many different pages. If thousands of articles mention "Notion" in the same sentence or paragraph as "project management," a search engine or language model can learn that the two are connected, without a single hyperlink between them.
The terms can be words ("mortgage" and "interest rate"), a brand and a category ("Ahrefs" and "backlink checker"), or two entities ("Pixar" and "Toy Story"). The closer and more often they appear together, in more independent sources, the stronger the inferred relationship.
The underlying idea comes from linguistics, not SEO. The linguist J. R. Firth summed it up in 1957: "You shall know a word by the company it keeps" [1]. Meaning can be inferred from context, and context is measured by co-occurrence.
How does co-occurrence work in SEO?
In SEO, co-occurrence describes how a search engine can learn that a page, brand or term is relevant to a topic from the words around it, rather than only from links and exact keywords. Two uses matter most:
- Understanding page topics. A page about "jaguar" that also mentions "habitat," "prey" and "rainforest" is about the animal; one that mentions "dealership" and "V8" is about the car. Co-occurring terms disambiguate.
- Associating brands with categories. When a brand name repeatedly appears beside a category phrase on independent sites, that association can be learned whether or not those mentions are linked.
Google holds a patent describing the first use. Patent US 8,321,409, "Document ranking using word relationships," filed in 2011 and granted November 27, 2012, describes identifying significant terms on top-ranking pages and scoring how often they appear near each other, in the same sentence or paragraph, to help rank documents [2]. A patent shows what Google has invented, not what it uses. Google has not confirmed this patent is in production.
What is the difference between co-occurrence and co-citation?
Co-occurrence is about terms appearing together in text, while co-citation is about two items being cited or referenced together by a third source. The two are routinely mixed up in SEO writing, including in some of the pages that rank for both terms.
| Co-occurrence | Co-citation | |
|---|---|---|
| What appears together | Words, phrases, entities in the same text | Two documents or sites referenced by the same source |
| Origin | Linguistics and information retrieval | Bibliometrics, 1973 [4] |
| Example | "Canva" keeps appearing near "graphic design tool" | A blog links to both Canva and Figma, so Canva and Figma are co-cited |
| What it suggests | The term or brand is related to the topic | The two cited items are related to each other |
Co-citation has a precise academic history. Henry Small and Irina Marshakova independently introduced co-citation analysis in 1973, defining it as how often two documents are cited together by other documents [4] [5]. In SEO, co-citation is most useful for understanding link neighborhoods (sites cited alongside each other). Co-occurrence is the better term for brand and keyword associations in text.
Why do unlinked mentions work if there is no link?
Unlinked mentions can help because a search engine or AI model can read a brand name in text and associate it with the words around it, without needing a hyperlink. An unlinked brand mention on a relevant, trusted page still places the brand name beside a topic, which is co-occurrence.
Google has at least patented the idea of counting references that are not links. A Google patent granted March 25, 2014, inventors Navneet Panda and Vladimir Ofitserov, distinguishes an "express link," which a user can follow, from an "implied link," defined as "a reference to a target resource, e.g., a citation to the target resource, which is included in a source resource but is not an express link" [6].
Google staff have spoken about mentions in general terms. At Brighton SEO in September 2017, Gary Illyes said that content "highly cited on the internet," including mentions and people talking about a brand rather than only links, could help it rank [8]. Former Bing product manager Duane Forrester said at SMX West 2016 that Bing could identify unlinked mentions and use them in judging authority [8]. Neither statement explains how much weight mentions get, or how co-occurring words around them are used.
What is known versus claimed
What is documented: Google has patents describing word co-occurrence for ranking [2] and implied links [6], and Google and Bing staff have said mentions can count [8]. What is not documented: that Google uses co-occurrence as a named ranking factor today, how much any mention is worth, or whether a mention without a link carries meaningful weight compared with a link. A June 2020 Search Engine Journal review of co-citation and co-occurrence concluded that links remain more important [7]. Any figure claiming to value an unlinked mention as a fraction of a link is a guess.
When did co-occurrence become an SEO term?
Co-occurrence entered mainstream SEO vocabulary in November 2012, through a debate between Rand Fishkin of SEOmoz and patent analyst Bill Slawski [3]. The underlying ideas are far older: co-occurrence in linguistics, and co-citation in library science from 1973 [1] [4].
1957: "the company it keeps"
J. R. Firth's 1957 line, "You shall know a word by the company it keeps," became the foundation of distributional semantics, the idea that word meaning can be learned from surrounding words [1]. Modern word embedding methods such as Word2vec build directly on it [1].
1973: co-citation analysis
Henry Small published "Co-citation in the Scientific Literature: A New Measure of the Relationship Between Two Documents" in 1973 [5]. Irina Marshakova proposed the same measure independently, in Russian, the same year [4].
November 2012: the Whiteboard Friday that changed its own title
In November 2012, Rand Fishkin recorded a Whiteboard Friday titled "Prediction: Anchor Text is Dying…And Will Be Replaced by Co-citation," arguing that some pages ranked for phrases they had few or no anchor text links for [3]. Bill Slawski argued that what Fishkin described was co-occurrence, terms appearing on the same pages, not co-citation, and the video was retitled "Prediction: Anchor Text is Weakening…And May Be Replaced by Co-Occurrence" [3].
November 2012: Google's word relationships patent
In the same month, on November 27, 2012, Google was granted the "Document ranking using word relationships" patent, which Slawski covered the following day as a co-occurrence patent [2].
2014: implied links
Google's 2014 patent defining "implied links," references to a page that are not hyperlinks, gave the mentions argument a documented basis, though not proof of use [6].
Does co-occurrence matter for AI search?
Co-occurrence matters for AI search because large language models learn associations between words and entities from how often they appear together in their training text. That is Firth's distributional idea at very large scale [1]. If a brand name rarely appears alongside a category in the text a model learned from, the model is less likely to mention the brand when asked about that category.
Search Insider's practical distinction: in classic search, co-occurrence is one possible signal among many that a search engine may or may not use. In a language model, co-occurrence in training data is not optional; it is how the model's associations form. That is why brand mentions on widely read pages are a reasonable AI visibility goal even when they carry no link. No published study yet measures how many co-occurring mentions it takes to change a model's answers, and claims of a specific threshold are unsupported.
How do you check co-occurrence for your brand?
To check co-occurrence for a brand, search for pages that mention the brand and the target category together, and compare the count and quality of those pages with competitors. A simple manual method:
- Search Google for
"brand name" "category phrase"with both in quotes, and note which independent sites appear. - Repeat for the two or three competitors that AI assistants already recommend for that category.
- Ask ChatGPT, Gemini and Perplexity "What are the best [category] tools?" and note whether the brand appears and which sources are cited.
- Look at the gap: sites that mention competitors beside the category phrase but never mention the brand are the outreach list.
Result counts in Google are rough estimates, so treat them as directional, not as measurement.
When should you focus on co-occurrence?
Focus on co-occurrence when a brand wants to be associated with a category it is not yet known for, or when AI assistants recommend competitors but not the brand. In practice that means earning mentions in the same sentence as the category phrase on relevant, independent pages: reviews, comparisons, industry roundups and news coverage. Write your own pages so the brand name and the core topic appear together naturally, in headings and opening sentences. Do not stuff category phrases next to a brand name; repeated, unnatural phrasing reads as spam to people and is unlikely to fool systems trained on natural language.