Most SEOs think about headings as places to put keywords. Put your target keyword in the H1, sprinkle related terms into H2s and H3s, and move on. That’s not wrong, but it misses what headings actually do from Google’s perspective.
There’s a Google patent that describes how the search engine uses HTML structure – headings, lists, tables, divs – to determine how semantically close terms on a page are to each other. That closeness directly affects how relevant Google considers the page for queries containing those terms. Once you understand the mechanics, it changes how you think about content organization.
The Concept in 30 Seconds
Semantic distance is a measure of how far apart two meanings are. “Dog” and “cat” are semantically close. They share context: pets, fur, domestication. “Dog” and “carburetor” are semantically distant. Almost no shared context.
Search engines use semantic distance to match intent, not just keywords. That’s the general idea, and it underpins a lot of how modern search works.
But there’s a specific, on-page version of semantic distance that most SEOs aren’t thinking about: how Google interprets the structural distance between terms within a single page. Not the conceptual distance between words in a language model. The literal structural distance between terms as defined by your HTML.
The Patent
US7716216B1, “Document ranking based on semantic distance between terms in a document.” Filed 2004, granted 2010, assigned to Google. Inventors: Georges R. Harik and Monika H. Henzinger. A continuation patent (US8060501B1) was granted in 2011. Bill Slawski wrote the definitive breakdown on SEO by the Sea.
The patent describes a system that locates semantic structures in HTML documents such as headings, lists, tables, divs, even elements styled with larger font sizes, and uses those structures to calculate distance values between terms. Those distance values feed into ranking scores that determine how relevant the page is for a given query.
The key insight: the search engine doesn’t just count the number of words between two terms to figure out how close they are. It looks at the HTML structure and uses that structure to override simple word-count proximity.
How It Works
The classic example from the patent makes this concrete. Imagine a page with the heading “Saturn Facts” and a list beneath it:
- Orbit: 10,759 Days
- Rotation Period: 10.7 Hours
- Mass: 568.5 x 10²⁴ kg
- Volume: 82,713 x 10¹⁰ km³
- Distance from the Sun: 1,434 x 10⁶ km
Two things happen under the patent’s logic.
First, “Saturn” in the heading is considered semantically close to every item in the list, regardless of position. The word count between “Saturn” and the last list item doesn’t matter. The heading creates a semantic container, and everything inside that container is equally close to the heading. The page is equally relevant for “Saturn mass,” “Saturn volume,” and “Saturn distance from the sun.”
Second, terms within the same list item are closer than terms across different list items. Here’s the counterintuitive part: “Saturn” and “Distance” (the heading and the last list item) are considered closer than “Days” and “Rotation” (the last word of item 1 and the first word of item 2), even though that second pair is visually adjacent on the page. The list boundary between items creates semantic separation that overrides physical proximity.
The patent lays out three rules:
- Both terms in the same list item: close.
- One term in a heading, one in any list item under that heading: approximately equally close, regardless of which list item.
- Terms in different list items: farther apart than either of the above.
The patent also notes that Google looks beyond formal HTML heading tags. A larger font size used as a visual heading can be interpreted as a heading element even without an H1 or H2 tag.
Slightly off-topic, but this is something I have been stressing with people for years. Google can understand heading structures by the layout even if proper H tags are missing. That’s why when you “fix” the missing H tags on a page or correct the order, you don’t see ranking improvement.
Google is trying to understand the visual and structural hierarchy of the page, not just parse HTML tags.
Why This Changes How You Think About Headings
Most SEOs treat headings as keyword opportunities. The patent reframes them as semantic containers that define relationships between every piece of content beneath them.
Heading text creates relationships, not just labels. If your H2 says “Installation Costs by Material Type” and the paragraphs beneath discuss pricing for hardwood, tile, and carpet, Google considers “Installation Costs” semantically close to all three materials. That heading establishes a relationship between the cost concept and every material mentioned in the section.
Now consider what happens if that H2 instead says “Additional Information.” The same content sits beneath it, but the semantic container is weak. Google gets far less signal about how “installation costs” relates to the content below. The heading is doing less work.
What you group together under a heading matters. If you want Google to associate two concepts, put them under the same heading. If you split them across different sections with different headings, you increase the semantic distance between them. This is a content architecture decision that most people make based on readability alone, without considering the semantic consequences.
Lists create equal-distance relationships. Every item in a list is equidistant from the heading above it. You can’t accidentally push a concept further from the main topic by placing it lower in a list. That’s structurally different from running paragraphs, where terms physically farther from the heading accumulate more word-count distance in a traditional proximity model.
Heading hierarchy is semantic nesting. An H3 under an H2 inherits context from the H2. The H2 inherits from the H1. You’re building a tree of semantic relationships. A well-structured hierarchy tells Google not just what each section is about, but how sections relate to each other and to the page’s central topic.
The Connection to Entity-Based SEO
If you’ve been following the entity extraction and information gain notes, this concept slots in directly.
Entities and their attributes are the building blocks. Semantic distance is how you control the relationships between those building blocks on the page. When you place an entity under a heading, you’re telling Google that entity is semantically close to the heading’s concept. When you place two entities under the same heading, you’re telling Google they’re related to each other.
This is entity relationship management at the page level. Your heading structure isn’t just formatting. It’s the mechanism by which Google interprets which entities belong together and how they connect to the page’s central topic.
What to Do With This
Review your heading text. Are your headings specific enough to create meaningful semantic containers? Generic headings like “Overview,” “More Info,” or “Details” create weak containers. Specific headings that name the concept being covered create strong ones.
Check your content grouping. Are related concepts under the same heading? Are unrelated concepts accidentally sharing a section? Look at your most important pages and ask whether the content beneath each heading actually belongs together semantically.
Use lists deliberately. When you have a set of items that should all be equally associated with a concept, a list under a clear heading is structurally stronger than scattering the same items across running paragraphs. The list structure guarantees equal semantic distance from the heading.
Think about heading hierarchy as a relationship tree. Your H1 is the trunk. H2s are branches. H3s are sub-branches. Content under each heading is bound to it semantically. When you’re planning a page, think about which concepts should be siblings (under the same parent heading) and which should be nested (sub-heading under a parent).
Don’t over-optimize. John Mueller has noted that heading hierarchy order doesn’t need to be perfect from Google’s perspective. The patent describes one ranking signal among many. Structure your content for humans first. But know that when you make a structural decision, it carries semantic weight.
The Takeaway
Your page structure isn’t just organization. It’s communication. Every heading you write, every list you create, every section break is telling Google how the concepts on your page relate to each other. The patent is from 2004, but the logic holds. Google is still reading the structure of your pages to understand meaning. Give it a clear structure, and it has a better chance of understanding what your page is actually about.
