Semantic SEO guide: write for things, not strings
In 2012 Google described its Knowledge Graph with three words: things, not strings. Search had started to understand the people, places and objects behind words instead of only matching letters. Semantic SEO is writing for that kind of search. You cover a subject the way an expert would explain it, name the real things involved and how they connect, and let Google work out which searches the page answers.
How Google moved from matching words to reading meaning
Google moved from matching words to reading meaning in steps it documented itself over more than a decade. These are the milestones that matter for how you write.
The Knowledge Graph
Google's announcement described a model of real-world entities and their relationships, holding more than 500 million objects and more than 3.5 billion facts and relationships at launch.
Google blog, 16 May 2012BERT in Search
Google said BERT would help it better understand one in 10 searches in US English, especially longer, conversational queries where small words like "for" and "to" change the meaning.
Google blog, 25 October 2019The systems Google names
Its ranking systems guide lists BERT, neural matching (matching concepts in queries and pages), RankBrain (relating words to concepts) and passage ranking (judging sections of a page).
Ranking systems guide, updated 10 December 2025Synonyms and spelling
Google's How Search Works pages say its language systems, including its synonym system, improve results in over 30% of searches across languages.
Google, How Search Works, read 4 October 2026
Plenty of guides say Google's MUM model now ranks pages by meaning. Google's ranking systems guide says MUM isn't currently used for general ranking in Search; it's used for specific applications, such as improving featured snippet callouts and searches about COVID-19 vaccines.
What entities, attributes and connections mean on a page
On a real page, semantic SEO means naming the things the page is about, covering the details people want about them and showing how they connect to your business.
A distinct thing: a product, a service, a condition, a place, an organisation. "Heat pump hot water system" is one; so is "Geelong".
A detail about the thing that people care about: tank size, running cost, noise, where it can be installed.
How one thing relates to another: a heat pump uses electricity, so it relates to your tariff and to rooftop solar.
The meaning mapone page, three lines
- Heat pump
- Gas storage
- Tank size
- Rooftop solar
- Running cost
- Install time
- Noise
- Rebates
- Suburb
- Council
- Climate zone
- Housing stock
- Things lineThe products, services and parts a customer would expect the page to name.
- Questions lineThe details people ask about before they decide: cost, time, eligibility.
- Place lineWhere the job happens, and what about that place changes the advice.
Each station is something a thorough answer would mention. The page is where the lines meet.
An illustrative topic map written for this guide.
An example sheet written for this guide. Rebate rules change; link to the current state government page rather than repeating amounts that may go out of date.
Notice what isn't on the sheet: variations of one phrase. "Hot water replacement", "replace hot water system" and "new hot water unit" are the same need. A page that properly covers the entities above will be understood for all of them, which is the point Google's AI features guide makes when it says its systems understand synonyms and general meaning.
Semantic SEO advice that Google's documents don't support
Because "semantic" sounds technical, it attracts tools and tricks. These are the ones we'd skip, and the Google source that explains why.
Sprinkling "LSI keywords"
Latent semantic indexing is an old information-retrieval method. Google's ranking systems guide never mentions it; the language systems it does name are BERT, neural matching and RankBrain. Covering the topic properly does the job those keyword lists claim to do.
A page for every phrasing
Google's AI features guide warns that creating separate content for every variation of a search, mainly to manipulate results, breaks its scaled content abuse policy.
Chopping pages into "chunks"
The same guide says there's no requirement to break content into tiny pieces, because Google's systems understand pages that cover several topics.
Entity markup as a ranking lever
Structured data can make a page eligible for rich results, but Google says it isn't required for its AI features and there's no special markup to add.
Topic scores from tools
Content-grading tools compare your page to others' word choices. Google's AI guide says no third-party tool has access to its internal ranking systems. Use them as a prompt, not a target.
Writing longer to look thorough
Google's starter guide says content length alone doesn't matter for ranking.
Google Search Central, "A guide to Google Search ranking systems" (updated 10 December 2025), guide to generative AI features in Search (updated 10 July 2026) and "SEO Starter Guide" (updated 10 December 2025), read 4 October 2026.
How to write one semantically complete page
We work in this order whether the page is a service page, a product category or a guide.
Name the one need the page serves
"Replacing a household hot water system" is a need. "Hot water" is a category. Our keyword research guide explains how to find and group needs.
List the things involved
Products, services, rules, places and people a customer would expect to hear about. Pull them from enquiries, the "People also ask" box and your own experience on the job.
List what people want to know about each one
Costs, time, risks, choices, eligibility. These become your subheadings and paragraphs, in the order a customer would ask them.
Add what only you can say
Your prices, your process, photos of your work, the mistakes you see. Google's helpful content guide asks whether a page offers original information or insight beyond the obvious; this is where you answer yes.
Connect it to the rest of the site
Link to the pages that cover the connected things in more depth, with descriptive link text, and link back from them.
Read it as the customer
Would they need to search again after reading? If yes, something on the sheet is missing.
For local businesses, the place line deserves real detail. A suburb is an entity too, with a council, a climate and housing stock that change the advice. A page that explains why older weatherboard homes in a suburb often need a particular fix says far more than a page that just repeats the suburb's name. Our on-page SEO guide covers how to label each of these sections clearly.
From one complete page to a site that covers a subject
Semantic SEO works at page level. Its site-level cousin is topical coverage: a set of pages, each answering one need well, linked so a reader can move between them the way the meaning map suggests.
One page covers one need completely: its entities, their attributes and how they connect.
A set of such pages covers the wider subject, with links that follow real connections, never a page per phrasing.
A new page should exist because it answers a different need, never because it targets a different wording or swaps the suburb name on the same text. Google's spam policies call out pages generated mainly to manipulate rankings, however they're made. Our article on topical authority goes further into planning the set.
Forty posts, no visitors, and pages fighting each other
Owners who take "cover the subject" literally often end up with more pages and less traffic. These four keep turning up in Search Console help threads, and each is a meaning problem before it's a technical one.
You publish a blog post every week and almost nobody reads them
Posts on broad topics such as "why hot water matters" compete with bigger sites that answered the same question years ago, often in the same words. Weekly volume, especially machine-written volume, adds pages without adding anything a searcher can't already find.
InsteadWrite fewer posts, each tied to a question customers actually ask you before they buy, and fill them with what only you know: real prices, the steps of your job, the local quirks.
Two or three of your pages take turns ranking for the same search
A service page, a blog post and a suburb page all chasing "hot water replacement" split the signals between them, and Google may show a different one each week. Search Console shows it: in the Performance report, filter by the query and look at the Pages tab.
InsteadChoose one page per need. Fold the useful parts of the others into it, redirect the leftovers to it, and point internal links at it with descriptive words. Keep a separate page only when it answers a different question.
Search Console lists pages as crawled but currently not indexed
Google's help for the page indexing report says these pages were crawled but not indexed, may or may not be indexed later, and don't need resubmitting. When most of them are thin or near copies of each other, such as tag pages or suburb pages with the same text, that pattern usually means Google sees little reason to keep them.
InsteadDon't request indexing over and over. Merge or remove the near duplicates, give the pages that matter more substance and more links from your strongest pages, and check the report again after a few weeks.
Search results quote a directory about your business instead of your own site
When your own pages say little about what you do, where and for whom, a directory listing or a review site can simply be the clearer source.
InsteadMake your own pages the most complete description of the business: services, areas, prices or price factors, licences, who does the work. Keep your Business Profile and main listings consistent with it. Nobody can promise which source an answer will use.
Google Search Console Help, "Page indexing report", read 5 October 2026.
Semantic SEO and Google's AI features
Google's guide to its generative AI features, updated 10 July 2026, says AI Overviews and AI Mode are rooted in its core ranking and quality systems. So the same work applies; there's no separate semantic layer to work on.
The guide describes "query fan-out": the AI model runs a set of related searches at once to gather more results for a question. Its example is a question about a weedy lawn fanning out into searches about herbicides, chemical-free removal and prevention. That's the meaning map at work. A page that already covers the connected things in a subject has more chances to be relevant to those related searches, without you writing a page for each one.
Google also says content offering a unique, experienced point of view is what helps most over time. You can see how your pages appear in its AI features in the generative AI performance report in Search Console. No one can promise a page will be cited, and we won't. Our guide to SEO for AI search covers the rest.
Semantic SEO questions, answered
What is semantic SEO?
Semantic SEO is writing and structuring a page around meaning: covering a subject's real things, the details people want about them and how they connect, rather than repeating one exact phrase. It suits how Google's language systems, such as BERT and neural matching, understand queries and pages.
What is the difference between semantic SEO and traditional SEO?
Older keyword-led SEO aimed to match the exact words people typed, often by repetition. Semantic SEO aims to answer the need behind those words fully. Clear titles, crawlable pages and links still matter in both.
What is an entity in SEO?
An entity is a distinct, identifiable thing: a person, place, organisation, product or concept. On your page, entities are the things a thorough answer has to mention.
Do LSI keywords work?
Not in the way they're sold. Latent semantic indexing is an older retrieval technique, and Google's guide to its ranking systems doesn't mention it. Cover the subject properly and related words appear naturally.
What is semantic search?
Semantic search is search that works from meaning rather than exact word matches. Google's ranking systems guide describes RankBrain as returning relevant content even when it doesn't contain all the words searched, by understanding how words relate to concepts.
Does semantic SEO still need keywords?
Yes, as a guide to what people need and the words they use. You still put the clearest phrase in the title and main heading. You just don't need a page, or a paragraph, for every variation; one complete page can be understood for many wordings.
Does schema markup help semantic SEO?
It can describe your page's entities in a fixed format and make the page eligible for some rich results. Our schema markup guide lists what's still supported.
Questions drawn from Google autocomplete suggestions for "what is semantic seo", "semantic seo vs traditional seo", "what is an entity in seo", "lsi keywords" and "semantic search", checked from Australia on 4 October 2026.
Do your pages answer the whole question?
Our free audit checks your key pages for gaps a customer would notice.