Semantic SEO: how to write for meaning, entities, and AI search
Semantic SEO and entity SEO explained without the jargon, plus the four habits that make your brand easy for search engines and AI to understand.
A plain-text version of this site for AI assistants is at https://contengi.com/llms.txt, with the full product documentation in one file at https://contengi.com/llms-full.txt. Every page is also available as markdown by adding .md to its URL.
TL;DR
- Semantic SEO means writing for meaning: covering the topic and the intent behind a search, with the people, products, and ideas involved named clearly enough that a machine can tell them apart.
- Entity SEO is the part that deals with those named things. An entity is anything with a distinct identity, like your brand, your founder, or a concept such as content engineering.
- Google has ranked on meaning for over a decade, through the Knowledge Graph, Hummingbird, and BERT. AI answers lean on it even harder, because they summarise ideas instead of matching words.
- For a small team the job comes down to four habits: say clearly what you are, cover a topic properly, link related pages, and back it up with structured data.
- Repeating keywords no longer does much. Repeating your brand's facts consistently everywhere does.
There was a time when SEO meant typing "best accountant Leeds" into a page seven times and hoping. That time ended years ago, and plenty of advice still hasn't noticed.
Here's what semantic SEO and entity SEO are, why they count for more now that AI writes the answers, and what to do about them without hiring a data scientist.
What is semantic SEO?
Semantic SEO is the practice of optimising content for what a search means, instead of the exact words typed. That means covering a topic in enough depth to answer the real question, using the language people use around it, and making the relationships between ideas obvious.
Google started down this road in 2012, when the Knowledge Graph let search understand "things, not strings". Hummingbird followed in 2013. In 2019, BERT helped Google understand the context of words in searches, including small words like "to" and "for" that change what a query means.
Today Google runs several ranking systems built to understand language, including neural matching, which connects the concepts in a query to pages that never use its exact words. A page can rank for a search it doesn't contain, if it plainly answers it.
What is entity SEO?
Entity SEO is semantic SEO applied to the named things in your content. An entity is a person, place, organisation, product, or concept with its own identity. "Apple" the fruit and "Apple" the company share a spelling and nothing else, and search engines tell them apart by context.
Your brand is an entity. So is your founder, your product, and the category you want to own. The goal of entity SEO is simple: make it easy for search engines and AI models to know who you are, what you do, and what you're connected to.
What this looks like in our own Search Console
Our brand name gets misspelled a lot. People search for "contegi", "contengy", and "contegy", and Google still shows our pages near the top for most of them. That's entity recognition working in our favour: Google has learned that those near-misses point at one company.
It also works against us. "Content engineering" and "engineering content" use the same two words and mean completely different things, and our pages get shown for searches about marketing for engineering firms. Nobody clicks, because we're the wrong answer. Clearer definitions and examples on our own pages are the fix, so the meaning can't be mistaken.
Why semantic SEO counts for more in AI search
AI answers summarise ideas. When ChatGPT or Google's AI Overviews explain a topic, they pull passages that answer the meaning of a question, often from several pages at once. Keyword density has almost nothing to do with it.
Google's AI features also run query fan-out, a set of related searches the model fires off to gather material. A page that covers a topic's natural sub-questions has more chances to be one of the sources, and fan-out is a big part of why GEO and SEO overlap so heavily.
Entities count too. If AI models aren't sure what your brand is, they leave it out or describe it wrongly. Clear, consistent facts about your company are the raw material for accurate brand answers in LLMs.
Semantic SEO vs keyword SEO
| Keyword-first SEO | Semantic SEO | |
|---|---|---|
| Starting point | One target phrase | A topic and the intent behind it |
| Page scope | One page per keyword variant | One strong page per real question |
| Language | Exact match repeated | Natural wording, related terms, named entities |
| Structure | Keyword in title and headings | Headings that mirror the sub-questions people ask |
| Proof of relevance | Density and backlinks | Depth, internal links, structured data, consistent facts |
Keywords still have a job. They tell you what people type and roughly how they think about a topic. They just stop being the thing you repeat and start being the thing you research.
How to do semantic SEO: a small-team checklist
- Define yourself in one sentence. Put a plain "Contengi is a content engineering platform for small marketing teams" style sentence on your homepage and about page. Use the same wording everywhere: LinkedIn, directories, your founder's bio.
- Answer the question behind the keyword. Read the top results and the People Also Ask box, list the sub-questions, and make sure your page answers the ones your reader cares about.
- Name things properly. Use full names for tools, people, and concepts the first time, and say how they relate. "Claude, Anthropic's AI model" beats "the tool".
- Link related pages on purpose. Internal links with descriptive anchor text tell search engines which pages belong together and which one leads.
- Add structured data. Organization markup tells Google your official name, logo, and contact details, and the sameAs property links your site to your profiles elsewhere. Add Article and FAQ markup on blog posts.
- Make sure bots can read it. Schema or content injected by JavaScript can be invisible to AI crawlers. Run a page through our free AI bot view to check what GPTBot sees, including your schema.
- Say something original. Covering a topic fully gets you into the conversation. A first-hand example or your own data is what makes an engine pick you over the nine others saying the same thing, which is the whole idea behind non-commodity content.
Common semantic SEO mistakes
Cramming in "LSI keywords" is the big one. There's no such thing as LSI keywords in Google's ranking, and stuffing a page with a list of related words reads badly to people and does nothing for machines.
Writing one page per synonym comes second. "AI content tool", "AI content tools", and "content AI tool" are the same question. One strong page answers all three. Several thin ones compete with each other.
The third is inconsistency. If your homepage calls you a platform, your LinkedIn calls you an agency, and your directory listing calls you software, you've given every machine three answers to "what is this company?". Pick one and repeat it everywhere your brand appears, which is most of the work if you want to control what AI says about your brand.
Frequently asked questions
What is semantic SEO in simple terms?
Optimising content for the meaning behind a search, not only the exact words. You cover the topic properly, answer the related questions, and make the things you mention easy to identify.
What is the difference between semantic SEO and entity SEO?
Semantic SEO is the overall approach of writing for meaning. Entity SEO is the part focused on named things like your brand, people, and products, so search engines and AI models can recognise and connect them.
Is semantic SEO still relevant with AI search?
More than before. AI answers are built from passages that match the meaning of a question, and clear entity information makes it easier for them to name your brand correctly.
Do I need schema markup for semantic SEO?
It isn't required, but it helps. Organization markup with sameAs links is the quickest win for most small businesses, because it confirms who you are in a format machines read directly.
Are LSI keywords part of semantic SEO?
No. Google doesn't use latent semantic indexing. Use the natural language your audience uses and cover the topic properly, and related terms show up on their own.