How Do SEO Keywords Work? A Developer's Guide to Search Terms, Intent and Pages
Keywords are the link between what a person types and the page that answers it. Here is how search engines use them, where they belong in your files, and how to verify the result.

On this page
You ship a page about your product, wait, and nothing shows up in search. The page is fine. The problem is that nobody typed the words you used to describe it.
SEO keywords are the words and phrases people type (or say) into a search engine. They work as a matching signal: a search engine compares the query with the text, structure and links of pages it has crawled, then ranks the pages it judges most useful for that query and its likely intent. A keyword does nothing on its own. It works when a page that genuinely answers the question uses the language the searcher uses, in the places crawlers read.
This guide covers the mechanism, where keywords go in a codebase, and how to check whether they are working. If you want the planning layer on top of it, read keyword strategy.
What are SEO keywords?
A keyword is a query seen from the page's side. The searcher's query is "how to add a sitemap to next.js"; the page's keyword is the phrase it needs to answer that. Keywords come in a few useful shapes:
| Type | Example | What the searcher wants |
|---|---|---|
| Head term | sitemap |
Unclear; many possible answers |
| Long-tail | generate sitemap in next.js app router |
One specific answer |
| Question | how do seo keywords work |
An explanation |
| Comparison | surfer seo vs frase |
A choice between tools |
| Branded | seoagent cli |
A particular product |
"Keyword" is a convenient label, not a literal one. A page is rarely matched on one exact string. Google's SEO Starter Guide frames the work around understanding what people search for and making content that serves them, not around repeating a phrase.
How do search engines use keywords?
In rough order, a search engine does four things with a keyword.
- Crawl and parse. It fetches your page and reads the HTML: the
<title>, headings, body text, link text, image alt text and structured data. Content that only exists after client-side rendering can be missed or delayed, which is why React SEO and Next.js SEO both start with what the server returns. - Interpret the query. The engine works out what the searcher means, not just which words they typed. Synonyms, misspellings and related concepts are matched, which is the subject of semantic search vs keyword search.
- Infer intent. "How do seo keywords work" wants a guide. "Keyword research tool" wants a list of tools. The same words can need different page types, so engines favour pages whose format matches the task.
- Rank candidates. Relevance is only one input. Quality, usefulness, links and many other signals also feed the ranking, and nobody outside the search engine can state the exact weights.
The practical consequence: a keyword gets you into consideration, and the page earns the ranking by answering the question. Stuffing a phrase into the page does not help, and Google's spam policies list keyword stuffing as a violation.
Why intent matters more than the exact phrase
Before you pick a keyword, open the results for it and look at what ranks. If the first page is full of step-by-step guides, a product page will struggle to match, however many times it says the phrase. If the results are tool comparisons, a tutorial is the wrong format.
Label each target query with one intent:
- Informational: the reader wants to understand or do something.
- Commercial investigation: the reader is comparing options.
- Navigational: the reader wants a specific site or product.
- Transactional: the reader is ready to sign up or buy.
Then give each query one URL. Two pages chasing the same intent compete with each other, which is the cannibalization problem described in the content mapping example.
Where to put keywords for SEO
Put the primary phrase where it describes the page truthfully, and let the rest of the page use natural variants. These are the locations worth checking, with the file you would edit in a typical repo:
| Location | Why it matters | Where it lives in a repo |
|---|---|---|
<title> |
Strongest on-page description of the page | metadata.title or frontmatter title |
| Meta description | Often used as the snippet; written for clicks | metadata.description or description |
| H1 | Tells readers and crawlers the page's subject | The first heading in the page or Markdown |
| URL slug | Short, readable description of the page | The filename or route folder |
| Opening paragraph | Confirms the page answers the query early | The first lines of body copy |
| Subheadings | Cover the follow-up questions | ## and ### headings |
| Image alt text | Describes the image; helps accessibility too | alt="" attributes or Markdown image alt |
| Internal link text | Says what the target page is about | Anchor text in other pages |
See what is an SEO slug and what are SEO tags for the details on two of these. Here is what a consistent set looks like for this very page:
---
title: 'How Do SEO Keywords Work? A Developer''s Guide to Search Terms, Intent and Pages'
description: 'SEO keywords are the phrases people type into search. See how engines match them to pages, where to place them in your repo, and how to check they work.'
---
# How Do SEO Keywords Work? A Developer's Guide to Search Terms, Intent and Pages
The file is named how-seo-keywords-work.md, so the slug carries the phrase too. Title, description, H1 and slug agree on one subject. That agreement is the point; none of them is stuffed.
How to use SEO keywords without hurting the page
- Write for the reader first. Read the page aloud. If a sentence only exists to hold a phrase, cut it.
- Use variants and related terms. A page on keywords can say "search terms", "queries" and "phrases" where they fit.
- Answer the follow-up questions. The subheadings of the page should match what the reader asks next.
- Do not chase one page per spelling. "SEO keyword" and "SEO keywords" are the same intent; one page covers both.
- Link to the page with descriptive anchors, from pages that are genuinely related.
- Do not promise a position. Nothing you do on the page controls where a search engine places it.
How to check whether your keywords are working
Google Search Console is the primary source. Its performance report shows, by page and query, impressions, clicks, click-through rate and average position. Look for these patterns:
- Impressions but few clicks: the page is being shown, so the title or description may not match what searchers want.
- Queries you did not target: the page is answering something adjacent. Either add a section for it or consider a separate page.
- No impressions after a reasonable time: check that the page is indexed, linked from other pages and in the sitemap.
Search Console does not show every query (Google anonymizes rare ones), so treat it as a sample, not a full log.
Doing this from your repo with a coding agent
If your site is built in a repository and you work in Claude Code, Cursor or Codex, the checklist above is a set of file edits. SEOAgent runs on your own model inside that coding agent and edits files in your repo: titles, descriptions, headings, alt text, internal links, sitemap entries. Each change shows up as a diff you approve. There is no separate subscription for a hosted writer and no per-credit meter on the free Skill and CLI.
A bounded instruction looks like this:
For content/blog/how-seo-keywords-work.md, check that the title, description,
H1 and slug describe the same subject. Propose edits as a diff. Do not touch
other files.
Limits to know about: it needs a coding agent, it has no dashboard-only workflow, and its keyword research and Search Console analysis (see keyword research) are part of the paid cloud tier. The agent writes changes into your repo and you ship them through your normal deploy; it does not publish for you. See SEOAgent for Claude Code for the full workflow, and competitor keyword research for finding terms rivals rank for that you do not.
Summary
SEO keywords work as a bridge. A searcher's words go in; a search engine matches them, by meaning and intent, to pages that answer the question. Choose the phrase by checking intent in the live results, give each query one URL, state the subject consistently in the title, description, H1, slug and opening, and then read Search Console to see what actually happened.
References
- SEO Starter Guide: The Basics (Google Search Central)
- Spam policies for Google web search (Google Search Central)
- Performance report: search results dimensions and groupings (Search Console Help)
Put SEO on autopilot in your own editor
SEOAgent runs as a free skill inside Claude Code, Cursor, and Codex — on the model you already pay for. Audit, plan, and write SEO content right in your repo, with every change reviewed before it ships. No second AI subscription.
What do you use to work on your site?
Pick one above and we take you to the right setup.