SEO copywriting: how to write for intent, not keywords

What SEO copy is

For a long time, SEO copy meant text written “for keywords”: a required number of occurrences, a set length, the key phrase in every paragraph. This approach doesn’t match what the search engines themselves say.

Google classifies excessive keyword use as spam — stuffing a page with keywords or numbers to manipulate its search ranking. One example is unnaturally frequent repetition of the same words. In its “SEO Starter Guide,” Google adds that repeating the same keywords over and over, even with small variations, tires out users.

Yandex lists “Use of SEO text” as a separate violation: it covers SEO copy that gives the user no added value. Yandex defines over-optimization this way: the text uses words or phrases excessively, artificially padding its length without solving the reader’s task. The same source recommends writing primarily for people, not for robots.

From this, a working definition follows. Today, SEO copy is text that fully answers the query, is clearly structured, and helps the search engine match the page to that query. It does contain keywords, but they follow from the topic rather than dictate it.

SEO copy, SEO copywriting and content marketing: what’s the difference

These terms are often confused. They differ in purpose and in the criterion used to judge the result.

	Purpose	Quality criterion
SEO copy	answer a specific search query on a specific page	the page satisfies the intent, is clear to the reader, and doesn’t compete with other pages on the site
SEO copywriting	writing that text — this is a process and a writer’s skill	the text matches the brief: structure, subtopics, facts, queries without keyword stuffing
Content marketing	attract and retain an audience with materials across different channels	the materials serve business goals not only in search, but also in newsletters, social media and sales

SEO copy is a specific kind of content: the kind built for search demand. A good blog article can be both SEO copy and part of content marketing, if it has a target query and a useful answer.

Where the text starts: intent and page purpose

Which query the page answers, and where it leads the reader next

Intent is what a person wants to get from a query: to understand a topic, compare options, find a contractor, reach a specific site. Similar-looking queries can hide different tasks, so intent is checked against the search results: which pages rank at the top — articles, directories, company pages. For more on the types of intent and how to tell them apart from the search results, see the separate article.

A page’s purpose is its role on the site. Before you write, answer three questions:

• Which query does the page answer? One main intent and its possible phrasings.
• What should the person understand or do after reading? Get the answer, choose an option, move on to solving their task.
• Where does the page lead next? To a related topic, a detailed breakdown, a service page or a product page.

The text itself depends on this purpose. An informational article answers the question and doesn’t pitch in every paragraph. A service page explains what the client will get and doesn’t recap a textbook. If the purpose isn’t defined, the text tries to solve every task at once and ends up solving none of them.

The answer to the third question turns into links. Wherever the reader’s next question comes up, you need links from the text to related pages — with a clear anchor and an explanation of why to follow it.

It’s useful to record the query, the purpose, and the questions the text must cover in a brief before work starts. What goes into that brief on our side is described on the page about how we prepare briefs and content for search demand.

Check: does another page on the site already cover the topic

If another page on the site already answers the query, the new text will start competing with it. This is keyword cannibalization: two pages target the same intent and get in each other’s way. How to find these overlaps using a keyword map and webmaster-panel data is covered in the article on the keyword map: how to find pages competing for the same query.

Before you write, check:

• whether the site already has a page that answers this question — even under a different heading;
• which pages of the site already show up for the query — from search and from webmaster-panel data;
• whether the new text would actually work better as a section of an existing page.

Duplicates are a separate case. If pages with the same content are available at different addresses, Yandex’s crawler may treat them as duplicates, and the page you actually want can drop out of search — the crawler picks a different one from the group. Google also shows only one version of duplicate content in its results. And Yandex may not include a page in search at all if it duplicates content the crawler already knows. The takeaway for the text: two near-identical pages don’t double your chances — the search engine still has to pick just one of them.

Update an existing page or create a new one

If a suitable page exists but doesn’t answer the query fully, it’s usually better to update it: add the missing subtopics, fix outdated information, restructure it. A new page is justified when the topic has its own distinct user task, the search results show separate pages for it, and the site doesn’t cover it even as a section.

Google advises periodically revisiting published content: updating it, or removing it if it’s no longer relevant. An update means changing the substance, not just the date. Google lists changing the date on pages whose content has barely changed, just to make them look more current, as a sign of content created for search engines rather than for people.

Structure of an answer page

An answer page is built so the person gets the main point right away, with the details available if they want them. A typical block order:

1. Answer — in the first paragraphs.
2. Sections or steps — following the logic of the user’s task.
3. Table or list — wherever something needs to be compared or listed.
4. FAQ — separate questions that didn’t fit into the sections.
5. Sources — for facts the reader can verify.

Answer first

The first paragraph answers the main question with no preamble. For a “what is…” query, that’s a one- or two-sentence definition. For a “how to…” query, it’s a brief outline of steps. The rest of the text expands on the answer rather than building up to it.

A simple check: will the person get the answer if they read only the first paragraph? If not, the opening needs rewriting.

Steps, tables, lists, FAQ

One of the ways Google describes helpful content: the text is easy to read and well organized, and long text is broken into sections and paragraphs with headings so users can find their way around more easily.

In practice, that means:

• a section heading states what the section is about — as a question or a statement;
• one idea per paragraph, with long sentences broken up;
• steps go in a numbered list, enumerations in a bulleted one;
• a table, when options are compared against the same criteria;
• a term is explained the first time it’s used.

The heading hierarchy exists primarily for the reader. Google says a logical heading order helps screen readers, while for search optimization purposes the order itself doesn’t matter.

An FAQ block makes sense when the topic has separate questions that didn’t fit into the sections. If the FAQ just repeats the main text, it only makes the page longer.

Facts and sources

Every checkable claim in the text — a number, a definition, a rule, a description of how a search engine works — needs something to back it up. The best source is the primary one: documentation, official help pages, a study with a described methodology. Phrasing like “studies show” with no link gives the reader nothing.

Google says its algorithms look for content that best meets the E-E-A-T requirements — Experience, Expertise, Authoritativeness and Trustworthiness. Of these, trustworthiness matters most. At the same time, E-E-A-T itself isn’t a ranking factor, though in some cases Google favors content that reflects it.

Our rule for this kind of material is simple: we don’t use unverified numbers or facts. If a claim can’t be backed up, we either rephrase it as an opinion or remove it.

If you’re curious how answer structure and verifiable facts relate to generative search, start with the article on what GEO — optimization for generative search is and how it differs from SEO.

Keywords in the text: how to use them without over-optimizing

Keywords are the queries people use to search for the page’s topic. For where the queries for a text come from and how they’re grouped, see the article on the keyword map. Here, it’s about how to use an already-built group of queries.

Semantic coverage: not just the main query

A page usually has not one query but a group: the main phrasing, its variants, clarifications, questions. Semantic coverage is how fully the text covers the subtopics behind that group. The task isn’t to insert every phrase — it’s to answer the questions those phrases express.

How to work with this:

• break the group’s queries down into subtopics: definition, step-by-step process, differences, mistakes;
• cover each subtopic that shares the same intent in its own section or paragraph;
• leave out subtopics with a different intent — they need a separate page, or a link to one.

Google advises thinking about the words potential visitors will use to search for your content: people who know the topic well and newcomers may phrase a query differently. You don’t have to anticipate every phrasing — Google’s systems can tell how well a page matches a range of queries even without the exact same wording on the page. One of the self-check questions in Google’s guidelines is whether the topics covered are explored fully and thoroughly.

Headings, the Title tag and opening paragraphs

Title. Google recommends giving every page a title in the <title> tag and keeping it concise and informative. The same source adds: don’t use too many keywords — overusing them can be treated as spam. Yandex asks that the title be informative, use several informationally important phrases, and not repeat them: repetition carries no useful information, and search may flag such a title as spam. Google builds the title link shown in search results from several sources — including the words inside the <title> tag and other headings on the page.

In practice: the main query phrasing goes into the Title and the H1, as a natural phrase. The H1 can be more detailed than the Title. Every page has its own Title — not a list of query variants.

Section headings. Subtopic phrasings belong in H2s and H3s where they accurately describe the section’s content. There’s no need to force a keyword into every heading: a heading’s main job is to help the reader find the right section.

Opening paragraphs. There’s no separate “keyword in the first paragraph” rule in the Google and Yandex guidelines we cite here. But if the first paragraph actually answers the query, the main phrasing tends to appear in it naturally — as the name of the topic.

Synonyms and related concepts instead of repetition

Google notes that when you write naturally and avoid repetition by using synonyms, a page has a better chance of appearing in search — simply because the text contains more varied phrasing.

Synonyms connect to another idea — entities and context. An entity is the specific thing being discussed: a concept, a process, a tool, an organization. Context is the surrounding concepts that clarify which meaning is intended. The word “key” can mean a part of a lock or, informally, a keyword. In a text about SEO, words like “query,” “SERP” and “Title” sit nearby — and the meaning becomes unambiguous for both the reader and the search engine.

What this means for the text:

• name the topic’s key concepts directly, and define them the first time they appear;
• use one name per term — synonyms work well for ordinary words, not for terms;
• show the connections: what a process consists of, how one concept differs from another, what follows what.

A quick check for keyword stuffing is to read the paragraph aloud. If a phrase sounds unnatural, or you wouldn’t have written it if search engines didn’t exist, replace the repetition with a pronoun, a synonym, or just cut it.

Text drafted with AI: what to check

Search engines’ positions on AI-generated content keep being updated. Below are the exact statements from Google’s and Yandex’s official materials as of October 2026.

Google. According to Google, generative AI can be especially useful for researching a topic and organizing source material. But creating many pages with automated tools without giving visitors content that’s more useful than the source material can count as a violation of the scaled content abuse policy. The spam policies describe this kind of scaling regardless of exactly how the content was produced; one example given is many pages created with generative AI without adding value for users. Using automation, including AI, to create content purely to manipulate search results is something Google explicitly calls a violation of its spam policies.

Yandex. Yandex Webmaster’s help section lists sites with automatically generated content that’s useless to the user among the signs of low-quality content. In its webmaster blog in March 2025, Yandex wrote that using neural networks by itself isn’t a negative factor, as long as the content is good quality, has been checked by an expert, doesn’t mislead the user, and genuinely helps solve their question.

In both these statements, the tool itself isn’t called a violation: what matters is the value for the reader, the quality, and the purpose behind the content. So a draft prepared with AI is checked just as strictly as any other — and even more strictly on facts:

What to check	What to look at
Facts and numbers	Every claim is checked against the primary source. A draft may contain inaccurate or fabricated data — without a source, it gets removed
Links and sources	Every link has been opened: the page exists and confirms exactly what the text says
Match to intent	The text answers the page’s query and doesn’t drift into adjacent topics
Repetition and filler	No near-identical paragraphs, empty introductions, or repeated key phrases
Claims about the company	Clients, cases, numbers, timelines and result promises — only from confirmed company data
Metadata	Title, description and alt text are accurate: Google asks that you pay particular attention to the accuracy, quality and relevance of these when content is produced with automated tools
Independent value	The text contains something not in the source material: analysis, an example, experience. Among its self-check questions, Google asks whether content drawn from other sources has been reworked enough to provide substantial value

The decision to publish is made by a person — an editor or specialist who’s responsible for the content.

Common mistakes

• Text built around a keyword list. The page has no defined purpose, and the text is assembled around keyword occurrences.
• Chasing a keyword-density “norm.” The key phrase is repeated to hit a number, not because the sentence needs it.
• Length for its own sake. Among the signs of content written for search engines, Google names articles of a certain length written because the author heard that Google considers it optimal — and adds: “That’s not true.”
• The answer buried in the middle. A long introduction, with the main point only after several screens of scrolling.
• A new page instead of an update. The topic is already covered on the site, and the two pages end up competing.
• Near-identical pages for query variants. The search engine picks one; the rest serve no purpose.
• Facts without sources. Numbers and “studies” with no link to the primary source.
• An unchecked draft. AI-written text published without fact-checking or editing.

Checklist before publishing

1. The main query, intent and purpose of the page are defined.
2. No other page on the site shares the same intent; if a suitable page existed, an update was chosen instead of a new page.
3. The first paragraph answers the main question.
4. The subtopics of the query group are covered; topics with a different intent are moved out or replaced with a link.
5. There’s a single H1; the Title is unique, informative, and not a list of keywords; section headings describe the content.
6. Key phrases read naturally; there’s no repetition just to hit a count.
7. Every fact and number has a source; claims about the company are confirmed only.
8. Links lead to related topics and to the target page, with descriptive anchor text.
9. An AI-drafted text has been checked against the table above.
10. The meta description describes the content of this specific page.

Frequently asked questions

How many keywords should a text have?

There’s no universal number. The Google and Yandex guidelines we cite here set no rules for keyword density or occurrence counts; they warn against an excess of keywords. The real guideline is different: does the reader actually need this phrase right here?

Do keywords need to be in the page’s URL?

A short, clear URL is convenient for the reader. But Google says that, from a ranking standpoint, simply having keywords in the domain name or URL path is unlikely to have any effect.

Can you base a text on competitors’ content?

Studying which subtopics the pages in the search results cover is useful. Rewriting them isn’t. Google advises against using someone else’s content, in whole or in part, and against rephrasing material that’s already been published elsewhere. Yandex classifies sites that copy or rewrite information from other resources as low-quality.

Do you need to tell readers that a text was prepared with AI?

Google recommends telling the audience how the content was created, and if materials are generated automatically, disclosing that to visitors in an appropriate way. In its helpful-content guidelines, Google clarifies that this kind of disclosure is worth adding if readers might have questions about how the content was created. We haven’t found an official Yandex statement on this in the Webmaster help section.

Should an article credit its author?

Among the questions Google suggests asking about your content: is it clear to visitors who created it, and is authorship credited where readers would expect it? For how-to and analysis articles, a byline helps the reader understand who’s responsible for the content.