Dispatch
What Is a Keyword in SEO?
A keyword is the word or phrase a person types or speaks into a search engine to find information, a product, or a website. That definition sounds simple, and for...
On this page
A keyword is the word or phrase a person types or speaks into a search engine to find information, a product, or a website. That definition sounds simple, and for the first decade of commercial search engines it was simple: search engines mostly matched the literal text of a query against the literal text on a page. That is no longer how modern search works, and treating a keyword as nothing more than a string to match is one of the more outdated habits in SEO.
Today a keyword is better understood as a label for an underlying need. When someone searches, Google’s systems try to work out the topic, the entities involved, and the intent behind the words, not just the words themselves. Google’s own SEO documentation describes ranking systems that look at the overall meaning and context of a page in relation to a query, not a checklist of matching terms (see the SEO Starter Guide). A keyword is still the entry point for research and measurement, but it represents a query and the intent behind it, not a string you need to repeat verbatim on a page.
Types of Keywords
SEO practitioners commonly group keywords by length and search volume. There is no strict, universally agreed cutoff between these categories, but the general pattern looks like this:
| Type | Typical length | Search volume | Example |
|---|---|---|---|
| Head term | 1 word | High | "shoes" |
| Body / mid-tail | 2 to 3 words | Moderate | "running shoes women" |
| Long-tail | 4+ words, or highly specific | Low individually, large in aggregate | "best running shoes for flat feet marathon" |
Head terms are broad, competitive, and often ambiguous about what the searcher actually wants. Long-tail terms are narrower and more specific, which usually means clearer intent and less competition for any single phrase, even though each individual long-tail phrase gets far fewer searches than a head term.
Keyword Types by Search Intent
A separate, and arguably more useful, way to categorize keywords is by the intent behind the search rather than the word count. The foundational version of this framework comes from computer scientist Andrei Broder’s 2002 paper “A Taxonomy of Web Search,” which split web queries into three categories: navigational (find a specific site), informational (learn about a topic), and transactional (complete an action, such as a purchase or download). You can read the original paper through the ACM Digital Library or a public mirror hosted by SIGIR.
SEO practitioners later added a fourth, informal category that sits between informational and transactional: commercial investigation (sometimes called “commercial intent”), which covers searches where someone is comparing options before they buy, such as “best wireless earbuds under $100” or “Ahrefs vs Semrush.” This fourth category isn’t part of Broder’s original academic taxonomy, but it has become standard in SEO practice because comparison and review searches behave differently from both pure information-seeking and a direct transaction.
- Informational: “what is a keyword,” “how does crawling work”
- Navigational: “Google Search Console login,” “Ahrefs”
- Commercial investigation: “best keyword research tools,” “Semrush vs Moz”
- Transactional: “buy Ahrefs subscription,” “Semrush pricing”
Matching content to the right intent category matters more than matching the literal keyword text. A page built to answer an informational query will rarely rank well for a transactional one, even if it contains the exact same words.
Keywords and On-Page Optimization
Keywords still play a role in on-page SEO, but the role is narrower than it used to be. Google’s spam policies explicitly call out keyword stuffing, defined as filling a page with keywords or phrases in an unnatural, repetitive way to try to manipulate rankings, as a violation that can hurt rather than help a page’s visibility (see Google’s Spam Policies for Google Web Search).
Practical, low-risk uses of keywords still matter:
- Reflecting the primary topic in the title tag and H1.
- Using the term and its natural variations in headings and body copy where they read naturally.
- Writing descriptive URLs and alt text that reflect what the page is actually about.
- Covering the topic thoroughly enough that related terms and questions appear naturally, without forcing them in.
The goal is topical clarity, not keyword density. A page that reads naturally and answers the query well will tend to pick up relevant variations on its own.
Why Pure Keyword Matching Stopped Being Enough
Three Google algorithm changes mark the clearest turning points away from literal string matching:
Hummingbird (2013) rebuilt how Google parses a query, shifting emphasis toward understanding the meaning of an entire query rather than matching individual words. Google announced it at a press event at its original Menlo Park garage; Search Engine Land’s contemporaneous reporting remains one of the most detailed public explanations of what changed (see “FAQ: All About The New Google ‘Hummingbird’ Algorithm”).
BERT (2019) added a language-understanding model that helps Google interpret the relationships between words in a query, including small connecting words like prepositions that change meaning (“to” vs “from,” for instance). Google’s own announcement states that BERT affected roughly one in ten English-language searches in the US at launch (see Google’s blog post, “Understanding searches better than ever before”).
MUM (2021) went further. Google describes MUM, the Multitask Unified Model, as roughly 1,000 times more powerful than BERT, trained across 75 languages, and able to work across both text and images rather than text alone (see Google’s blog post, “MUM: A new AI milestone for understanding information”). The practical effect for keyword strategy is the same direction as Hummingbird and BERT, just further along it: a page is associated with a topic through the breadth and clarity of what it actually explains, not through phrase matching.
None of these updates eliminated keywords. They changed what a keyword needs to do: it still has to signal the topic clearly, but it no longer needs to appear in a page verbatim, in a specific order, or at a specific frequency to be associated with that topic.
Keywords and Entities
Modern search also reasons about entities, real-world people, places, organizations, and concepts, rather than only strings of text. This is part of why a page can rank for a query that never uses the exact phrase typed into the search box, as long as the page’s content clearly establishes what topic and entities it covers. This doesn’t mean keywords are irrelevant; it means the keyword is treated as one signal pointing to a topic, alongside the page’s overall content, structure, and the way other pages and queries relate to that same topic.
A Note on Keyword “Value”
Keywords are sometimes described as having intrinsic economic value because of how they’re priced in paid search auctions. That’s directionally true, advertisers do bid more for keywords closer to a purchase decision, but be careful with specific dollar figures. A hypothetical example: if a business pays an average of $50 per click for a transactional keyword and converts 5% of clicks into a $2,000 sale, the math behind that bid is straightforward to model. That is an illustrative example only, not a benchmark or industry average; actual cost-per-click figures vary enormously by industry, competition, and geography, and anyone wanting real numbers should pull them from their own Google Ads account or current platform data rather than treat any single quoted figure as universal.
Practical Keyword Research, Briefly
Full keyword research methodology (tools, clustering, filtering by intent and difficulty) deserves its own treatment, but the starting point is simple: identify a handful of “seed” terms that describe your core topics, then expand outward using tools like Google Search Console (which shows queries already driving impressions to your site), Google Trends (relative interest over time), and dedicated keyword databases such as Ahrefs or Semrush. Group the resulting terms by intent before you group them by volume; a smaller cluster of clearly transactional terms is often worth more than a larger cluster of vague informational ones.
Frequently Asked Questions
Is a keyword the same thing as a search query?
Not exactly. A search query is the literal string a person types or speaks. A keyword is usually the broader topic or term that a group of similar queries share, used as a unit for research and tracking purposes.
How many keywords should one page target?
There’s no fixed number. A page should have one clear primary topic, but it can and usually does rank for many related terms and questions if it covers that topic thoroughly. Trying to force a single page to target several distinct primary topics tends to produce weaker, less focused content.
What is keyword cannibalization?
It’s when two or more pages on the same site target the same or very similar keyword and intent, splitting ranking signals between them instead of consolidating into one strong page.
Does keyword stuffing still happen, and does it work?
It still happens, and it doesn’t work. It’s explicitly listed in Google’s spam policies as a manipulative practice that can lead to lower rankings.
Are long-tail keywords still worth targeting?
Yes. Individually they bring less traffic, but in aggregate, and because of their clearer intent, long-tail terms are often a more efficient source of qualified traffic than competing directly for head terms.
Has voice and AI-driven search changed what a keyword is?
It has changed query patterns more than it has changed the underlying concept. Conversational interfaces, including Google’s AI Overviews and AI Mode, accept longer, more natural-language queries instead of compressed two- or three-word phrases. The keyword research task shifts toward identifying the topics and questions behind those longer queries rather than guessing at exact phrasing.
It’s worth being precise about what “ranking for a keyword” means once AI Overviews and AI Mode are in the picture. Google’s own documentation states that these features draw on the same underlying Search index and ranking systems used for standard results, not a separate “AI algorithm,” and that a page generally needs to be indexed and eligible to appear with a normal snippet in Search before it can be cited as a supporting link in an AI Overview (see Google’s AI features documentation). So a keyword still needs to earn a page a place in that underlying index and ranking the same way it always did. What’s changed is what happens after that: a well-cited page may satisfy the searcher inside the AI-generated summary itself, with or without a click, which is a new layer on top of ranking rather than a replacement for it.