Keyword Research Step by Step (With Examples)

Keyword research is the process of finding the terms your audience types into Google, estimating how many people search for each one, judging the competition and deciding which ones deserve a page. The result is a prioritized list of topics to write, not a spreadsheet of numbers. The method below has 8 steps and works with free tools.

This guide is part of the path of the complete SEO guide and is the first step of the content layer: before writing or optimizing any page, you need to know what people actually search for.

Quick summary

A keyword is the term or question someone types into search. It can have several words.
Search volume is an estimate, not an exact fact. Use it to compare topics, not to promise traffic.
"Difficulty" is a metric calculated by third-party tools. The most reliable check is to look at who occupies the first page.
Every search has an intent (learn, compare, buy). If your page does not match it, volume does not matter.
Each main page should target one main keyword, so your pages do not compete with each other.
Search Console shows the searches that already bring people to your site, and it is the best source for refining the list.

How to read this guide: if you are starting out, follow the 8 steps in order and use the hypothetical example as a model. If you already have experience, go to the "For SEO professionals" blocks, the metrics table and the prioritization step.

What is a keyword and why does it matter?

A keyword is any term or phrase a person types into a search engine, such as "pilates for lower back pain" or "how much does a pilates class cost". It matters because it is the bridge between what you offer and the question the customer asks. If you write with the company's vocabulary and the customer uses another, nobody finds the page.

In plain English: think of keyword research as listening to customers at the counter before setting up the shop window. Instead of guessing what they want to see, you write down the questions they ask and organize the store around them.

Keywords are usually split by length and specificity:

Type	Example	Characteristics
Short and generic ("head")	pilates	High search volume, vague intent, heavy competition
Medium	pilates for beginners	Clearer intent, medium competition
Long and specific ("long tail")	pilates for lower back pain in Austin	Low volume per term, very clear intent, less competition

Long searches add up to a lot. Google reported in 2019 that 15% of daily queries are new, according to Pandu Nayak, vice president of Search. No tool lists all those phrases, which is why writing to answer real questions tends to work better than chasing a single term.

How do you do keyword research in 8 steps?

The method has 8 steps: define the topics, gather ideas, measure demand, assess competition, classify intent, prioritize, map each term to a page and review with real data. Do each one in a simple spreadsheet: one row per term and columns for volume, intent, competition and decision.

Define topics, not words. List 5 to 10 subjects your business can speak about with authority. For a bike shop, "maintenance", "choosing a bike" and "accessories" are topics.
Gather search ideas (seed terms). Use what customers already say: support questions, sales objections, reviews, groups and forums. Complete it with Google's own suggestions: autocomplete, the "Related searches" block and the "People also ask" block, which show searches and questions related to the original search.
Measure demand. Combine two free Google tools, explained in the next section.
Assess competition on the results page itself. Search the term and look at who appears: giant portals, stores, small blogs, videos. If only huge brands occupy the first page, the competition is heavy for a new site.
Classify the intent. Note whether the search is informational, comparative, transactional or local. The guide to search intent shows how to read this on the results page.
Prioritize. Give each term a score (explained further below) and choose the top 10.
Map term to page. Each main keyword goes to one page, existing or new. Terms with the same intent stay on the same page; different intents call for different pages.
Review with real data. After a few weeks, check in Search Console which searches your pages appear for and adjust.

Which free tools should you use and what does each one show?

Three free tools cover most of the work: Google Search Console (what already happens on your site), Google Trends (relative interest over time) and Google Ads Keyword Planner (ideas and volume estimates). Each answers a different question, and none should be used alone.

Search Console. The performance report shows clicks, impressions, CTR and average position by query and by page, according to Google's help. It is the only source of real data about your site. Queries with impressions and a position between 8 and 20 tend to be the best candidates for optimization, but that range is a heuristic, not a Google rule. See the Search Console guide.
Google Trends. It shows relative interest in a topic, from 0 to 100, normalized by the topic's share of all searches in the period and region. The Trends FAQ warns that these are not absolute volumes and that the tool is not a scientific survey. It is useful for comparing terms and spotting seasonality (for example, "sunscreen" rises before summer).
Keyword Planner. It was built for paid search campaigns: Google explains that it helps you discover keywords and see search and cost estimates. You need a Google Ads account, and the help page mentions that access to the basic features requires completing the account setup, including billing information.

For SEO professionals: read the Keyword Planner columns carefully. According to Google Ads' definitions, "average monthly searches" is the average, over 12 months by default, of searches for the term and its close variants. And the "Competition" column measures how many advertisers showed up for the term relative to all of Google's terms, not how hard it is to rank organically. If someone says a term has "low competition" because Keyword Planner showed that, they are mixing ads with SEO.

How do you interpret volume and difficulty without being misled?

Volume tells you how much demand exists; difficulty tries to tell you how hard it is to appear. Both are estimates. Volume is an average that groups close variants, and difficulty is a score calculated by each tool with its own formula, unrelated to Google. Use both only to compare terms against each other.

In plain English: volume is foot traffic on the street. A street with 10,000 curious pedestrians can sell less than one with 100 people determined to buy. That is why volume alone does not choose the keyword: intent and your chance of appearing weigh just as much.

A manual difficulty check that does not depend on a paid tool takes two minutes per term:

Who holds the top positions? Large, established sites, or pages from small sites?
What is the format? Articles, product pages, videos, comparison sites, forums? If you only offer a format different from what Google shows, your chances drop.
Is the answer complete? If the results are shallow, outdated or off topic, there is room.
Are there pages from sites the same size as yours? If so, it is a sign you can get in.

How do you prioritize the list?

Prioritize by combining relevance to your business, a realistic chance of appearing and demand. A simple formula prevents impulsive decisions. It is our own heuristic, not a Google standard: adjust the weights to your case.

Score = Relevance (0 to 3) × Chance of appearing (0 to 3) × Demand (1 to 3)

Relevance: 3 = leads straight to a sale; 0 = unrelated to the business
Chance:    3 = first page without giant sites; 0 = only huge brands
Demand:    3 = high within the list; 1 = low

For SEO professionals: group by intent and topic, not by isolated term. In the Search Console performance report, filter by one query and open the pages tab: if two of your URLs get impressions for the same query, you already have an overlap signal to resolve before publishing more. For the initial list, export the queries of pages that already have impressions and use them as seed terms.

Terms with a score of 0 leave the list, because zero relevance or zero chance cancels out the rest. Among those that remain, start with the highest scores and with the ones where you already have a page to optimize: improving what exists is usually faster than creating from scratch.

Hypothetical example: a pilates studio in Austin

Hypothetical example: a pilates studio wants to attract new students through Google. It has no blog and only one services page. Volume numbers are left out on purpose, because they vary by tool and by time.

Search	Intent	Perceived competition	Decision
pilates	Vague	High (portals and big brands)	Do not prioritize
pilates for lower back pain	Learn and evaluate	Medium	Guide article
pilates in Austin	Local and transactional	Medium	Optimize services page
how much does pilates cost	Compare prices	Medium	Article with price ranges and what influences them
pilates or weight training for weight loss	Compare	Low	Comparison article

The final decision is to map five pages: one optimized (services) and four new. The studio decides to keep "pilates", the generic term, only as the site's theme, without trying to rank for it directly. After a month, it opens Search Console and sees which of these searches already generate impressions. The ones that show up at positions 11 to 20 become the first round of adjustments.

Which keyword research mistakes should you avoid?

The most common mistakes are choosing by volume alone, ignoring intent, using the same term on several pages and overusing the keyword in the text. Google treats artificial stuffing as spam: the SEO Starter Guide lists keyword stuffing among the practices that violate the policies and recommends writing naturally, anticipating the terms people would use.

The same keyword on several pages. The pages compete with each other, which is called keyword cannibalization.
Thinking only about isolated words. Pages that cover a full topic reinforce each other, as the guide to topic clusters explains.
Forgetting AI searches. Longer, conversational questions change the type of term. See how people search in AI and the concept of query fan-out, in which one question becomes several searches.
Never reviewing. Demand changes. Redo the list every quarter with Search Console data.

Frequently asked questions

What is a keyword in SEO?

It is the term or phrase a person types into a search engine to find something. It can be one word ("pilates") or several ("pilates for lower back pain in Austin"). In SEO, you identify which terms your audience uses and create pages that answer them.

What is the best keyword research tool?

To start, the free combination of Search Console, Google Trends and Keyword Planner covers the essentials. Paid tools help at scale, with more data and competitor analysis, but their difficulty metrics are their own estimates and vary between vendors.

Is search volume reliable?

It is a useful estimate for comparing terms, not an exact number. In Keyword Planner, volume is a 12-month average that includes close variants of the term. For your own site, real impression data in Search Console is more reliable than any external estimate.

How many keywords should I use per page?

One main one and a few natural variations that fit the same intent. There is no ideal number, and Google does not require a minimum density. Write to answer the question fully and use the terms the way a person would in conversation.

Are long-tail keywords worth it?

Usually, especially for new sites. They have less volume individually, but clearer intent and less competition. Since a large share of daily searches is new, pages that answer specific questions find demand that tool lists do not show.

Do I need to research keywords for every post?

Yes, but lightly. Before writing, confirm the main term, the intent the page needs to meet and whether you already have another page on the subject. This step avoids duplicate content and keeps the text aligned with what people are looking for.

Next step

The next step is understanding what is behind each search: read search intent: types, examples and how to serve each one. With the list ready and the intents classified, apply the on-page SEO checklist to each page. If the list became a queue of topics and the bottleneck is now producing them, Contentor generates articles with AI using templates with validated structure and WordPress publishing, and the text should still go through your review.