How to do keyword research: A step-by-step guide

Around 94% of all webpages receive no traffic from Google, according to SE Ranking. That is not a web design problem or a technical issue. In most cases, it comes down to keyword selection. The page targeted the wrong term, misread what searchers actually wanted, or went after a head keyword the site had no realistic chance of ranking for.

Keyword research fixes that. This guide covers nine clear steps: defining goals, building seed lists, expanding with tools, filtering by intent, assessing difficulty, analyzing competitors, clustering, accounting for AI search, and prioritizing with a framework that connects every keyword decision to business outcomes.

What keyword research is (and why most teams do it wrong)

Keyword research is the process of identifying the exact words and phrases your target audience types into search engines, then evaluating each term for relevance, ranking potential, and business value. The goal is not to build the longest keyword list. The goal is to find the right terms and connect them to content that ranks, gets clicked, and drives qualified traffic into the funnel.

Most teams rush straight to a tool, pull a list of high-volume terms, and start writing content. That approach produces pages competing against sites with far more authority, and the content rarely ranks. Rand Fishkin, co-founder of Moz and SparkToro, framed it clearly: "Effective keyword research is about understanding the problems your audience is trying to solve, not just the words they use to search."

Keyword research is audience research with a search data layer on top.

The business case for getting this right

Organic search drives an average of 33% of overall website traffic across seven major industries, according to Conductor's 2025 State of SEO report. The top organic result on Google averages a 39.8% click-through rate, per FirstPageSage. Position two drops to 18.7% and position three to 10.2%.

The gap between page one and everything else is significant. Keyword research is how you build a clear path to the terms that actually matter to your business.

Step 1: Define your goals before you open any tool

Set a clear objective before starting keyword research. Without one, you will collect data with no filter. Your goal determines which intent types matter, which funnel stages to prioritize, and what a winning keyword looks like for your specific business. Write your goal in one sentence before touching Ahrefs or Semrush.

Goals shape every downstream decision. A SaaS company building top-of-funnel awareness needs different keywords than one capturing purchase-ready buyers comparing plans. A local service business targets different intent than a B2B consultancy building thought leadership.

"The first thing we ask new clients is: what does a qualified lead look like for your business? Keyword research should start with that answer, not with a tool." Tanner Medina, Co-Founder and Chief Growth Officer, Launchcodex

Two questions that frame your keyword goals

Before opening any tool, answer these:

1. What action do you want the reader to take after finding this content? Subscribe, book a call, request a demo, or buy a product.
2. Which stage of the funnel does this content serve? Awareness, consideration, or decision.

A keyword like "what is programmatic SEO" serves awareness. "Programmatic SEO agency pricing" serves a decision-stage buyer. Both are valid, but they need different content, different calls to action, and different success metrics. Mixing them without a plan produces a content calendar with no strategic direction.

Common mistakes at this stage

• Chasing volume without asking whether the traffic converts
• Targeting decision-stage keywords before the site has enough authority to rank for them
• Building a keyword plan that serves no specific stage of the funnel

Step 2: Build your seed keyword list

Seed keywords are broad, short terms that anchor your research. They are not final targets. They are the starting point for expanding into the full range of phrases your audience uses. Start with what you know about your business, your customers, and the problems you solve, then confirm those assumptions with real search data.

A seed keyword for a content marketing platform might be "content strategy." From that single term, a research tool can surface hundreds of variations, questions, and long-tail phrases that real people search every month. The quality of your seed list determines the quality of everything that follows.

Four sources for strong seed keywords

1. List the core services or products you offer. Write them in plain customer language, not internal company jargon.
2. List the specific problems your customers describe when they first contact you or during discovery calls.
3. Review your Google Search Console data for queries already driving impressions to existing pages.
4. Use AnswerThePublic to surface question-based variations pulled directly from Google autocomplete.

Build a starting list of 10 to 20 seed terms. From each one, a tool like Ahrefs Keywords Explorer or the Semrush Keyword Magic Tool will generate dozens of related variations filtered by volume, intent, and difficulty.

Step 3: Expand and validate with research tools

Use a keyword research tool to turn your seed list into a full dataset with search volume, keyword difficulty, and traffic potential for each term. The right tool gives you enough data to filter, prioritize, and spot content clusters you would never find manually. Use at least one paid tool alongside Google Search Console for first-party validation.

Each major tool has a different strength.

For most teams, the strongest combination is Google Search Console for existing sites plus one paid tool like Ahrefs or Semrush.

What to collect for each keyword

Build a working spreadsheet with these columns:

• Keyword phrase
• Monthly search volume
• Keyword difficulty (KD score)
• Traffic potential (distinct from raw volume)
• Search intent type
• SERP features present (AI Overview, featured snippet, People Also Ask)
• Business relevance score (rate 1 to 3 manually)

Traffic potential deserves attention here. A keyword with 500 monthly searches where the top-ranking page pulls 4,000 visits across all its related variations is far more valuable than the raw number suggests. Ahrefs surfaces this metric directly, and it is one of the most underused filters in the tool.

Step 4: Filter by search intent

Search intent is the most important filter in keyword research. It defines what the searcher actually wants to find. Target the wrong intent type and Google will not rank your content, regardless of how well optimized it is. Check the SERP for every shortlisted keyword before committing resources to it.

Google recognizes four intent types:

1. Informational: The searcher wants to learn. Example: "how does retargeting work."
2. Navigational: The searcher is heading somewhere specific. Example: "Semrush login."
3. Commercial: The searcher is comparing options. Example: "best CRM for SaaS companies."
4. Transactional: The searcher is ready to act. Example: "buy Semrush annual plan."

How to identify intent without guessing

Open Google and search the keyword. Look at the top five organic results.

• Results are blog posts and how-to guides: informational intent.
• Results are product pages and pricing pages: transactional intent.
• Results are comparison articles and review roundups: commercial intent.

That is the format Google has determined satisfies this query. Match it. Publishing a long-form educational guide when Google ranks product pages for the same keyword means your content will not rank, regardless of its quality or depth.

Intent mapping across the funnel

Map intent types to funnel stages when planning content:

• Awareness content targets informational queries
• Consideration content targets commercial queries
• Conversion content targets transactional queries

A balanced content plan needs keywords across all three stages. A site that targets only informational terms builds traffic but rarely converts it.

Step 5: Assess keyword difficulty and traffic potential

Keyword difficulty (KD) is a directional score, not an absolute measure. It tells you roughly how competitive a keyword is based on the backlink authority of pages currently ranking. Use it alongside your own site's authority and the actual traffic potential of the keyword to decide whether it is worth pursuing.

According to Backlinko's analysis of 306 million keywords, 91.8% of all search queries are long-tail keywords. Long-tail terms typically carry lower KD scores and lower search volume, but they convert at higher rates because they reflect more specific intent. Most content plans underweight them.

A KD of 35 on a site with a Domain Rating of 15 is an unrealistic target. The same KD of 35 on a site with a Domain Rating of 55 is very achievable. Always compare your site's authority against the average authority of pages currently ranking for that keyword.

Why volume should not be your primary filter

Rand Fishkin has said publicly that he avoids leading with volume metrics because they pull teams toward the wrong decisions. A senior marketer at a B2B SaaS company put the business-first view plainly, as cited by DOJO AI: "I'd rather rank for a keyword with 400 searches a month that drives demo requests than chase 10K searches that do nothing for the business."

That framing is especially relevant in B2B and SaaS markets where a single conversion can justify months of content investment.

KD thresholds as a starting reference

These are directional ranges, not fixed rules. Always review the actual SERP to check what content ranks, how many backlinks it has, and whether you can build something meaningfully better or more current.

Step 6: Analyze competitor keywords

Competitor keyword analysis reveals proven demand. If a competitor ranks for a term and drives traffic from it, that keyword has validated commercial interest. Your job is to find the terms they rank for, identify where their content falls short, and create something more useful, more thorough, or more current.

There are two types of competitors to analyze:

• Direct competitors: Businesses selling the same product or service you sell.
• Content competitors: Sites ranking for the keywords you want, even if they serve a different audience. This often includes media publications, industry blogs, and SaaS knowledge bases.

Both matter. Content competitors can dominate informational keywords your potential customers search early in the buying process, well before they reach a branded or transactional query.

How to run a keyword gap analysis

1. Identify three to five direct competitors and two to three content competitors.
2. Enter each domain into Ahrefs Site Explorer or Semrush's Organic Research tool.
3. Review their top organic keywords sorted by traffic, not volume.
4. Filter for keywords where you do not have a ranking page.
5. Flag any keyword where the competitor holds position 6 to 20. A weak ranking means a better piece of content can displace them.
6. Add qualified keywords to your spreadsheet with a "gap" tag.

Four signs a competitor's content can be beaten

• It was published more than two years ago and has not been updated
• It covers the topic at surface level with no specific examples, data, or frameworks
• It misses the current AI search angle entirely
• It targets informational intent on a query where commercial intent is now dominant

All four are common. A SERP review and a quick scan of the competing page are enough to spot them.

Step 7: Cluster keywords and map them to content

Keyword clustering groups related terms by topic and intent so a single page can target multiple queries rather than a single phrase. This approach builds topical authority, prevents keyword cannibalization, and creates a more efficient content structure. Google's algorithm rewards comprehensive topic coverage across a site, not just individual optimized pages.

As Nightwatch's SEO team explains, "Grouping keywords by topic or intent is critical for modern SEO. It helps you build topical authority, create focused, high-performing content, and improve internal linking and on-page structure."

How to cluster keywords manually

1. Group keywords that share the same core topic and intent. "Email marketing strategy," "email marketing best practices," and "how to create an email marketing plan" likely belong in one cluster.
2. Identify the primary keyword for each cluster. This is the term with the highest traffic potential and the closest match to your content goal.
3. Assign secondary keywords to the same cluster. These are variations and related phrases that belong on the same page.
4. Separate clusters with clearly different intent. "Best email marketing tools" (commercial) and "how email marketing works" (informational) are two separate clusters even though the topic overlaps.

Tools like Semrush's Keyword Manager and Ahrefs can automate the initial grouping. Use them to generate cluster suggestions, then review manually to catch intent mismatches the algorithm misses.

Keyword mapping: connecting clusters to pages

After clustering, assign each cluster to a specific URL.

• Existing page: Assign the cluster to the page that already covers the topic. Update it to include the primary and secondary keywords naturally.
• New page needed: Flag it for content creation and add it to the editorial calendar.
• Cannibalization risk: If two existing pages already target the same cluster, decide which one to keep, then redirect or consolidate the weaker page.

Build a keyword map in a spreadsheet with columns for cluster name, primary keyword, secondary keywords, target URL, and page status.

Step 8: Account for AI search and GEO

AI search is reshaping which keywords drive organic clicks. When a Google AI Overview appears in results, the top organic result loses approximately 34.5% of its clicks, according to Ahrefs data. Keywords that consistently trigger AI Overviews still carry brand visibility value, but factor the reduced click-through potential into how you prioritize and structure content around them.

AI-referred sessions grew 527% year-over-year in the first five months of 2025, according to Previsible's 2025 AI Traffic Report. ChatGPT, Perplexity, and Google AI Mode now answer queries directly, often without sending users to a website at all. This changes how keyword decisions translate into actual traffic.

"AI tools cut keyword clustering time from hours to minutes, but they still miss the business context only you can provide. Validate every AI-generated keyword idea against real intent data before it goes on your map." Derick Do, Co-Founder and Chief Product Officer, Launchcodex

How to identify AI-affected keywords

Before finalizing your target list, search each keyword in Google and note whether an AI Overview appears. If it does:

• The keyword still matters for GEO citation and brand visibility.
• Click-through rates to organic results will be lower than position-based averages suggest.
• Content must be structured so AI systems can extract and cite it, rather than only ranking it.

Queries with informational intent trigger AI Overviews most often. Commercial and transactional queries are less affected, making them higher-value targets for organic click traffic.

Optimizing keyword-targeted content for GEO

GEO (Generative Engine Optimization) is the practice of structuring content so AI search engines cite it in generated responses. For each piece of content built around a keyword cluster, apply these rules:

• Open every major section with a direct, citable answer to the implied question.
• Use structured formatting throughout: numbered steps, clear definitions, comparison tables.
• Cite credible sources, reference named experts, and use precise terminology. AI engines favor well-cited, entity-rich content.
• Target question-based keywords with dedicated answer-led sections. These map directly to the prompts users type into ChatGPT and Perplexity.

A 2024 study by Princeton, Georgia Tech, and IIT Delhi tested nine GEO content methods across thousands of samples. Keyword stuffing scored lower than doing nothing. Authoritative, structured content significantly outperformed it.

Step 9: Prioritize with a business-first scoring method

Not every good keyword deserves equal investment. Prioritize by combining business relevance, intent type, keyword difficulty, and traffic potential into a single score. This keeps your content roadmap tied to revenue-generating potential rather than raw traffic volume.

The KOB (Keyword Opportunity and Business value) framework, developed by Siege Media, formalizes this approach. It scores each keyword across three dimensions: how much traffic potential it carries, how achievable the ranking is for your site, and how directly it connects to what you sell.

A simplified prioritization scoring model

Score each keyword across three dimensions using a 1 to 3 scale:

Add the three scores together. Target keywords with totals of 8 or 9 first. Keywords scoring 5 or below belong in a backlog for when your site's authority grows.

When low-volume, high-intent keywords outperform

A keyword with 200 monthly searches and a business relevance score of 3 often outperforms a keyword with 5,000 monthly searches and a relevance score of 1. This holds especially in B2B and SaaS contexts where buyer intent is specific, purchase cycles are long, and a single conversion justifies significant content investment.

How to turn keyword research into a repeatable system

Done once and filed away, keyword research stops working fast. Research from Passionfruit shows that 31% of high-value keywords change significantly in intent or volume every six months. Google's own data confirms that 15% of all daily searches have never been made before. New queries emerge constantly. Keyword research is an ongoing system, not a one-time project.

What to review each quarter

• Check your keyword map. Which target pages have moved up, plateaued, or dropped?
• Look for new competitor content entering the SERP on your target keywords.
• Review Google Search Console for new queries driving impressions without dedicated content.
• Run a fresh competitor gap analysis to catch opportunities that were not available last time.
• Flag and retire pages not gaining traction on business-critical terms. Consolidate them where possible.

Building a quarterly keyword review workflow

1. Export current keyword rankings from Ahrefs or Semrush at the start of each quarter.
2. Flag any keyword that has dropped more than five positions since the prior review.
3. Pull Google Search Console data for pages with strong impressions but low click-through rates. These usually signal an intent mismatch or a title that does not match what searchers expect.
4. Add new seed keywords based on recent product changes, new audience segments, or competitor moves.
5. Update the keyword map with new clusters and set priorities for the next quarter's content calendar.

When Launchcodex builds a keyword strategy for a new client, this review cycle starts on day one. The keyword map is a living document updated every quarter alongside performance data, not a static spreadsheet built at launch and forgotten.

Using AI tools to accelerate the process

86% of SEO professionals already use AI tools in their keyword research workflow, and 65% report improved results, according to Ranked.ai. The most practical use cases:

• Generating seed keyword variations fast using a prompt like: "List 20 specific questions a B2B SaaS marketing manager would search when evaluating an SEO agency."
• Clustering a large keyword list by grouping terms with the same implied topic and intent.
• Identifying intent mismatches by asking an AI to categorize a list of keywords by the four intent types.

AI tools speed up the process. They do not replace the judgment required to connect keyword decisions to business goals. Validate AI-generated keyword ideas against real search data in Ahrefs or Semrush before adding them to your map.

FAQ

What is the best free keyword research tool?

Google Search Console is the best free starting point for any site already generating impressions. It shows exactly which queries are driving clicks and where gaps exist between impressions and rankings. For new sites with no traffic yet, Google Keyword Planner gives volume ranges at no cost. Pair both with the free tier of Keywords Everywhere for lightweight daily research.

How many keywords should one page target?

Target one primary keyword and two to four secondary keywords per page. The primary term drives the page structure, title tag, and main heading. Secondary terms should appear naturally in subheadings and body copy. Trying to rank a single page for 10 unrelated keywords splits its focus and rarely works.

How often should I redo keyword research?

Run a full keyword research process when launching a new site or a new content program. After that, review the keyword map quarterly. Check Google Search Console monthly for new query opportunities. Major algorithm changes or new product launches also warrant a focused re-evaluation.

Does keyword research still matter with AI search?

Yes, and it matters more than before. Understanding which queries trigger AI Overviews, which terms get cited in ChatGPT and Perplexity responses, and how to structure content for GEO all depend on the same foundational keyword intelligence. The outputs adapt to a new channel. The process does not go away.

What is keyword cannibalization?

Keyword cannibalization happens when two or more pages on the same site target the same keyword. Google splits its evaluation between the competing pages, which usually means neither ranks as well as it could. Fix it by deciding which page best serves the searcher's intent, then redirecting or consolidating the weaker page into the stronger one.

What is the difference between search volume and traffic potential?

Search volume is the average number of monthly searches for a specific keyword phrase. Traffic potential estimates the total organic visits the top-ranking page actually receives across all related keyword variations, not just the exact phrase. Traffic potential is almost always the more useful number when making content investment decisions.