Keyword research is the practice of identifying the specific search queries your potential customers type into Google, ranking them by commercial value and ranking difficulty, and using the prioritized list to guide content production and on-page optimization. It is the load-bearing first step of any serious SEO program, and the step most often skipped or done badly. This post walks through what real keyword research involves in 2026, where the leverage actually sits, and the common patterns that produce traffic without producing customers.
What good keyword research actually produces
A complete keyword-research output for a small or mid-sized business should produce three documents:
- The seed keyword list — 200-500 queries the business plausibly wants to rank for, organized by topic cluster.
- The priority subset — 30-60 keywords scored on commercial intent, search volume, and ranking difficulty. This is what content production targets in the first 6-12 months.
- The competitive gap analysis — keywords competitors rank for that the business does not, ranked by opportunity.
What good keyword research does not produce: a thousand-row spreadsheet with no prioritization, a list of head-term queries no small business can realistically rank for, or a list of long-tail queries with so little volume that ranking them produces no traffic.
The four signals to score every keyword on
- Search volume. How many people search this query per month? Tools like the major keyword research tools give estimates. Volume matters but is less important than the other three signals.
- Commercial intent. Does someone searching this query intend to buy? “Best HVAC service near me” has clear commercial intent; “how does HVAC work” does not. Commercial-intent queries are worth 10-100x non-commercial queries.
- Ranking difficulty. How hard is it to rank for this query against existing competitors? Most tools give a 0-100 difficulty score based on competitor backlink profiles and domain authority. Difficulty above 50-60 is usually out of reach for small business sites in year 1.
- Topical fit. Does this query map to a service or product the business actually offers? Ranking for queries you do not serve produces traffic that does not convert.
The four-bucket model for organizing the priority list
Once keywords are scored, organize them into four buckets based on the realistic ranking timeline:
- Quick wins (0-3 months). Long-tail, low-difficulty, high-intent queries. These are the queries you can rank for fastest. Typical pattern: “[service] [city]” or “[service] [specific use case]” with 50-500 monthly searches and difficulty under 25. Hit these first to produce early ranking signal and early lead flow.
- Medium-term (3-9 months). Mid-difficulty commercial queries. Volume 500-2000, difficulty 25-50. The bulk of a 12-month content roadmap targets this bucket. Pillar pages and the deeper cluster posts fit here.
- Long-term (9-24 months). High-volume head-term queries. Volume 2000+, difficulty 40-70. These are the queries that compound over years. They require the topical-authority signal built in months 1-9 before they become realistic targets.
- Aspirational (year 2+). Very high-volume queries dominated by major established competitors. Volume 5000+, difficulty 60+. Rank for these once the moat is built, not before.
The intent classification framework
Commercial intent is not binary. A more useful framework splits intent into four categories:
- Transactional. Someone ready to buy or sign up. “Buy [product],” “[service] near me,” “hire a [profession].” Highest commercial value. Convert at the highest rate.
- Commercial investigation. Someone researching a purchase but not yet ready. “Best [product],” “[product] reviews,” “[service] vs [alternative].” High commercial value. Convert at moderate rates with longer cycles.
- Informational with commercial fit. Someone learning about a topic that ties to a product or service. “How does [thing] work,” “what is [concept].” Lower direct conversion but builds topical authority and captures top-of-funnel demand that converts later.
- Pure informational. Someone curious with no purchase intent. “History of [topic],” “fun facts about [topic].” Avoid unless the topic genuinely supports brand-building or thought-leadership goals.
Most serious SEO programs allocate roughly 50% of content effort to transactional + commercial-investigation queries, 40% to informational-with-commercial-fit, and 10% to pure informational thought-leadership.
Common mistakes
- Targeting head-term keywords too early. A new site cannot rank “best CRM software” against established competitors in months 1-12. Going after them anyway produces 18 months of frustration and no rankings.
- Ignoring local-intent variants. “Plumber” has massive search volume nationally but is impossible to rank for from a single-city service business. “Plumber [city]” is realistic, has clear local intent, and converts well.
- Chasing volume over intent. A 5,000-monthly-search informational query is usually worth less than a 200-monthly-search transactional query. Intent matters more than volume.
- No mapping to existing content. Identifying keywords without identifying what page targets them produces a keyword list that does not connect to actual content production.
- One-time keyword research that never updates. Queries shift over time. The keyword research that worked in 2024 has stale assumptions about AI-search behavior, voice-search patterns, and emerging long-tail queries. Refresh quarterly.
How AI search changes keyword research
The rise of AI search engines (ChatGPT, Perplexity, Google AI Overviews) has changed what counts as a useful keyword to target. New considerations:
- Question-based queries get surfaced in AI answers more often than keyword-based queries. “How do I [task]” outperforms “[task] guide” for AI citation rates.
- Long-tail commercial queries get cited more often because they have clearer intent that AI engines can match to specific content.
- Branded queries matter more under AI search because AI engines weight brand-mention authority heavily.
- Comparison queries (“X vs Y”) have high AI-citation rates because AI engines preferentially cite content that gives clean comparisons.
Tools that actually help
The keyword research stack we use:
- Paid keyword data providers for raw keyword data + ranking data. Multiple options exist at different price points; the right tool depends on volume needs.
- Google Search Console for queries the site already gets impressions on. Free, accurate, and a source of truth for actual user query behavior on your site.
- Google Keyword Planner for volume estimates on queries you do not yet rank for. Free with a Google Ads account.
- Manual SERP inspection for the top 30-60 priority queries. Look at what is actually ranking, what the AI Overview says, what features are in the SERP. Tools cannot replace looking at the real SERP.
The bottom line
Keyword research is one of the highest-leverage hours of work in any SEO engagement, and one of the most often skipped. Done well, it produces a 12-month content roadmap that compounds. Done badly, it produces a spreadsheet that gets ignored and a content program that targets the wrong queries for two years. The work is not glamorous, but the difference between disciplined keyword research and “we will figure it out as we go” is usually the difference between an SEO program that hits its targets and one that does not.
For our SEO methodology, see the SEO services pillar. For the SEO terminology used in this post, the SEO glossary has definitions.