Keyword Research

What keyword research actually involves

Keyword research is the process of finding the search queries your audience uses, estimating how much demand and competition each one carries, and deciding which topics deserve a page. The output is not a spreadsheet of keywords — it is a content plan where every planned page maps to a query (or query cluster) with real demand and a realistic chance of ranking.

Three questions drive the whole process:

1. Is anyone searching for this? Demand validation through search volume estimates.
2. What do searchers expect to find? Intent analysis by reading the current results — see search intent.
3. Can this site realistically rank? Difficulty assessment based on who currently occupies the SERP and how strong their sites are.

Skipping any one of these produces a familiar failure: content nobody searches for, content in the wrong format for the query, or content targeting queries dominated by sites you cannot displace.

How keyword metrics are calculated

Tools like Ahrefs and SEMrush estimate their metrics from clickstream panels and search data samples, so treat every number as a directional estimate rather than a measurement.

• Search volume is the estimated average number of monthly searches for an exact query, usually averaged over 12 months. Seasonal queries can swing far above or below their stated average in any given month.
• Keyword difficulty (KD) is a tool-specific score, typically derived from the link profiles of the pages currently ranking. A KD of 40 in Ahrefs is not comparable to a KD of 40 in SEMrush — each vendor uses its own formula.
• Cost per click (CPC) reflects what advertisers pay in Google Ads for that query. High CPC signals commercial value: someone is willing to pay for that click. Google Keyword Planner reports bid ranges directly from auction data, which makes it the most reliable free source for commercial value signals.

A practical composite: prioritize queries where (relevance to your offer) × (volume) ÷ (difficulty) is highest. The formula is informal, but the discipline of weighing all three at once prevents volume-only thinking.

Where keyword research fits in the workflow

The chain runs: audience need → search query → content plan → optimized page → measured performance. Keyword research sits at the second and third steps. Done well, it also feeds the later ones — the queries you choose define what "ranking success" means when you measure results in Google Search Console.

A step-by-step process

1. Seed list. Write down 10–20 terms your audience uses, taken from sales calls, support tickets, community threads, and competitor navigation — not from your own internal jargon.
2. Expand. Feed seeds into a keyword tool and pull related terms, questions, and autocomplete variations. Free options like Ubersuggest cover early-stage needs; Ahrefs and SEMrush give deeper competitive data.
3. Group into clusters. Queries that show substantially the same top-10 results belong on one page. "keyword research tools" and "best tools for keyword research" rarely need separate pages; "keyword research" and "keyword research for YouTube" almost always do.
4. Check intent manually. Search each priority query and read the results. Note the dominant format: guides, product pages, comparison lists, videos. Your page must match that format to compete.
5. Assess difficulty honestly. Look at who ranks. If the top 10 is all major publishers and established tool vendors, a new site should target a longer, more specific variant first.
6. Map one primary query per page. Assign each planned page a primary keyword and a handful of secondary variants. Two of your own pages targeting the same primary query will compete with each other.
7. Revisit quarterly. Demand shifts, SERPs change, and pages that rank at position 8 often justify a content refresh more than a new page does.

Common scenarios

New site with no authority. Skip head terms entirely. Target long-tail queries (typically three or more words with clearly specific intent) where the current results are thin or outdated. Volume per query is small, but the aggregate adds up and the wins build topical credibility.

Established site entering a new topic. Build a cluster: one pillar page on the broad term plus supporting pages on subtopics, interlinked. This concentrates relevance signals — the same logic behind structured learning paths.

E-commerce category planning. Keyword research often reveals the category structure customers expect ("running shoes for flat feet" as a facet, not a blog post). Let query language shape information architecture, not just content topics.

Common mistakes

• Choosing by volume alone. A 200-searches-per-month query that matches your offer beats a 20,000-searches-per-month query that does not.
• Ignoring the live SERP. Tool metrics summarize; the actual results page tells you the required format, the strength of competitors, and how many clicks organic results realistically get.
• Treating close variants as separate topics. This produces ten thin pages where one strong page would rank for all ten variants.
• One-time research. Keyword data decays. Queries gain and lose volume, and Google reshapes results pages — what was a ten-blue-links SERP last year may now lead with an AI-generated answer.
• Researching without a conversion path. Traffic with no connection to your offer inflates charts and nothing else. Tie every target query to a next step the visitor can take.

FAQ

How many keywords should one page target? One primary query plus its close variants — the cluster that shares the same top results. A well-built page commonly ranks for hundreds of long-tail variations of its primary term without targeting them individually.

Are search volume numbers accurate? No. They are modeled estimates, and different tools disagree routinely. Use them to compare relative demand between queries, not as a forecast of traffic. Search Console impression data on queries you already rank for is the only first-party demand number you have.

Is keyword research still relevant with AI search summaries? The mechanics still apply: people express needs as queries, and you need to know which needs exist. What changes is click expectation — informational queries answered directly in the results send fewer clicks, which raises the relative value of queries where searchers need a tool, a product, or depth that a summary cannot provide.

What is a realistic difficulty ceiling for a new site? There is no universal number, since each tool scores differently. The honest method is manual: if every top-10 result comes from a site with far more topical history and links than yours, pick a more specific query. Tool difficulty scores below the tool's "easy" band are a reasonable starting filter.

Free tools or paid tools first? Start free: Keyword Planner for commercial signals, Search Console for queries you already touch, autocomplete for question mining. Pay for Ahrefs or SEMrush when you need competitor keyword gaps and link-profile-based difficulty data — usually the point where you are planning content at scale.