Keyword Research Without Guesswork: Use Your Own Data First

Why Most Keyword Research Is Backwards

Most keyword research starts with a tool. You type in a seed keyword, the tool returns 500 related keywords with estimated search volume and "difficulty," and you pick the ones that look promising. Six months later, you have a content calendar full of posts that rank for queries no one is buying for.

The problem is that the tool is guessing. It does not know what your actual customers search for. It does not know which keywords already drive revenue to your site. It does not know which keywords are realistic for you to rank for given your domain authority and content depth. And it certainly does not know which keywords match the search intent of someone close to a purchase decision versus someone in a casual research phase.

Better keyword research inverts the order. It starts with your own data, what queries already trigger your pages, where you are close to ranking, what your customers are actually buying for, and then uses tools to expand and validate. This is the same evidence-based discipline behind evidence-based SEO applied to research specifically.

This guide covers the complete keyword research framework: how to find your own opportunities, how to expand them with tools, how to validate intent, and how to organize keywords into pages that actually rank.

The Three Sources of Keyword Ideas

Good keyword research pulls from three sources, in this order.

Source 1: Your Own Search Console Data

The first place to look is queries that already trigger your pages. Go to Google Search Console → Performance → Queries. Sort by impressions, descending. Export the top 1,000.

You now have a list of queries Google has already decided your site is relevant for. These are not aspirational, they are real. The opportunities live in three patterns:

Quick-win queries (positions 4-10). You are already on page one or close to it. A small content or internal-linking improvement often pushes them to a top-3 position where the click-through rate triples. The What To Fix First feature automates this triage.

High-impression, low-CTR queries. You rank for the query but the title and meta description are not winning the click. Rewrite them.

Queries with no matching page. You appear for a query but have no page specifically targeting it. Build the page. This is one of the highest-EV moves in SEO because Google has already validated the relevance.

Source 2: Customer Conversations

The second source is what your customers actually say. This is the source most SEO teams skip, and it is the highest-value source for commercial keywords.

Specifically, mine:

• Sales calls. What questions do prospects ask before they buy? Those are queries.
• Support tickets. What problems do customers have after they buy? More queries.
• Reviews and testimonials. What language do customers use to describe your product? Use that exact language in titles and headers.
• Internal site search. What do visitors type into your search box? They are telling you what they want to find.
• Email replies. What questions come up repeatedly in your inbox?

Customer-conversation keywords are usually long-tail, high-intent, and lower-competition. They convert better than head terms because they are closer to a real purchase decision.

Source 3: Tools (Last, Not First)

Once you have exhausted your own data and customer conversations, then use tools to expand. The right way to use a keyword tool is to seed it with what you already know works and ask it for variations and adjacent queries.

Useful tool categories:

• Search Console expansion (the Keyword Intelligence tool can pull related queries from your existing GSC data)
• Competitor gap analysis (what queries do your top 3 competitors rank for that you do not, see competitor SEO analysis)
• Question discovery (AlsoAsked, AnswerThePublic, for content keywords)
• SERP feature analysis (which queries trigger featured snippets, People Also Ask, video carousels, these affect how you should structure the page)

What to ignore in tool data:

• Search volume. It is an estimate and often wrong by an order of magnitude. Treat it as relative, not absolute.
• Keyword difficulty. It is a third-party score with no relationship to actual ranking factors. Useful as a directional signal, dangerous as a decision-maker.

The Three Types of Keywords to Target

Once you have a list of candidate keywords, classify them by type. The strategy for each type is different.

Quick-Win Keywords

Definition: Keywords where you currently rank position 4-10 in Search Console, with at least 100 monthly impressions, on a page that is commercially relevant to your business.

Strategy: Optimize the existing page. Better title and meta description, deeper content, more internal links from related pages, updated information, schema markup. The goal is to push from position 4-10 to position 1-3 where CTR is 3-5x higher.

Time to results: 2-6 weeks after the change ships.

Why this matters: Quick wins are the highest-EV work in SEO because Google has already validated that the page is relevant. You are not trying to rank from scratch, you are improving an existing ranking. Most small businesses can move 5-10 quick-win keywords per quarter, which compounds to meaningful traffic over a year.

Content Gap Keywords

Definition: Keywords where competitors rank but you do not. You have no page targeting the query, or the page you have is too thin.

Strategy: Build a new page or substantially expand an existing one. The page should match the search intent better than competitors and be 1.5-2x as comprehensive. Use the Content Studio to draft from your own evidence rather than starting from a blank document.

Time to results: 2-6 months. New pages take longer than optimizing existing pages.

Why this matters: Content gaps are how you grow into new query territory. Quick wins keep you on the same hill; gaps move you to new hills. Most sites need a mix, heavy on quick wins for short-term results, ongoing investment in gaps for long-term growth.

Long-Tail Variations

Definition: Specific, low-competition queries that are easier to rank for than broad terms. Usually 4+ words. Often question-form. Lower volume per query but higher conversion intent.

Strategy: Either add the long-tail as an H2 section on a related parent page, or, for high-value commercial long-tails, build a dedicated page. The decision depends on whether the long-tail represents a meaningfully different user intent.

Time to results: 1-3 months for new pages, 2-6 weeks for new sections.

Why this matters: Long-tails are how small sites compete with big brands. You will not outrank Amazon for "blue widgets," but you can outrank them for "blue widgets for narrow cabinet drawers under €30." The aggregate traffic of dozens of long-tail wins often exceeds a single head-term win, with much higher conversion rates.

How to Validate Search Intent

A keyword without validated intent is worthless. The same query can mean different things to different searchers, and you need to know which intent your page is targeting before you write it.

The four classic intent categories:

Informational. The searcher wants to learn something. "What is keyword research?" "How does SEO work?" These queries trigger blog posts, guides, and educational content. Conversion rates are low; traffic value comes from brand-building and email captures.

Navigational. The searcher wants a specific site or page. "UpSearch login." "Google Analytics dashboard." Brand-name queries are navigational. You either own these or you do not, you cannot really compete for someone else's brand.

Commercial investigation. The searcher is comparing options before a purchase. "Best AI SEO platform." "UpSearch vs Ahrefs." "Plumber Dublin reviews." High-intent. Conversion rates are 3-5x informational.

Transactional. The searcher is ready to buy. "Buy AI SEO platform." "Plumber Dublin call now." "UpSearch pricing." Highest intent. Conversion rates are 5-10x informational.

To validate intent, run the actual query in an incognito Google window and observe:

• What types of pages rank in the top 5? (Blog posts? Product pages? Comparison pages? Tools?)
• What SERP features appear? (Featured snippet, PAA, shopping ads, local pack?)
• What does the dominant pattern tell you about searcher intent?

If the top 5 are all blog posts, the query is informational. Building a product page will not rank. If the top 5 are all product pages, the query is commercial, a blog post will not win. Match the page type to the intent of the SERP.

Organizing Keywords Into Pages

The final step is mapping your keywords to specific pages. The rule is one primary keyword per page, with related secondary keywords supporting it on the same page.

Primary keyword. The query the page is built to rank for. Goes in the title tag, the H1, and the first 100 words. The URL should reflect it.

Secondary keywords. Related queries with the same intent and topic. Goes in H2/H3 headers, body content, and internal anchor text. A page can rank for 10-50 secondary keywords if the topic coverage is good.

Cross-page keywords. Queries that are related but represent different intents. These go on different pages, with internal links between them.

The trap to avoid is targeting the same primary keyword on multiple pages, which causes keyword cannibalization. If you find yourself wanting to write two pages for the same query, consolidate into one stronger page.

For larger keyword sets, consider topic clusters: a pillar page covering the broad topic, plus several supporting pages targeting specific long-tails, all internally linked. The Marketing Hub Strategy tool generates these cluster maps from your existing data.

A Sample Keyword Research Workflow

Here is the full workflow applied to a fictional small business, a Dublin plumbing company.

Step 1: GSC export. Pull last 90 days of queries. Filter for impressions over 50.

Step 2: Quick wins. Identify queries in position 4-10. Examples: "emergency plumber Dublin 6" (pos 6), "boiler repair Dublin" (pos 4), "plumber near me Dublin" (pos 8). Three quick wins identified.

Step 3: Customer conversations. Mine recent calls and reviews. Note recurring phrases: "weekend emergency," "no hot water," "leaking radiator," "old house plumbing." These are query patterns.

Step 4: Tool expansion. Use a tool to find variations. Discover "weekend plumber Dublin," "no hot water Dublin," "Victorian house plumbing repair", all related to customer conversations.

Step 5: Intent validation. Run each query in incognito. Confirm intent. "Emergency plumber Dublin 6" returns service pages, commercial intent, build a service page. "How does a boiler work" returns blog posts, informational, build a guide if useful for content marketing.

Step 6: Page mapping. Map keywords to pages:

• /services/emergency-plumber-dublin-6 → "emergency plumber Dublin 6," "weekend plumber Dublin 6," "24/7 plumber Dublin 6"
• /services/boiler-repair-dublin → "boiler repair Dublin," "no hot water Dublin," "boiler not working Dublin"
• /blog/how-boilers-work → "how does a boiler work" (informational, content marketing only)

Step 7: Prioritize. Quick wins first (boiler repair and emergency Dublin 6 are highest impressions). Content gaps next. Long-tails last.

That is keyword research. Six steps. A few hours of work for a small business. Produces a 90-day content and optimization plan grounded in real data.

Common Keyword Research Mistakes

Five mistakes account for most failed keyword strategies.

Mistake 1: Starting with a tool, not your data. Tools are for expansion and validation, not ideation. Start with Search Console.

Mistake 2: Optimizing for search volume. A 10-search-per-month query that converts at 20% is worth more than a 10,000-search-per-month query that converts at 0.1%.

Mistake 3: Ignoring intent. Building a blog post for a transactional query is wasted effort. Always match page type to SERP intent.

Mistake 4: Targeting the same keyword on multiple pages. Causes cannibalization. One primary keyword per page, period.

Mistake 5: Treating keyword research as a one-time exercise. Search behavior changes. Run the workflow quarterly. The SEO roadmap framework bakes this into the cycle.

FAQ

How do I find keywords my competitors rank for?

Use a competitor analysis tool (UpSearch Competitor Intelligence, Semrush, or Ahrefs) for the systematic approach, or manually check the top 5 results for your target queries to see who consistently ranks. The competitor SEO analysis guide covers the gap analysis playbook.

Should I target high-volume keywords or low-competition keywords?

Both, in different proportions. Target high-volume head terms with strong commercial pages (homepage, main service pages) for long-term growth. Target low-competition long-tails with supporting blog posts and detailed pages for fast wins. A mix of 30% head terms and 70% long-tails is a typical small business split.

How many keywords should I target per page?

One primary keyword per page. The page can rank for 10-50 related secondary keywords as a side effect of good topic coverage, but only one keyword should drive the title, H1, and URL. If you want to target two distinct keywords, build two pages.

How long does it take to rank for a new keyword?

For new pages: 2-6 months for clear progress, 6-12 months for stable top-5 positioning. For optimizations to existing pages: 2-6 weeks. The variation depends on competition, domain authority, and content quality. Quick wins always rank faster than content gaps.

What is the difference between a keyword and a query?

A query is the exact string a searcher typed into Google. A keyword is the broader topic a page is built to rank for, which usually includes many related queries. Modern SEO is more about keyword topics than exact-match keywords, Google's understanding of synonyms and related concepts is strong enough that "blue widget price" and "how much do blue widgets cost" both reward the same well-built page.

Should I use AI for keyword research?

Yes, but only at the expansion and analysis steps, not the ideation step. Use evidence-led AI to mine your GSC data faster, summarize competitor patterns, and group keywords by intent. Do not use generic AI to invent keywords, it has no idea what your customers actually search for.