Semantic Keyword Research: The Question Map Method

Semantic keyword research is not the collection of individual phrases into a list — it is mapping every question, intent and concept inside a topic. The output is not a 500-row spreadsheet but a structural plan showing which page answers which group of questions.

The difference is easy to state. Classic research answers "how many times a month is this searched?". Semantic research adds a second layer — what does the person asking actually want to know, what do they search before and after, and which entity is the question attached to? The theory behind this sits in the semantic SEO and entity SEO articles.

What is wrong with the traditional list

The output of ordinary keyword research usually looks like this: "seo service", "seo services", "seo service price", "professional seo service", "cheap seo service". Five rows — and most of them are the same entity and very nearly the same intent.

Working from that list creates three problems:

• Page bloat. Creating a separate page for every row leads straight to cannibalisation.
• Intent blindness. "seo pricing" and "what is seo" may sit side by side in a spreadsheet, but they are entirely different people: one is ready to buy, the other is still learning.
• Coverage gaps. Dozens of questions that belong to the topic but never made the list are missed — and those are the main source of long-tail traffic.

Google's documentation on AI features makes this more pressing: the system uses a query fan-out technique, issuing several related searches behind a single query. A site that covers a topic broadly and deeply matches more of those additional searches.

Search intent: four types

Before processing a query you have to establish its intent. In practice four types are enough:

There is a simple way to establish intent: search the query on Google and look at the format of the results. If the first ten results are blog posts, Google treats it as an informational query — bringing a sales page to that fight will almost never work.

The five stages of semantic research

Stage 1 — The central entity. Write it in one sentence: "This cluster is about technical SEO." The more precise the entity, the easier every following step becomes.

Stage 2 — The entity network. List 20 to 40 concepts connected to it. For technical SEO: robots.txt, sitemaps, canonical tags, indexing, crawl budget, Core Web Vitals, JavaScript rendering, hreflang, 301 redirects, status codes, structured data, mobile usability.

Stage 3 — Harvesting questions. Collect real questions for each concept. Five templates work in almost any topic: what is it · how is it done · why is it needed · which tool checks it · how long or how much.

Stage 4 — Intent sorting. Label every question with one of the four types above.

Stage 5 — The page plan. Group questions that share an intent and require the same answer, then assign each group to one page. One page = one question group.

Free and reliable sources

Paid tools speed things up but are not required. The most valuable data is free:

Search Console → Performance report. Per Google's documentation the report supplies four metrics — clicks, impressions, CTR and average position — and allows filtering across six dimensions: query, page, country, device, search appearance and date. It shows which queries your site already appears for, which is the best possible starting point for finding gaps.

The same document notes two limitations: search results vary "by time, place, device, and recent history of the person searching", so a query in the report may not show your site when you run it yourself; and the newest data is preliminary and may still change.

The search results themselves. The "People also ask" block, autocomplete suggestions and related searches at the foot of the page reflect Google's own understanding of the query space directly.

Customer questions. Sales threads, phone calls, social media comments. This source is almost always overlooked, yet it carries the most accurate wording. Our own SEO questions section is built from exactly this kind of collected question.

Assigning queries to pages

This stage turns research into a plan. A simple table is enough:

The "Target URL" column doubles as your internal linking plan: blog posts should point to the relevant service pages, and service pages should point back to the blog posts that go deeper.

Example: mapping "technical SEO"

A simplified map of a real topic looks like this:

Every row is a potential page — but they should not all be written at once. The rule of priority is simple: start with the intent closest to the business goal, then work outward into informational queries.

Preventing cannibalisation

Cannibalisation is when several of your own pages compete for the same query. Google cannot settle on which to show, and both perform worse.

How to spot it: in Search Console open Performance, filter by the query, then switch to the Pages tab. If two or three URLs appear for one query, you have a problem.

How to fix it:

• Merge. Turn two weak articles into one strong one and 301-redirect the old URL to the new one.
• Differentiate. If the pages really are distinct subtopics, separate the titles and content much more clearly.
• Signal with internal links. Use anchor text to indicate which page is the primary one.

How to use queries in the copy

Once the queries are collected, the biggest risk is stuffing them into the text artificially. Google's spam policies name keyword stuffing explicitly: "repeating the same words or phrases so often that it sounds unnatural".

The right approach:

• Build headings from questions. H2 and H3 headings phrased as real questions help the reader and perform better in fragmented answer formats.
• Answer first. Open every section with a direct two- or three-sentence answer.
• Use natural variation. Different expressions of the same concept — synonyms, abbreviations, question forms — should appear naturally, not as a quota to fill.
• Add numbers and specifics. Google's helpful content documentation emphasises original information, evidence and clear sourcing.