Finding the right keywords means choosing the exact queries your future customers type when they are ready to take action. These are not necessarily the ones with the highest search volume. Targeting "shoes" attracts casual browsers. Targeting "waterproof safety work boots" attracts a buyer who already has their credit card in hand.
Keyword research is the process of identifying, filtering, and sorting your market's queries to assign each one to a specific page on your site. When done properly, it serves two engines at once: Google and AI answer engines.
Key takeaways
- A good keyword ticks three boxes: intent aligned with your offer, reachable competition, and visible commercial value, such as through cost per click (CPC).
- Every query falls into one of 4 intents: informational, commercial, transactional, or navigational. A single page should serve only one.
- The basic rule: 1 main keyword per page, plus its close variants, to prevent two of your pages from cannibalizing each other's rankings.
- According to Google, about 15% of daily searches have never been seen before. Long, precise questions remain a wide-open playing field.
- A new or lesser-known site ranks faster for long-tail keywords (3 words and more) than for generic terms locked down by major players.
What does finding the right keywords actually mean?
A good keyword combines three conditions: an intent that matches what you sell, competition your site can beat, and measurable value for your revenue. Search volume is merely one indicator among others. A high-volume term with no purchase intent fills your analytics reports, not your bank account.
This is the difference between traffic and qualified traffic. The former flatters reporting. The latter pays the bills.
Consider two queries. "Invoicing software" is broad: students, accountants, casual browsers, competitors. "Invoicing software for craftspeople with mobile quotes" is narrow, but every person typing it has a specific need and a budget. They are past the browsing stage. They convert much better.
Keyword or semantic search?
Search engines no longer read isolated words. They understand topics. Semantic search therefore involves mapping out an entire subject: the main query, its synonyms, related questions, and connected entities (brands, standards, materials, uses). A page covering this full scope is deemed more useful than one repeating a single term ten times.
Google states this clearly in its guidelines on creating helpful, people-first content: content written solely to manipulate rankings is not what gets rewarded.
How does an effective keyword research process work in 4 steps?
Effective keyword research follows four steps in this exact order: listing seed keywords, expanding the list, filtering by metrics, and then sorting by intent and grouping by page. Skipping step 4 is the most frequent mistake. It leads to cannibalizing pages and content that answers no one's needs.
Step 1: list your seed keywords
Start from your business, not a tool. Your seed keywords come from three sources:
- your product or service catalog, using the names your customers use rather than internal designations;
- your personas: their job, their problem, their industry;
- your sales and support teams, who hear real questions every single day.
Aim for 10 to 20 seed keywords. No more. The next step will multiply them.
Step 2: expand the list
Each seed keyword turns into a family of dozens of variants. Use Google autocomplete, the "People also ask" box, AnswerThePublic for questions, and Google Search Console for queries where your site already appears. Add a gap analysis: queries where your competitors rank and you do not. Tools like Semrush or Ahrefs automate this part.
Step 3: filter by metrics
Three figures matter. Monthly volume indicates demand. Keyword difficulty (KD) indicates the effort required to reach the first page. CPC indicates what advertisers are willing to pay, reflecting the commercial value of the click. Add Google Trends to spot seasonality: a query that peaks in November might be dead in March.
A high CPC with moderate difficulty often signals a goldmine. High volume with near-zero CPC often signals traffic with no buyers.
Step 4: sort by intent and group
Categorize every query by its intent: "how to fix" (informational), "best comparison" (commercial), "buy online" (transactional), "brand login" (navigational). Then group close variants into clusters. Each cluster receives a single page with a primary keyword and secondary variants.
For an ecommerce site, this breakdown directly structures your category tree and product pages, as we explain in our guide to ecommerce SEO.
Step-by-step example: from "shoes" to a query that converts
A safety equipment distributor targeting "shoes" has virtually zero chance of ranking, and the rare visitors who arrive are looking for sneakers. By adding three specifics (use, constraint, context), the query shifts from a vague audience to an identifiable buyer ready to order.
Here is how the query narrows down step by step.
| Query | Probable intent | Competition | Who types it | Page to create |
|---|---|---|---|---|
| shoes | vague | crushing (major retailers) | everyone | none |
| safety shoes | commercial | strong | buyers and browsers | main category |
| waterproof safety shoes | commercial / transactional | medium | buyer with a specific constraint | sub-category |
| waterproof safety shoes for construction sites | transactional | low to medium | site manager, construction buyer | dedicated page with selection and standards guide |
| what standard for safety shoes on wet construction sites | informational | low | buyer in the decision-making phase | blog post linked to the dedicated page |
Look at the last column. Each useful row corresponds to a page, and the pages link to one another. The informational article captures the buyer at the question stage. The dedicated page captures them at the purchase stage.
The real win is here: the fourth query has far less search volume than the second, but the person typing it has already decided what they want. They are comparing two or three offers, not fifty. And this is precisely the type of precise question an AI answer engine like Perplexity reformulates when recommending a supplier.
The translation trap
If you sell across multiple markets, do not simply translate this list. A German buyer types "Sicherheitsschuhe", a British buyer often uses "safety boots", and an American buyer says "work boots". Same products, different queries, different volumes. A word-for-word translated list targets queries nobody types. A native list starts from the actual search behavior of each market.
How to choose keywords that ChatGPT, Gemini, and Perplexity pick up
AI answer engines like ChatGPT, Gemini, and Perplexity process complete natural language questions that are longer and more contextualized than traditional Google searches. To get cited, you must target entire questions, then answer them right in the first sentence of the section in an autonomous and factual manner.
This is the boundary between SEO and GEO. SEO aims for a ranking in a list of links. GEO aims for a citation in a generated answer. Both start from the same keyword research, but GEO requires three additional habits.
- Turn your clusters into questions. "Waterproof construction safety shoes" becomes "What safety shoes should I choose for a wet construction site?". Use these questions as section headings.
- Answer first, elaborate second. An AI extracts a passage. If it begins with "In this section, we will see", it is useless. If it begins with a fact (a standard, a criterion, a figure), it is citeable.
- Clearly name entities. Your brand, your products, standards, use cases. An AI cites what it can identify without ambiguity.
Then comes measurement. A Google ranking is verified in Search Console. An AI citation is verified by actually querying ChatGPT, Gemini, and Perplexity on your target queries repeatedly. The rest is theater: a inflated "visibility score" without the underlying prompts and answers proves nothing. We detail this shift in our analysis of SEO to GEO.
Who is this approach essential for, and who can skip it?
Keyword research is essential whenever your growth depends on customers who do not yet know you: a new website, an online store, expansion into a new market. It becomes secondary when your customers already find you by name or when nobody is searching for your type of product yet.
It is essential if…
- Your site is young or has few backlinks. You will not beat market leaders on generic terms. The long tail is your entry point, and only structured research reveals it.
- You sell online. Every category and product page must match a real query.
- You are opening up new markets. Every language has its own queries. Without local research, you are publishing blindly.
- You publish content without results. Often, the content is fine but targets queries with no demand or purchase intent. A 7-step SEO audit quickly highlights this.
You can relegate it to the background if…
- Your business relies entirely on local word of mouth, with a full order book and a few brand queries. A well-maintained Google Business Profile is often enough.
- You sell B2B to a handful of known accounts. Your buyers do not use open search. Direct outreach carries more weight.
- Your product creates an entirely new category. If no one is searching for the need yet, there is no demand to capture. Create it elsewhere first, then return to keywords once demand emerges.
Before investing, do the math: how much is a customer worth, and how many customers can a page generate. Our method to calculate your SEO break-even point gives you this number in an hour.
Frequently asked questions
How many keywords should you target per page?
One primary keyword per page, accompanied by a few secondary variants and related questions. If two of your pages alternate for the same query in Google Search Console, it is a sign of cannibalization. Merge them or redistribute their targets.
What free tools can you use to find keywords?
Google Search Console, Google Keyword Planner, Google Trends, Google autocomplete, and AnswerThePublic cover the essentials. Keyword Planner requires a Google Ads account, but no active spending. Without an active campaign, it displays volume ranges rather than precise figures.
Is it better to target high search volume or the long tail?
A new site should start with the long tail: less competition, more qualified visitors, faster conversions. Generic terms come later once your site has built authority through already-ranking pages. The two strategies complement each other; they do not oppose one another.
How often should you redo your keyword research?
Review your list at least once a quarter, and before every seasonal peak spotted in Google Trends. Queries evolve, new questions emerge, and AI answer engines change their sources. A list left untouched for a year often points to outdated pages.
Take action: your one-week keyword calendar
Finding the right keywords is useless if the list sits forgotten in a spreadsheet. Here is a plan to turn it into published pages starting next week.
- Day 1: export three months of queries from Google Search Console. Identify those where you rank between position 8 and 20. These are your quickest wins.
- Days 2 and 3: list 10 to 20 seed keywords, expand them, and then filter by volume, difficulty, and CPC.
- Day 4: categorize every query by intent and group them into clusters, one cluster per page.
- Day 5: rephrase each cluster into a natural question—the kind you would ask ChatGPT or Perplexity.
- Days 6 and 7: schedule one page per week on a calendar, starting with low-competition transactional queries.
If you prefer not to do this manually, that is what we do. UPGeoSEO runs a free SEO + GEO audit of your site, builds your market-by-market keyword calendar, writes native content in every language matching your brand voice, publishes on WordPress, and tracks your actual citations in ChatGPT, Gemini, and Perplexity. You approve, we publish. Start with the free audit and see which queries you can win as early as this month.
