How to find keywords your customers actually use

Most keyword research fails before anyone opens a tool. The problem is not access to data. It is starting with the wrong language. If you begin with industry jargon, internal product names, or whatever a generic SEO tool suggests first, you end up building pages for terms your customers do not actually use.

The practical way to do it is simpler. Start with the jobs customers need done, collect the words they already use, expand those into a working list, check what those searches really mean, then choose the terms you can realistically win. That is the core of how to do keyword research well, especially for small and medium businesses with a local-language website and no in-house SEO team.

Start with the jobs customers need done

Before Google, Search Console, or any keyword database, start with evidence from customers.

Ask one basic question about each service or product you sell. What is the customer trying to get done?

Not your category name. Not your feature list. The actual job.

A few examples:

• A payroll firm does not only sell “payroll services”
• It helps an owner “run monthly payroll without mistakes”
• A B2B software company does not only sell “workflow automation”
• It helps a team “stop chasing approvals in email”
• A dental clinic does not only offer “implantology”
• It helps someone “replace a missing tooth”

Those job statements give you better seed keywords than your brochure does.

Start in four places.

First, sales calls. If you record calls, review them. If you do not, ask the sales team for recent notes. Look for repeated phrases around pain, urgency, and comparison. Customers often say things like:

• “We need this to work with Stripe”
• “Can this connect to Shopify?”
• “We need a German version”
• “We want something for small hotels, not enterprise”
• “We need a local supplier in Ireland”

Second, emails and contact forms. These are useful because people write in their own words. Export a few months of enquiries. Highlight nouns and verbs that recur. Pay attention to modifiers such as “near me”, “for accountants”, “same day”, “for export”, “in Polish”, “for WordPress”, or “for small business”.

Third, your product and service pages. Not because they are right, but because they show the promises you already make. Pull out the core service terms, industries served, locations, integrations, and compliance topics. Then check whether customers use the same language. Often they do not.

Fourth, support tickets and onboarding questions. These are especially good for article ideas. Service pages target buying intent. Support language often reveals informational searches, such as setup steps, comparisons, errors, and process questions.

Put all of this into one sheet. One row per phrase. Add columns for:

• Source
• Exact wording
• Service or product
• Customer type
• Country
• Language
• Likely intent
• Notes

At this stage, do not tidy too early. Keep rough phrasing. If five customers say “booking system for guesthouse” and none say “hospitality reservation platform”, that tells you something important.

For businesses outside English-speaking markets, this matters even more. Direct translation usually loses the wording people actually search with. In German, Polish, Swedish, Latvian, or Finnish, buyers often use different compounds, case forms, or local business terms than a translated English keyword list would suggest. That is one reason we research in the site’s own language and publish natively, not by translating English content afterwards. We covered that in more detail in what keyword research means for a small business site.

Turn rough ideas into a usable keyword list

Once you have seed terms from real customer language, expand them.

Google suggestions are still one of the fastest ways to do this. Type a seed phrase into Google and note:

• Autocomplete suggestions
• “People also ask” questions
• Related searches at the bottom
• Variants with location, price, industry, or problem words

If your seed is “accounting software for restaurants”, you may see variants around small business, VAT, stock control, payroll, or country-specific terms.

Use Search Console next. This is where many small businesses already have useful data and simply do not look at it properly. In Search Console, open Performance and review queries for the last 3 to 6 months. Filter by page if needed.

Look for:

• Queries with impressions but low clicks
• Queries where you rank on page 2 or 3
• Queries landing on the wrong page
• Long-tail searches you did not know you were appearing for

If one service page gets impressions for “warehouse barcode software” and “stock scanner app”, that is a clue. The market may use both. You might need clearer copy, a stronger heading structure, or separate pages.

Then review competitor pages. Not to copy them, but to see how the market is segmented. Search your seed terms and inspect the pages that consistently appear. What words do they use in title tags, headings, FAQs, and anchor text? Are they targeting industries, locations, use cases, or product types that you have ignored?

Be careful here. Competitors can be wrong too. If they all use imported English jargon on a local-language site, that does not make it good keyword research.

Local language variants matter. In the UK and Ireland, this may mean “solicitor” versus “lawyer”, “quote” versus “estimate”, or regional service wording. Across the EU, the differences are often larger. Germany, Austria, and Switzerland may use different preferred terms. The Nordics often mix English product terms with local-language problem statements. In the Baltics and Poland, inflected forms and local business terminology can change the way a phrase is searched.

Build the list in clusters, not as one long dump. A practical structure is:

• Core service
• Problem
• Industry
• Location
• Comparison
• Integration
• How-to
• Compliance or regulation
• Brand and competitor terms

For example, if you sell e-commerce accounting software, a cluster might include:

• accounting software for Shopify
• accounting software for WooCommerce
• VAT for online shop
• bookkeeping for Amazon sellers
• e-commerce accountant Germany
• inventory accounting software small business

If you are an agency managing several sites, keep separate tabs by client and language. If you want a clearer process for multilingual and local-language SEO work, our guide on how a web agency should handle SEO for local-language sites sets out the workflow we use.

Check intent before you check volume

This is where many keyword lists go wrong. A phrase can have plenty of searches and still be useless to you if the searcher wants something else.

Intent comes first.

For each keyword, ask what the searcher is trying to do. Usually it falls into one of these groups:

• Learn something
• Compare options
• Find a provider
• Buy now
• Solve a problem
• Reach a specific site or brand

Check the search results manually. Search the phrase in the target country and language. Then look at what ranks.

If the first page is mostly guides, the intent is informational. If it is mostly service pages, category pages, or product pages, the intent is commercial or transactional. If the results are maps, local packs, and local firms, location matters heavily. If the results are dominated by government pages, directories, or major marketplaces, ranking may be harder than the keyword tool suggests.

A simple example. “How to register a company in Poland” and “company registration service Poland” are related, but they do not want the same page. The first likely needs a guide. The second needs a service page.

This is especially important when you are deciding whether to publish an article or improve a money page. We often see small businesses write blog posts for queries that clearly want a service page, then wonder why rankings do not convert.

Sort your list by page type:

• Service page
• Product page
• Category page
• Location page
• Article
• Comparison page
• Help or documentation page

Then remove mismatches. If you cannot serve the intent properly, drop the keyword for now.

Also check whether the wording signals research mode or buying mode. Words like “best”, “software”, “agency”, “near me”, “pricing”, “cost”, “service”, and “for small business” often suggest stronger commercial intent. Words like “what is”, “how to”, “why”, and “examples” usually sit earlier in the journey.

There is no hard rule, but there is a clear habit worth building. Before you look at search volume, search the term yourself and decide what kind of page deserves to rank.

Choose keywords you can realistically rank for

Now you can narrow the list.

A small business should not chase every broad term in the market. Relevance beats raw volume. A keyword with lower search demand but clearer fit often brings better leads and better conversion.

Use five filters.

First, relevance. If the term does not describe what you actually sell, cut it. If you provide implementation services for ERP systems, ranking for a broad “ERP software” term may bring the wrong audience.

Second, difficulty. Most tools provide some form of keyword difficulty score, but do not treat it as fact. Check the search results yourself. If page one is full of national brands, software giants, government sites, or long-established publishers, that is a warning. If the results include smaller firms, niche providers, and local businesses, the term may be realistic.

Third, location. Decide whether the query is local, national, or international. “Accountant in Cork” is not the same as “tax adviser Ireland”. In the EU, country targeting matters because search behaviour, legal terms, and buying expectations differ. A page written for Germany will not automatically satisfy Austria or Switzerland, even in the same language.

Fourth, language. Do not assume one translated page can cover all variants. If buyers in Sweden use one phrase and buyers in Norway use another, treat them separately. The same goes for multilingual countries and regions. In the UK, language variation is usually smaller, but terminology still changes by sector.

Fifth, page type. Ask whether you already have the right page, or whether you would need to create one. If the keyword deserves a service page and you only have a blog post, your ranking chance is lower until the site structure matches the query.

A quick scoring model helps. Give each keyword a score from 1 to 3 for:

• Relevance
• Intent fit
• Ranking realism
• Business value
• Existing page readiness

Then total the score. This is rough, but it stops you prioritising a high-volume term that you have little chance of winning.

For smaller sites, the best opportunities are often:

• Long-tail service terms
• Industry-specific variants
• Location-specific phrases
• Integration terms such as WordPress, Shopify, Webflow, or Ghost
• Problem-led searches with clear commercial intent
• Queries where you already get impressions in Search Console

If you want to see which other sites are taking the attention you should be getting, our guide on finding the rivals stealing your best search traffic is a good next step.

Build a simple plan you can publish from

A keyword list is only useful if it turns into pages.

The simplest plan is a keyword map. One sheet, with one primary keyword cluster per page.

For each existing or planned page, list:

• URL
• Page type
• Primary keyword
• Secondary keywords
• Search intent
• Target country
• Target language
• Internal links needed
• Status
• Publish or update date
• Results after 30, 60, and 90 days

Start with your highest-value pages. Usually these are service pages, category pages, and key product pages. Make sure each has one clear primary topic. If one page tries to rank for three unrelated jobs, split it.

Then add supporting articles. Good article topics usually do one of three things:

• Answer a pre-sales question
• Help the service page rank by covering related subtopics
• Capture searches from buyers who are not ready yet

For example:

• Service page: payroll outsourcing for small business
• Supporting article: when to switch from in-house payroll to outsourced payroll
• Supporting article: payroll mistakes that cause HMRC problems
• Supporting article: outsourced payroll pricing, what affects cost

In the UK, tax and employment content should be careful with local terminology and authority. In EU markets, make the same adjustment for local law, VAT treatment, and country-specific procedures. Where rules differ between the UK and the EU generally, separate the content rather than forcing one page to serve both.

Prioritise quick wins first:

• Pages already ranking between positions 8 and 20
• Queries with good impressions in Search Console
• Service pages with weak copy but strong demand
• High-intent local terms with weaker competition
• Articles that can support a money page quickly

Publishing cadence matters less than consistency. One useful service page update and one solid article per month will usually beat a burst of generic AI content that says nothing new.

That is one place many businesses have been burnt. They were sold volume. Fifty translated articles. No local phrasing. No intent check. No review by someone who writes the language properly. The output looked busy and produced little.

Our approach in Seonis is narrower and more checkable. We research keywords in the site’s own language, write and publish natively, run each piece through a language gate, and report in the owner’s language. We also track whether buyers asking tools such as ChatGPT, Gemini, Perplexity, Claude, and Grok actually see your business named in answers. That does not replace search tracking, but it gives you another measurable view of visibility.

Finally, track results with patience and a short feedback loop. Review monthly:

• Which pages gained impressions
• Which queries moved up
• Which pages got clicks but no leads
• Which pages attract the wrong intent
• Which articles assist service-page traffic

Then update the plan. Keyword research is not a one-off task. Products change. Customer language changes. Competitors change. Google changes. AI answer engines change too.

If you are an agency doing this across multiple client sites, the process needs to stay simple enough to repeat. If you are a business owner, it needs to produce pages you can actually publish, not a spreadsheet that never leaves the folder.

Done properly, keyword research is not about finding the biggest term in the market. It is about finding the language your buyers use when they are close enough to act, then building the right page for that search. That is what brings the right traffic, and gives you something you can measure.