Key takeaway: In 2026, search intent beats search volume. A low-volume transactional keyword like "hotel management agency Montreux" can earn a booking, while a high-volume informational term like "what is hotel management" earns mostly zero-click impressions. Classify intent first, then prioritise.
Why the old volume-first method is failing
For years the workflow was simple: type a seed keyword into a tool, sort the results by monthly volume, and target the biggest numbers. That model is breaking down for a few concrete reasons.
First, a large share of informational searches now end without a click. When a user asks a question and an AI Overview, ChatGPT, or another assistant answers it directly in the results, the user never visits any website. The clicks left over are concentrated in fewer, higher-intent results. Second, a search can surface in many places in one day — Google, YouTube, Reddit, TikTok, and AI chatbots. A pocket of demand that looks tiny in one tool can be meaningful when you map it across every surface real buyers use. Third, and most importantly, volume says nothing about the intent behind a query. Thousands of people can search "what is SEO" every month, but almost none of them are ready to buy anything.
That is the core shift. You are no longer optimising for the person who types a keyword; you are optimising for the problem behind the keyword. The keyword is just a label for a moment of intent.
Step 1: Collect real customer language before touching a tool
Almost every keyword list fails because it starts from an analyst's assumptions instead of the customer's words. The best raw material for keyword research already lives in your business. Start there.
- Google Search Console. Export the queries that already bring you impressions and clicks. These are proof of existing demand, and many will be long-tail terms you never thought to target.
- Support tickets and sales calls. Write down the exact verbs and problems customers use. People describe the same service as "help with my taxes" rather than "accounting services." Those natural phrases are usually cheaper to rank for and far more relevant.
- Community platforms. Search Reddit, YouTube comments, and industry forums for your topic. Real objections, questions, and comparison statements surface here that no keyword tool will show you.
Gather a few dozen of these raw phrases. Treat them as seed material, not as the final list. They anchor everything else in language you know is real instead of language a tool invented.
Step 2: Expand with tools, then classify search intent
Run your seeds through a keyword tool — Ahrefs, Semrush, Google Keyword Planner, or a free alternative like Ubersuggest all work. You are looking for related terms, long-tail variations, and question-based queries. Cast a wide net here; you will narrow it down next.
Now the step that separates this method from the old one: classify every keyword by search intent before you look at volume. There are four main types:
- Informational. Someone learning, e.g. "how does SEO work." Needs a guide; rarely converts directly.
- Navigational. Someone looking for a specific site or page.
- Commercial. Someone comparing options, e.g. "best SEO agency for hotels." A review or comparison page is the right match.
- Transactional. Someone ready to act, e.g. "book SEO audit." Needs a pricing, booking, or checkout page.
Match the content format to the intent. An informational query should point to a guide, not a pricing page, and a transactional query should point to a page that lets the visitor act immediately. When format and intent align, conversion follows; when they misalign, even a top-ranking page underperforms.
Step 3: Check the AI visibility dimension
Ranking in the classic blue links is only half the job in 2026. If your page gets cited by ChatGPT, Perplexity, or a Google AI Overview, it earns visibility that no organic listing can give you. So add an AI check to your shortlist.
For each keyword, run three quick tests. Search it on Google and note whether an AI Overview appears and which sources it cites. Ask the same question to ChatGPT and Perplexity and check whether your site, or a competitor, gets mentioned. Then decide whether the current content on those queries is the kind of short, answerable passage an assistant would quote. Question-based keywords tend to trigger AI answers most often, which is exactly why understanding how AI assistants choose their sources is so valuable.
If your site is structured with clear headings and self-contained answers, it earns an AI citation profile. If it is not, that is an opportunity your competitors are missing.
Step 4: Score by realistic ranking odds, not just difficulty
Keyword difficulty is a useful signal, but it is a vanity number on its own. The question that matters is: given the strength of the sites already ranking, can you realistically get into the top results? For a new or small site, target keywords where the current top results are weaker or less authoritative than yours would be. A keyword with modest difficulty but a gap in quality is worth more than an "easy" keyword dominated by brands with enormous backlink profiles.
Be honest about your domain, too. Chasing a keyword that a giant brand owns with matching intent and flawless content is usually a slow, demoralising grind. The fastest wins come from long-tail terms and question-based queries where the competition is thin and the intent is clear — the same niches AI assistants are constantly citing.
Finally, build a small cluster instead of isolated pages. Choose one pillar topic, then create supporting articles that each target a related long-tail query linked back to the pillar. Search engines and AI assistants both reward this kind of connected, topical coverage. If you are unsure how to structure a comprehensive site that covers a topic well, our SEO audit checklist walks through the technical foundation beneath it all.
Step 5: Decide what a keyword is actually worth
Traffic is a means, not the end. Before you commit to any keyword, ask what one click on it might be worth to your business. A transactional keyword that leads to a booking or a signed service contract is worth a lot more than an informational keyword that will never produce revenue. This is why revenue per click, not monthly volume, is the metric to guide your shortlist.
For a service business like a local SEO or web design agency, this often means prioritising commercial and transactional queries even when their volumes are small. One client founded in this pattern — a hotelier searching for management help — is worth more than a hundred readers of a general guide. Local intent, in particular, tends to convert well because the searcher usually has an immediate, geographically relevant need; our local SEO guide for Swiss businesses covers how to capture it.
And remember that ranking is not the same as converting. A page that ranks but fails to turn visitors into enquiries is doing nothing for your bottom line. Pair your keyword research with sound conversion work, or you are simply paying to generate traffic that never pays you back — the subject of our website conversion fixes article.
What good keyword research produces
Done properly, this process gives you three things. A prioritised list of keywords you can realistically rank for, mapped to intent. A content plan where each page matches the format a searcher actually wants. And a citation strategy that earns visibility in both classic results and AI assistants. The output is not a spreadsheet of big numbers; it is a roadmap from search demand to revenue.
The landscape has changed, but the fundamentals have not. Find the problem behind the keyword, match your page to the moment of intent, and be honest about whether you can win. Do that and the volume obsesses over itself.
Frequently Asked Questions
Want to see whether your current keyword targets are the right ones? Run your site through the lens of our SEO audit checklist first, then bring the list to a conversation about content that actually converts.