Imagine Markus. Mechanical engineering, 80 employees, a solid business. Markus thinks he’s done everything right over the past year: a new website – fast, attractive – and even a blog. His stated goal is printed out and pinned to the whiteboard: “Number 1 on Google for industrial automation”. Three months, a change of service provider and a four-figure sum later, his site is in 38th position. Right above him: Siemens, Bosch, a specialist publisher with 200,000 backlinks.
Markus is battling opponents with corporate budgets for a single word. A word that his customers don’t actually type in at all.
Two streets away, a significantly smaller competitor ranks number 1 – for “retrofit control old plant costs”. No glamour, 90 search queries a month. But every single one of them is a buyer with a specific problem and a budget. The smaller player wins the enquiries that Markus never even sees.
That is the whole difference between a big keyword and a winnable one.
A keyword isn’t valuable simply because lots of people search for it. It’s valuable when the right people are searching for it and you can realistically rank for it.
Why the biggest keyword is almost always the wrong target
There is a reflex that ruins almost every initial keyword idea in small and medium-sized enterprises: You take the one word that best describes your service – such as ‘online shop’, ‘tax advice’ or ‘CRM software’ – and declare it your target. That’s understandable. But it is precisely these generic terms that represent the most fiercely competitive, most expensive and least purchase-ready territory in the entire search index.
The figures behind this are clear. A much-cited Ahrefs analysis of billions of keywords comes to a simple conclusion: Around 93 per cent of all keywords in Ahrefs’ US database have fewer than ten search queries per month. These unassuming long-tail terms thus make up the overwhelming majority of all existing search queries. The few high-volume head terms are the loud tip of the iceberg. Taken together, the long, quiet mass beneath offers far more entry points than the tip could ever provide.
Google itself provides the second half of the evidence: around 15 per cent of all daily search queries are ones the search engine has never seen before. Google explicitly confirmed this figure in 2025. It has even remained stable, whilst queries have become more conversational and longer thanks to AI. In other words: your next customer is currently reformulating their search, in their own words, based on their specific problem. No one types in ‘industrial automation’ and then pulls out their credit card.
Over 90 per cent of all Google searches are long-tail. Your ideal customer doesn’t type in a generic term. They type in a specific question.
For an SME, this isn’t bad news – it’s the best news there is. You don’t have to beat Siemens. You just need to find the 200 specific terms that Siemens doesn’t even compete for, but which your customers are still searching for.
Three questions determine whether a keyword is winnable
‘Winnable’ sounds like a gut feeling. It isn’t. Whether a search term is worth your while can be broken down into three straightforward questions. Every serious keyword passes through all three.
Firstly: What does the searcher really want? This is the search intent, and it trumps search volume every time. The SEO world has been categorising search queries into four types for years. Informational: someone wants to learn (“what is a retrofit”). Navigational: someone is looking for a specific brand or website. Commercial: someone is comparing options before making a purchase, recognisable by words such as ‘best’, ‘comparison’, ‘test’, ‘alternative’ and ‘provider’. Transactional: someone wants to take action now, using terms like ‘buy’, ‘price’, ‘cost’ and ‘near me’. For an SME looking to generate enquiries, the last two are gold. ‘CRM software’ is informational and non-committal. ‘CRM for a trades business with time-tracking costs’ indicates a specific purchase intention.
Secondly: How fierce is the competition? This is precisely what Keyword Difficulty is for – a metric ranging from 0 to 100 that tools such as Sistrix, Ahrefs or Semrush estimate based on the strength of the pages already ranking, primarily from their backlink profile. As a rough guide for a medium-sized business without significant domain authority: terms below around 30 are realistically achievable; anything above 60 is a waste of energy for years on end, as long as you don’t have the backlinks of the big players. You’re looking for the sweet spot: high purchase intent, acceptable search volume, manageable competition.
Thirdly: Is the search volume accurate, and how do you interpret it correctly? Search volume is an estimate, not a speedometer. 90 searches a month sounds paltry until you realise that these are 90 buyers with the same pressing problem. Ten such terms outperform a single top-level term with 10,000 searches, the vast majority of which are just people browsing. Always interpret volume in conjunction with intent, never in isolation.
The tools: what you use to search systematically
You don’t need a €400 subscription to get started. You need a sensible sequence.
Start with your own data. The Google Search Console is the most underrated keyword tool of all, and it costs nothing. It shows you which terms you’re already getting impressions for without ever having optimised for them – often terms ranked between positions 8 and 20 that, with a bit of work, can move up to page 1. These are the quickest wins you’ll find, because Google already takes you reasonably seriously for these topics.
Next comes the groundwork. The Google Keyword Planner provides related terms and volume estimates, with one honest caveat: without actively running ads, you’ll only get rough ranges rather than precise figures, and you’ll need a Google Ads account. For the DACH region, there’s hardly any way round Sistrix; this free keyword tool shows related searches and, much more valuable, the actual questions users are asking about your topic. Tools such as AlsoAsked or AnswerThePublic turn a seed term into a whole tree of questions, and Google Trends tells you whether a topic is on the rise or on the wane.
The one trick most people forget: Google reveals long-tail keywords for free. Type your term into the search bar and read the autocomplete suggestions. Scroll down to ‘Similar questions’ and ‘Similar searches’ at the bottom of the page. This is Google’s own insight into what people are actually typing in: free, unfiltered, straight from the source.
Since AI Overviews, it’s no longer just about volume
In 2026, a new factor will come into play that older research guides simply don’t cover. For an increasing number of search queries, Google is displaying an AI-generated answer directly above the results: the AI Overview. Fewer people are clicking through as a result.
Sistrix has analysed this across over 100 million German keywords in February 2026. The result should reshape every keyword strategy: an AI Overview appeared in around 20 per cent of all German search queries in this analysis – a figure that Sistrix most recently confirmed in July 2026. Where it appears, the click-through rate for the top result drops from 27 per cent to 11 per cent. A drop of almost 60 per cent. In total, this costs websites in Germany an estimated 265 million clicks per month. And in almost 79 per cent of cases, the AI response ranks above the organic results.
Search volume alone has never told the whole story. Since the advent of AI Overviews, it doesn’t even tell half the story.
What does this mean in practical terms for your list? Take two terms that look almost identical in the tool. ‘what is crm’ has decent volume, but is purely informational: the AI Overview answers the question in full, so the user never scrolls down to your page. The click is dead before you’ve even ranked. “crm comparison tradesmen providers” has a fraction of the volume, but behind it is someone making a purchasing decision who wants to see real providers, not a machine-generated summary. Two years ago, you would have sorted by volume and placed the informational term at the top. Today, you flip the list around.
The strategic implication is the same as before, only more urgent: shift your focus away from broad informational terms towards specific, purchase-oriented long-tail keywords. And, at the same time, plan to be cited in the AI responses in the first place. That is the new visibility issue, which we examine in the post on AI visibility.
The new Gen-AI report in Search Console – and what it’s not telling you
On 3 June 2026, Google launched the Search Console reports for generative AI features. For the first time, they officially show you how often URLs from your website appear in AI Overviews and in AI Mode. For keyword research, this is the most relevant new data point in years – and at the same time the one most prone to misinterpretation.
According to Google’s own documentation, the report has exactly four dimensions: Pages, countries, devices, date. No search queries. And no clicks – only impressions. So you’ll find out that one of your pages is being used in AI responses, but not in response to which question. It is precisely this mapping of keywords to pages – the foundation of traditional search – that ends here. What’s more, Google is initially rolling out the report to only a subset of properties. If it’s missing from your property, this isn’t a configuration error.
The fact that the search query dimension is missing isn’t an oversight, but a natural consequence of the nature of the service. Google itself states that an average AI Mode search is three times as long as a traditional search query. Such fully formed sentences are practically never repeated word for word. In its long-tail analysis, updated in May 2026, Ahrefs estimates that over 95 per cent of these conversational search queries have no measurable search volume. A keyword tool cannot identify them at all; there is simply nothing to count.
In practice, therefore, you will be working on two levels from now on. At keyword level, the research from the previous sections remains valid: your customers will continue to type purchase-related long-tail terms into the standard search bar, and in Germany they are more likely to click on these results than elsewhere. 62.1 per cent of German Google searches end without a click; in the UK, the figure is 69.5 per cent (SparkToro/Similarweb, January to April 2026). At page level, place the Gen AI report alongside the traditional performance report and compare by URL. Pages with high AI impressions and simultaneously declining organic clicks are analysed in AI responses without you actually getting the visitor: candidates for topic clusters closer to the point of purchase. Pages with no AI impressions do not appear at all in the new response layer. Both findings should be included in the next quarterly update of your keyword list – not as a replacement for research, but as a second metric alongside it.
Your systematic process: from gut feeling to list
This is the process you can actually repeat, every quarter, for every service.
You start with seed terms: five to ten words that describe the core of your business. You then expand on these. Search Console for what’s already up and running, Keyword Planner and Sistrix for the bulk of the data, and Google’s own suggestions and ‘Similar questions’ for the actual phrases your customers use. Now you have a long, raw list. And this is precisely where systematic research differs from mere gathering.
You sort each term by intention and group it into topic clusters, i.e. related queries centred around a purchasing topic. For example, around ‘retrofit control’ (the modernisation of old machine control systems), terms such as ‘PLC retrofit costs’, ‘subsidies for modernising old plants’ and ‘retrofit providers in NRW’ end up in the same cluster: the same purchase intent, different phrasing. These clusters aren’t just for organisation; they’ll later form the framework for your content. Then you filter ruthlessly: terms with a clear purchase intent, manageable difficulty and realistic search volume remain. Purely informational head terms – where you’d lose out to both large corporations and AI-generated answers – are discarded.
In the end, you prioritise. Not alphabetically, but according to what brings in money: which terms have the highest purchase intent with the lowest difficulty? That’s your starting list for the next three months. The rest goes into the backlog.
And Markus? Had he run this process before setting his ambitious whiteboard target, the board wouldn’t have read ‘No. 1 for industrial automation’, but rather a list of thirty small, unattractive, highly profitable terms for which he can realistically rank. It’s less glamorous. It just works.
Systematic keyword research isn’t magic – it’s a method. This is exactly where our work in digital marketing & SEO begins: first finding the keywords we can rank for, then building everything else on top of that. The method decides the rest, not the budget.