Why Keyword Tools Miss Luxembourg Demand

There is a number in Luxembourg's labour statistics that quietly invalidates most keyword research done for this market. At the end of 2025, cross-border commuters filled 47% of roughly 494,000 salaried jobs, according to STATEC's Regards 02/26. In February 2026 alone, 233,022 people crossed a border to come to work. France supplies around 54% of them, with Belgium and Germany making up most of the rest.

Read that as a labour statistic and it is interesting. Read it as a search statistic and it is a problem.

Because nearly half the people who work in Luxembourg do not search from Luxembourg. They search from Thionville and Metz, from Arlon and Bastogne, from Trier and Saarbrucken. They search on google.fr and google.de and google.be. And when a keyword tool is set to Luxembourg as the target location, most of that demand is not measured as Luxembourg demand at all. It is filed under three other countries, or it disappears into volume bands too small for the tool to report.

This article is about what to do instead. It draws on public statistics that are genuinely available for this market, and it is equally explicit about the things nobody has published, because the second list matters as much as the first.

Why a keyword export for Luxembourg comes back thin

Anyone who has pulled planner data for Luxembourg has seen the same pattern. Head terms return volumes that look implausibly small for a wealthy economy. Long-tail commercial phrases return zero. Whole categories that obviously have buyers appear to have no searchers at all.

Three things are happening at once, and they compound.

The population base is small before you split it

Luxembourg had 690,959 residents on 1 January 2026, growing 1.3% over the year according to STATEC. That growth rate has been slowing: it was 1.5% the year before, and it ran between 2.0% and 2.5% through much of the last decade. Internet penetration is close to saturation at about 98.8% at the end of 2025 by DataReportal's estimate. So the online population is essentially the whole population.

Now divide that population across four languages. Then split it by commercial intent. The resulting cells are small enough that keyword tools, which round and suppress low volumes, will report many of them as zero. A zero in that context does not mean nobody searches. It means the tool declines to tell you.

Half your buyers are geolocated somewhere else

This is the bigger effect, and it is specific to Luxembourg in a way that few other markets replicate. A recruitment firm, a fiduciary, a private clinic, a car dealership, a language school: each of these serves a client base where a large minority of individuals physically sit outside the country when they search. Their queries land in French volume for France or German volume for Germany. Set the tool to Luxembourg and you have excluded them by construction.

The tools themselves sometimes return nothing at all

During research for this piece, a market-intelligence platform query for Luxembourg digital-marketing demand returned no data. That is worth stating plainly, because it is a tooling limitation rather than a finding. Small countries frequently sit below the reporting thresholds of commercial datasets. An empty export is not evidence of an empty market, and treating it that way is how agencies end up telling a Luxembourg client that their category has no search demand when the client's own sales pipeline says otherwise.

The four language datasets, and why merging them ruins the plan

Ask what language a Luxembourg website should be published in and you will be handed a statistic. The trouble is that at least four different statistics circulate, they measure four different things, and they point in different directions. Merging them produces a site architecture that looks evidence-led and is not.

What is measured	Figures	Source and behaviour
Languages used at work by employed residents	French 69.2%, Luxembourgish 54.4%, English 40.0%, German 29.5%, Portuguese 14.8%	STATEC, from the 2021 census. Multiple answers allowed, so these are not shares and do not total 100%. Excludes cross-border workers
Main working language of the company, reported by business leaders	French 56%, Luxembourgish 20%, English 18%, German 6%	STATEC via the national portal. Single choice, so this one does behave like a share
Self-reported ability to speak the language	French 98%, English 80%, German 78%, Luxembourgish 77%	Ministry of National Education study, 2018. Competence, not usage, and certainly not search behaviour
Main and spoken language of residents	Luxembourgish main language 48.9%, spoken 61.2%, German 22.5%	STATEC linguistic diversity data, updated January 2026. Resident identity, not commercial publishing

Look at what happens if you pick one and run with it. Take the 98% French ability figure and you conclude French is the only language that matters. Take the 48.9% Luxembourgish main-language figure and you conclude a Luxembourgish site is essential. Take the four official languages at face value and you build four equal trees, quadrupling review burden and multiplying technical failure modes without demand evidence behind each branch.

None of those four datasets measures search volume by language. Not one. That is the gap, and no public source closes it.

Geography and sector both override the national average

French use at work ranges from 43% to 83% depending on the municipality, and sits at 67% in Luxembourg City itself, according to the same STATEC linguistic diversity release. So a national average is a poor guide for a business with one location. English predominates specifically in finance, insurance, specialised professional services and the EU institutional cluster, which is exactly where the highest-value B2B queries live.

Written reality diverges from spoken reality

Here is a detail that surprises people. French is the sole language of legislative drafting in Luxembourg. Meanwhile the print press runs heavily in German, with one survey-based estimate putting print media at roughly 65% German, 25% French and 10% Luxembourgish. So the language of law, the language of media and the language of the office are three different answers in the same small country. Luxembourgish is overwhelmingly a spoken and identity language rather than a written commercial one, which is why it can be the main language of nearly half the residents and still carry thin commercial search demand.

Our working position, stated as a position rather than a proof: French and English first for internationally oriented B2B, finance and professional services. German where the client genuinely serves DACH counterparties or the eastern border. Luxembourgish reserved for civic, cultural, municipal and brand-navigational content where nothing else would read as authentic. Then test it against the client's own Search Console and CRM data and be willing to be wrong.

The market you are actually selling into

The resident population understates this market so badly that planning against it produces the wrong content entirely. Two structures sit on top of those 690,959 residents.

An investment fund industry measured in trillions

Investment vehicles domiciled in Luxembourg held EUR 6,731.325 billion in total net assets at 30 June 2026, spread across 2,968 undertakings for collective investment, up 16.31% over twelve months, according to the CSSF's monthly statistics. Luxembourg is the largest fund domicile in Europe and the second largest worldwide after the United States. Banking assets run at roughly 13 times GDP and insurance at roughly 3 times, on IMF figures from its financial sector assessment.

Now notice the direction of two of those numbers. Assets rose 16.31% over the year while the count of vehicles fell, from around 3,100 a year earlier to 2,968. That is consolidation. Fewer, larger, better-resourced buying organisations. Which means longer procurement cycles, named reviewers, documented process, and content that has to survive a committee rather than charm an individual.

The financial sector's weight in the wider economy is often quoted as a single number, and it should not be. Estimates range from 22.4% to about 30% depending on whether you measure gross value added or GDP, which year, and whether ancillary services are included. The OECD's 2025 economic survey puts it at roughly 25% of GDP and about 10% of employment. A conservative single-year anchor is 23.5% of GDP in 2023 on STATEC figures cited by the Chamber of Commerce. The range is a definitional artefact, not a contradiction, and saying so is more useful to a client than picking whichever number sounds most impressive.

An EU institutional cluster, and a very ordinary SME economy underneath

Luxembourg City is the second largest European Union duty station after Brussels, with something on the order of 12,000 EU staff, including roughly 2,200 at the Court of Justice and around 4,000 at the European Investment Bank. Those figures come from secondary compilations rather than a single official register, so treat them as indicative. The point stands regardless: a large, highly educated, multilingual, English-working professional population that no resident-language statistic describes well.

Underneath all of that sits a normal small-business economy. Around 40,400 small and medium enterprises operated in the non-financial business economy in 2023. SMEs make up about 99.5% of non-financial enterprises and roughly 67% of employment. Plumbers, dentists, restaurants, garages, IT resellers. Their search demand is local, often French or Luxembourgish, and frequently invisible in a planner export while being perfectly real in their phone log.

A keyword protocol that survives this market

Given all of the above, the research method has to change rather than the tool. What follows is the protocol we use, and every step of it exists because one of the problems above would otherwise corrupt the output.

What this protocol deliberately does not produce

It does not produce a single national search volume for Luxembourg. It cannot, because the demand genuinely originates in four countries and no method reunites it honestly. It does not produce a language split of Luxembourg queries, because no public dataset measures that. And it does not produce a euro figure for the size of the Luxembourg SEO services market, because no statistical office or independent research firm publishes one. Allocating a European estimate by population or GDP would manufacture a number that looks precise and means nothing.

Your keyword tool only knows about one of the engines

Here is an assumption buried so deep in most planning that nobody states it. Keyword volume estimates describe Google. That is fine in markets where Google holds ninety-something percent of search. Luxembourg is not one of those markets.

Statcounter's Luxembourg figures for July 2026, measured across all platforms, put Google at 83.11%. Bing follows at 7.58%, DuckDuckGo at 6.01%, then Yandex at 1.01%, Ecosia at 0.88% and Yahoo at 0.7%. So roughly one query in six runs somewhere other than Google, and your keyword dataset is silent about every one of them.

Two of those numbers are unusual enough to deserve attention rather than a footnote.

DuckDuckGo at 6.01% is several times the share it holds globally. Bing at 7.58% also runs well above its global average. Both patterns are consistent with what we know about the population: high income, high education, unusually privacy-attentive, and an institutional employment base in regulated finance and EU bodies that runs heavily on Microsoft 365 and Edge. That is a plausible explanation and not a proven one, since no study establishes the causal link for this market. Treat it as an observed pattern worth testing rather than a mechanism.

The practical consequences are concrete. Verification should happen in Bing Webmaster Tools alongside Search Console, not Google alone, because otherwise a sixth of the market goes unmonitored. Rank tracking configured for Google only will report a visibility picture that is systematically incomplete here in a way it would not be in, say, Italy. And when a client asks why their Bing impressions look disproportionately large relative to what the agency dashboard suggested, this is the answer.

One caveat on the source, because it matters. Statcounter measures activity across its own panel of sites rather than conducting a census of queries, so the level should be read as directional. The trend is arguably more interesting than the level anyway. Google's Luxembourg share was measured at 89.44% in March 2024. If both readings are sound for their respective dates, that is roughly six percentage points of share moving away from Google in a little over two years, which is a strategic fact rather than a rounding error. Re-read it at the source before quoting it in a plan, because it will keep moving.

One more trap, and it has nothing to do with keywords

While researching who actually sells search services into this market, two patterns showed up often enough to be worth a warning to buyers.

The first: a .lu domain and the phrase "SEO agency in Luxembourg" on a website do not establish that anyone is in Luxembourg. At least one provider marketing itself that way lists a contact address in Portugal. The second: a directory entry showing a European city does not establish it either. At least one provider listed under a Netherlands city turns out, on its own company materials, to be headquartered in India.

Neither of those is dishonest by itself, and offshore delivery is a legitimate model. We deliver from Indonesia and say so on every page. The point is narrower: if location matters to your procurement, or to your data-protection assessment, verify it against a trade register entry and a named legal entity rather than a domain extension. That is a five-minute check that changes the shortlist more often than people expect.

What to do on Monday

If you are responsible for a Luxembourg site and this article has landed uncomfortably, three things are worth doing before any content gets commissioned.

First, open Search Console and split the last twelve months by country. If a meaningful share of impressions comes from France, Belgium or Germany, your keyword model is currently describing a smaller market than the one you serve. Second, ask which of the four language statistics your current architecture was built on, and whether anyone checked what that statistic actually measures. Third, look at your last twenty inbound enquiries and note which language each arrived in. That sample is small and biased, and it is still better evidence about your buyers than a national average.

None of that requires a new tool. It requires treating this market as four overlapping markets in four languages, which is what it is.

---

Frequently Asked Questions

Why does keyword research for Luxembourg return such low volumes?

Three effects compound. The resident population is 690,959, small before you split it across four languages and by commercial intent, and keyword tools suppress or round low-volume cells to zero. Nearly half the workforce, 47% of roughly 494,000 salaried jobs at end-2025, commutes in from France, Belgium and Germany and searches from those networks, so their demand is filed under other countries. And small markets frequently sit below the reporting thresholds of commercial datasets, so an empty export can be a tooling limitation rather than a finding about demand.

Should I add French, Belgian and German search volume to my Luxembourg figure?

No. Keep each dataset under its own country label and never sum them. A merged total destroys the one piece of information you most need, which is where the searcher physically was. That determines which domain they used, which language they searched in, which competitors they saw, and whether your local landing page or your regional page should be the one ranking. Report four numbers with four labels, and let the client decide how to weight them.

Which language should a Luxembourg business publish in?

For internationally oriented B2B, finance and professional services, French and English first, German where there is genuine DACH or eastern-border business, and Luxembourgish for civic, cultural or brand-navigational content. That order comes from separating four datasets that measure four different things: workplace language use, main company working language, self-reported speaking ability, and resident main language. None of them measures search volume by language, because no public dataset does. So set the architecture at template level, then validate against first-party Search Console and CRM data.

Is there real commercial search demand in Luxembourgish?

Nobody has published a measurement, and we will not invent one. What the evidence supports is that Luxembourgish is spoken-first: main language of 48.9% of residents and spoken by 61.2%, while written commercial publishing skews heavily to French and German, with French the sole language of legislative drafting and print media running at roughly 65% German. The reasonable inference is meaningful demand in civic, municipal, cultural and brand-navigational queries rather than in the commercial long tail. That is an inference, clearly labelled as one.

Why does the financial sector's share of the economy keep changing between sources?

Because the sources measure different things. Figures circulate between 22.4% and about 30% depending on whether the metric is gross value added or GDP, which reference year is used, and whether ancillary professional services are counted. The OECD's 2025 survey puts it at roughly 25% of GDP and about 10% of employment. A conservative single-year anchor is 23.5% of GDP in 2023. The spread is definitional rather than contradictory, and the honest presentation is a range with the definition attached.

How do I check whether an agency is genuinely based where it claims?

Look for a named legal entity and a trade register entry, not a domain extension or a directory city. During research for this article, one provider using a .lu domain and Luxembourg positioning listed a contact address in Portugal, and one listed under a Netherlands city appeared from its own materials to be headquartered in India. Offshore delivery is a legitimate model and we use it ourselves. The problem is only when location is asserted rather than verifiable, which matters for procurement and for any data-protection assessment.

What is the size of the SEO market in Luxembourg?

No statistical office, trade body or independent research firm publishes a figure, and four separate research passes for this article failed to find one. We state that as unavailable rather than allocating a European or global estimate by population or GDP, which would produce a precise-looking number with no measurement behind it. What can be described honestly is market structure: a small resident base, a workforce nearly half of which commutes in, a fund industry holding EUR 6,731 billion at mid-2026 across 2,968 consolidating vehicles, and roughly 40,400 SMEs underneath.

---