GEO Barometer Q3 2026: 88% of Nordic SMB websites lack llms.txt

1,252 SMB websites in Denmark, Norway and Sweden, measured in terms of AI readiness: 88 per cent are missing llms.txt, but only 4 per cent are effectively unreadable by AI. Figures, methodology and margins of error.

We have analysed the websites of 1,252 small and medium-sized enterprises in Denmark, Norway and Sweden — plumbers, accountants and solicitors — across 15 technical dimensions of AI readiness. Three findings stand out: 88.2% (±1.8pp) have no llms.txt file, with Swedish law firms topping the list at 93% (±4.3pp) — but only 3.6% (±1.0pp) are actually unreadable by AI crawlers. The big picture, therefore, is not that “AI cannot read Nordic SMBs” — it is that almost no one has taken any active steps yet. The panel is fixed and is re-surveyed every quarter; the next survey is on 3 October 2026.

The barometer is the Nordic extension of our Danish July survey of 465 corporate websites — the same methodology, now with Norway and Sweden included in the panel.

Key figures, country by country

Measured in July 2026. n = number of sites scanned in the panel; ± is the 95% margin of error. Population = official number of companies in the sector for which register data is available at sector level — sourced from Statistics Denmark’s Statistics Bank and Statistics Sweden’s (SCB) Statistics Database.

Denmark (465 sites)

Industry	n	Population (DST, 2024)	Median GEO readiness	Without llms.txt	Effectively unreadable
Plumbing	200	2,793	85/100	90% (±4.2pp)	3%
Accountants & bookkeepers	147	4,677	75/100	85% (±5.8pp)	5%
Solicitors	118	1,891	70/100	91% (±5.2pp)	7%

Norway (383 sites)

Sector	n	Median GEO readiness	Without llms.txt	Effectively unreadable
Plumbers	123	80/100	89% (±5.5pp)	2 %
Accountants & auditors	126	85/100	86% (±6.1pp)	5
Solicitors	134	80/100	90% (±5.1pp)	4

No population figures are given for Norway: Statistics Norway’s open sector statistics are not available at the required level of detail, and we do not make estimates.

Sweden (404 sites)

Sector	n	Population (SCB, 2025)	Median GEO readiness	Without llms.txt	Effectively unreadable
Plumbing firms	132	9,328	85/100	81% (±6.7pp)	4 %
Accounting & auditing	136	25,848	85/100	88% (±5.5pp)	2%
Lawyers	136	7,397	85/100	93% (±4.3pp)	1 %

What the figures actually mean — and what they do not mean

First, the good news: the median score is 70–85 out of 100 across all nine series. Most Nordic SMB websites are technically sound enough for AI search engines to read them — only around 4 per cent actually block AI crawlers or serve content that machines cannot read. The narrative that ‘AI can’t see small businesses’ doesn’t hold up in the data.

The real picture is more interesting: almost no one has taken any active steps to be selected by AI. llms.txt — a file that tells AI crawlers what your site is about — is missing from 88 per cent of sites across the Nordic region. It’s not because the file is difficult to create; it’s because the category is so new that almost no one has got started yet.

And now for the honest caveat you rarely get from people selling GEO: llms.txt is a controversial signal. Ahrefs’ analysis of their crawl data showed that the vast majority of llms.txt files have never been crawled by AI crawlers. We measure it as one of 15 signals — not as a miracle cure. The reason the figure is interesting nonetheless is what it reveals: that the race for AI visibility in the Nordic region hasn’t really begun yet. The lead is still up for grabs.

The difference between being found and being chosen: a technically sound website is the ticket to entry — not the victory. When we ask AI engines real customer questions (“best accountant in Aalborg”), they typically mention a handful of companies and cite a few sources. To become one of them requires citable content, external signals and a clear corporate identity — the other dimensions in the survey. See also the industry rankings from the Danish survey.

Method

• Panel: 1,252 company websites (DK 465 · NO 383 · SE 404) across three sectors, identified via a systematic search using sector-specific terms × largest cities per country and filtered to exclude portals, media outlets and public authorities. The panel is fixed: the same sites are re-assessed every quarter, allowing trends to be tracked longitudinally.
• Measurement: Each site is technically scanned (raw HTML readability, robots.txt access for AI crawlers, llms.txt, structured data/JSON-LD, page signals) and is scored on a scale of 0–100. No AI-generated assessments are included in these figures — only deterministic checks.
• “Effectively unreadable” means that the site blocks AI crawlers in robots.txt or serves content that cannot be read without JavaScript rendering.
• Margins of error are 95% confidence intervals on proportions; series with fewer than 25 sites are not published.
• Population figures: Statistics Denmark (workplaces, 2024) and the SCB Statistics Database (number of companies, 2025) — both broken down to industry code level. Norway is deliberately excluded (see above).
• Known limitations: (1) The panel was identified via a search and therefore likely over-represents the most search-visible companies — the actual industry figures are unlikely to be any better than ours. (2) A small proportion of the new Norwegian and Swedish panels may contain portals/directories; the panel is cleaned prior to the Q4 survey, and changes to the panel composition are documented. (3) The technical readiness metric measures whether AI can read your content — not whether you are cited; we measure that separately.
• Next survey: 3 October 2026 (automatic, same panel). Quarterly changes are only published when they exceed the margin of error.

Frequently asked questions

What is llms.txt?

A text file on your website (such as robots.txt) that tells AI systems what your site is about and which pages are most important. It is a new, optional standard proposal — easy to create, but not yet consistently used by AI crawlers.

Is my website invisible to AI if I don’t have a llms.txt file?

No. Only around 4 per cent of the sites analysed are actually unreadable by AI crawlers. llms.txt is just one active signal among many — the absence of it mainly indicates that almost no one in your industry has got started yet.

Why do Swedish solicitors top the list for the absence of llms.txt?

93 per cent (±4.3pp) of the 136 Swedish law firms surveyed are missing the file — the highest proportion in the entire survey. At the same time, their technical median score is among the highest (85/100). The pattern is consistent across the board: solid technical capabilities, but no active AI optimisation.

How can I see how my own firm is performing?

Run a free assessment at geoa.app/geo-score/check — same engine, same 15 dimensions as the barometer, 60 seconds, no login required.

Key findings (citable)

1. 88.2% (±1.8pp) of 1,252 Nordic SMB websites have no llms.txt (Geoa GEO Barometer, July 2026).
2. Only 3.6% (±1.0pp) are effectively unreadable by AI crawlers — the problem is not access, but a lack of active optimisation.
3. Swedish law firms have the largest llms.txt gap: 93 per cent (±4.3pp) of 136 assessed — out of 7,397 law firms in Sweden (SCB, 2025).
4. Median readiness is 70–85/100 across all nine series — Nordic SMB websites are technically sound but passive in the AI race.
5. The survey is repeated quarterly on the same fixed panel; next data point: October 2026.

Sources: Geoa GEO Barometer Q3 2026 (own survey, July 2026, n=1,252, methodology as above) · Statistics Denmark, workplaces by sector (2024) · SCB Statistics Database, number of companies by SNI (2025) · Ahrefs, llms.txt crawl analysis