Automatic SEO: what can a machine take over — and what can’t it?

"Automatisk SEO" is searched for in Danish, Swedish and Norwegian — but half of the work cannot be automated by anyone. We measured where the line is drawn: in 576 AI responses, the technical setup was a strong predictor of whether a website was cited as a source (p=0.000), and had no bearing on whether the company was recommended by name (p=0.988). Here is the overview, the cost of the manual alternatives — and how to tell if the automated system is working.

"Automatisk SEO" is searched for in Danish, Swedish and Norwegian — and behind the word lie two very different types of work. We have measured where the line is drawn: in 576 AI responses concerning 465 real companies, technical setup was a strong predictor of whether a website was cited as a source (p=0.000), and nothing about whether the company was recommended by name (p=0.988). In short: the half that involves making your site readable by machines can actually be automated — and there’s plenty to get to grips with, as 88.2 per cent of 1,252 Nordic SME sites still lack the very basics. The other half, which is about getting selected, cannot be automated. Here’s what belongs where, what manual work costs by way of comparison — and how you can tell if the automation actually works.

What does ‘automatic SEO’ actually mean?

The term covers three different things sold under the same name. Automatic analysis scans your website and finds the errors — tools have been able to do this for many years. Automatic execution fixes them without you having to edit the code yourself. Automatic content generates text on your behalf. The three aren’t equally mature, and they don’t achieve the same results.

The distinction isn’t an academic one. The technical work relies on well-documented signals, whilst the aspect relating to being recommended appears to lie in a completely different realm: Ahrefs compared 75,000 brands with their visibility in Google AI Overviews and found that online mentions of the brand were more closely linked to visibility (correlation 0.664) than backlinks were (0.218). This is a correlation, not a proven cause – but it’s not something you can simply automate your way out of.

The term is worth taking seriously because it is the language that Nordic SMEs actually uses. In a collection of 1,482 unique Google suggestions in Danish, Swedish, Norwegian and English (own collection, August 2026), one suggestion appeared in all three Nordic languages:

automatisk seo

Technical terms such as “GEO” and “AI SEO”, on the other hand, are found almost exclusively in English. People aren’t searching for the category. They’re searching for a way to avoid the work.

What can actually be automated — and what does the measurement show?

The technical layer can be automated, and this has been proven to be valuable. The Princeton and Georgia Tech research behind the GEO study (KDD ’24) measured a boost of up to 40 per cent in visibility in generative search engines by adding cited sources, statistics and a clear structure to the content. Our own figures point in the same direction. In our own measurement of 576 AI responses with web search enabled — four search engines, twelve Danish cities, four unbranded customer queries, compared against 465 real companies across three sectors (our own measurement, July 2026) — the companies whose websites were cited as sources had significantly higher technical scores than the rest: 77.9 versus 72.7, and structured data on 68 per cent versus 57 per cent of the sites. The difference is statistically significant (p=0.000). The methodology and caveats are clearly stated in the analysis itself.

What is interesting is which factors distinguished the cited sites from the rest: crawler access, content readable without JavaScript, structured data, and a clear structure. It is all repetitive, rule-based work — exactly the sort of thing a machine is good at, and exactly the sort of thing a human will never manage to cover entirely. A figure from the same dataset drives the point home: of the 4,034 sources cited by the AI responses, 61 per cent were the companies’ own websites. Your own site is your biggest source of citations — provided it can be read.

And there’s plenty to get to grips with. In the GEO Barometer for Q3 2026, we scanned 1,252 SME websites in Denmark, Norway and Sweden across 15 technical dimensions. 88.2 per cent (±1.8pp) had no llms.txt file – but only 3.6 per cent (±1.0pp) were actually unreadable by AI crawlers. So the issue isn’t that the task is difficult. It’s that almost nobody has done it. That’s what makes automation a sensible idea here: not because the machine is smarter than you, but because it won’t forget to do it again next month.

What can’t automation do?

It cannot get AI to recommend your business. In those same 576 responses, the businesses that were recommended by name were technically identical to those that were not: 74.9 versus 74.9 in score (p=0.988). The biggest explanatory factor was the sector — solicitors were recommended by name in 57 per cent of the responses, plumbers in 38 per cent and accountants in 32 per cent. These differences are far greater than any technical difference between the businesses. The difference between the two outcomes is the whole point of being cited or recommended, and it’s worth bearing in mind before you buy anything at all.

Two popular automation ‘recipes’ also fell short when we tested them against the dataset:

• “Let the machine write more text.” Cited pages averaged 1,083 words compared with 895 for the rest — but the difference is not statistically significant (p=0.32). More text in itself is not what makes the difference.
• “Generate a llms.txt, and you’ll be mentioned.” Our own dataset provides no support for this claim — the figures even point in the opposite direction in two out of three sectors. The file is still worth creating, but for the right reason: it makes it easier for machines to understand what the site contains. The guide takes half an hour, and we do not use our own figures as an argument in its favour.

What can be automated, and what can’t? The overview

How much does it cost to do the work manually?

Three approaches, three price points. Do it yourself: free in Danish kroner. Our manual weekly routine takes around 45 minutes a week, and it works — provided you stick to it. Hiring: a SEO specialist in Denmark costs at least 35,000 kr a month; this figure has been rounded down from two modelled salary sources (ERI SalaryExpert 2026: 515,705 kr/year, PayScale 2026: approx. 500,480 kr/year). Agency: entry-level prices from around 1,000 kr/month — setion.dk’s own published starting price, verified in August 2026 — and upwards from there.

By way of comparison, our own plans start at DKK 299/month for one website and 25 monitored questions, DKK 799/month for three websites and 100 questions, and DKK 1,499/month for ten websites and 300 questions (all excluding VAT). Automatic implementation of corrections is included in all paid plans — not an extra charge. There’s a 14-day free trial with no credit card required, and a free account gives you GEO Score without a subscription. Right now, set-up is included free of charge (normally 1,995 kr), and you’ll keep the price you start on.

The point of these figures isn’t that the cheapest option wins. It’s that the choice between the three options depends on how much of the work is actually repetitive. If it’s mostly the technical side, you’re paying too much for a human. If it’s about reputation and relationships in your industry, no machine can help you — not even ours.

How do you know if the automation works?

By measuring the two outcomes separately, each with its own denominator. This is the most important implication of the entire finding: a tool that displays a single ‘AI visibility score’ conflates something that technology demonstrably influences with something that technology demonstrably does not. Consequently, you cannot tell whether the work was effective.

You can measure the foundation for indexing in 60 seconds with the free test — can the search engines read, understand and index your site? The recommendation page requires someone to ask your actual customer questions over and over again under the same conditions and note down who is mentioned. This can be done manually as a spot check; it’s also precisely the part where it makes sense to let a machine handle it, because the value lies in the repetition, not in the effort.

Where should you start?

Start by getting the facts on the table before you buy any sort of automation. Run the free GEO-test on your own domain and see how many of the 15 technical dimensions are in place. If there are gaps in crawler access, structured data or readability, that’s the half that can be automated — and the part that, when measured, correlates with being cited. If everything is technically in place and you’re still not being mentioned, you’re in the ‘recommendation’ track, where the answer isn’t a tool, but work on reputation, publicity and demand within your industry. Check out the plans if you want the metrics and corrections to run automatically every week.

Frequently asked questions

What is automatic SEO?

A collective term for three things: automatic analysis of your website, automatic implementation of technical corrections, and automatic content generation. Only the first two are based on documented signals. In our analysis of 576 AI responses, technical setup was strongly correlated with being cited as a source (p=0.000) — whilst more text in itself made no significant difference (1,083 versus 895 words, p=0.32).

Can SEO be fully automated at all?

No. The technical layer can be automated — crawler access, structured data, llms.txt, content structure — because it is rule-based and repetitive. Two things, however, cannot be automated: the knowledge about your business that the machine would otherwise have to invent, and everything to do with reputation and mentions outside your website. In our data, the businesses that the AI recommended by name were technically identical to those it did not recommend (74.9 versus 74.9, p=0.988).

Can automatic SEO get my business mentioned in ChatGPT?

It can make your website readable, understandable and quotable — and this has been documented as what set the quoted businesses apart from the rest. But no one can promise you a placement or a mention in a third-party AI response, and it’s a red flag if anyone does. You can measure, correct and track progress; you cannot order the result.

How much does automatic SEO cost compared to an employee or an agency?

An SEO specialist in Denmark costs at least 35,000 DKK per month (rounded down from ERI SalaryExpert and PayScale, 2026). Agencies start at around 1,000 DKK per month and go up from there. Our own plans are priced at DKK 299, 799 and 1,499 per month excluding VAT, with automated execution included in all paid plans, plus a 14-day free trial with no credit card required.

Is automatically generated content a good idea?

Only if it’s based on your actual business knowledge and is proofread by a human before it’s published. Our analysis provides no evidence that more text in itself makes a difference (p=0.32). The quality must come from what only you know — what you do, for whom, and what sets you apart.

How many Nordic companies actually have the technical basics in place?

Fewer than you might think. We scanned 1,252 SME websites in Denmark, Norway and Sweden in July 2026: 88.2 per cent (±1.8pp) had no llms.txt file. Conversely, only 3.6 per cent (±1.0pp) were actually unreadable by AI crawlers. So the task isn’t difficult — it’s just not being done.