I analysed 40 Finnish-language searches on 10 September 2026, and Google’s AI-generated response appeared in 24 of them. However, the distribution is stark: the response appeared in 70 per cent of information searches, but only in 38 per cent of cases where the searcher was looking to make a purchase. Another figure was even more surprising — the top result in the search appeared as the source of the AI response in only 58 per cent of cases. So, the top position does not guarantee a place in the response. This article presents the full data set and the conclusions drawn from it.
No measurements have been published on this in Finnish. The figures often cited in the industry are from the US — a different language, different search behaviour, different competition. That’s why I did my own.
What was measured
- 40 searches that Finnish small business owners actually carry out: websites, online shops, visibility, setting up a business, bookkeeping, law, data security.
- The search term list was finalised before the measurement. No search terms were removed afterwards. The list is included in its entirety in the measurement script so that it can be checked.
- Finland, Finnish language, mobile search, top 20 organic results.
- For each one, we noted: whether an AI-generated response appeared, which websites it linked to, and whether those websites also appeared in the top ten organic search results.
The final point is the crux of the whole assessment. If the AI response points to the same websites that appear at the top of the search results, standard search engine optimisation is still sufficient. If it points elsewhere, the competition is a different ball game.
Result 1: In most cases, there is no AI-generated response in shopping searches
| Search type | Searches | The AI response was displayed |
|---|---|---|
| Search — “what is this?”, “how does this work?” | 27 | 19 (70 %) |
| Shopping — “Who makes this?”, “How much does it cost?” | 13 | 5 (38 %) |
| Total | 40 | 24 (60 %) |
No AI-generated results appeared at all for these searches: homepages for businesses, websites for businesses, website prices, homepages for small businesses, advertising agency, accountancy firm, search engine optimisation, search engine optimisation price, GDPR for small businesses, cookie legislation.
This is, in practical terms, the most significant finding, and it is reassuring. When a person is ready to make a purchase and searches for a creator, Google still displays a standard list of results. The search that brings in the customer is still the one that can be influenced by search engine optimisation.
An AI response comes into play at the stage where a person is just starting to look into something. That stage is often missed — but it is also the stage where the customer has not yet made a purchase.
Result 2: a top ranking does not guarantee a place in the answer
Of the 24 searches where an AI-generated answer appeared, the top organic result was the source 14 times — that is, in 58 per cent of cases. In every second result, Google’s own top search result was therefore not deemed a suitable source for the answer.
The responses averaged 5.5 sources, totalling 131. Of these, 100 (76 per cent) were also in the organic top ten, whilst 31 (24 per cent) came from outside it.
Three out of four sources therefore come from the top. Search engine optimisation still goes a long way, and it’s not worth abandoning it just because of some new acronym. But a quarter comes from elsewhere, and that means two things at once: a top ranking isn’t enough, and a top ranking isn’t a prerequisite.
There’s a huge difference between one search and another
The average hides the fact that the searches behave in completely different ways. In some cases, the AI response reproduces the list of results almost verbatim:
| Search | Sources | These are outside the top ten |
|---|---|---|
| WordPress security | 8 | 0 |
| website update | 7 | 0 |
| web hosting comparison | 6 | 0 |
| Setting up a sole trader business | 4 | 0 |
And in some respects, it draws almost all of its content from elsewhere:
| Search | Sources | These are outside the top ten |
|---|---|---|
| online shop for businesses | 6 | 5 |
| AI optimisation | 7 | 5 |
| Website redesign | 4 | 3 |
| How much does a website cost? | 6 | 2 |
The latter group has something in common: they are searches for which there is no single correct answer. ‘Online shop for a business’ can mean a platform, a service or an entire project, and at the top of the results list are large e-commerce sites that do not answer the question. When the search results don’t provide an answer, artificial intelligence looks further afield — and that’s when a page that isn’t at the top but provides a direct answer makes it onto the list.
What does this mean in practice?
- Do not purchase a separate ‘GEO service’ until your search engine optimisation is in order. Three out of four sources come from the top of the rankings. Search engine optimisation is still the same process, and it’s a prerequisite for appearing in AI-powered search results for most queries.
- Look at your own search terms; don’t generalise. The difference between 0 and 5 external sources is so great that you cannot deduce your own situation from the average. Go through your own list before you decide anything.
- Write answers, not introductions. The pages that made it into the top results answered the question directly. The AI quotes an independent statement, not a paragraph that refers to the previous one.
- Search queries are quiet for the time being. If your turnover comes from searches such as ‘company websites’ or ‘accountancy firm’, the AI response won’t take anything away from you just yet. This may change, which is why it’s worth reassessing the situation.
Where this measurement is not valid
It is worth pointing out these limitations, as they are not usually mentioned:
- One day, one measurement. 10 September 2026. The AI response varies from day to day and even from one search to the next; the same search may yield a different result ten minutes later. These figures represent a snapshot of the current situation, not a trend.
- The 40 search results are not a sample of all Finnish search queries. They have been selected for a specific audience — small business owners. In another sector, the distribution may be completely different.
- Mobile search, not logged in, Finland. Results may vary on a desktop or when logged in.
- This does not track clicks. It shows where the AI response appears and who it refers to — not how many people chose not to click. For that, you need data from your own website.
All content
All 24 searches in which an AI-generated response appeared, and how many of these sources also featured in the top ten organic results:
| Search | Sources | Kympissä |
|---|---|---|
| Doing the company’s own bookkeeping | 8 | 6 |
| WordPress security | 8 | 8 |
| website update | 7 | 7 |
| AI optimisation | 7 | 2 |
| accountant’s fee | 7 | 5 |
| web hosting | 7 | 5 |
| How much does a website cost? | 6 | 4 |
| online shop for businesses | 6 | 1 |
| domain name transfer | 6 | 5 |
| web hosting comparison | 6 | 6 |
| IT support for businesses | 6 | 5 |
| WordPress websites | 5 | 5 |
| What is geo-optimisation? | 5 | 4 |
| Google Business Profile | 5 | 3 |
| e-invoicing is mandatory | 5 | 4 |
| accessibility requirements | 5 | 3 |
| Privacy policy template | 5 | 4 |
| Website redesign | 4 | 1 |
| Setting up an online shop | 4 | 4 |
| What is the best online shop platform? | 4 | 3 |
| Cost of setting up an online shop | 4 | 4 |
| Setting up a sole trader business | 4 | 4 |
| light entrepreneurship or a trade name | 4 | 4 |
| Setting up a limited company | 3 | 3 |
The source set is more fragmented than I expected. Among those 131 source locations, 104 different websites appeared, 82 of which appeared only once. Twenty websites appeared twice. Only two appeared more than twice: zoner.fi (5 times) and hostingpalvelu.fi (4).
This is the most encouraging figure in the whole survey. No one has a monopoly on AI responses. There isn’t a handful of large sites that would take all the source positions — three out of four sources are sites that made it into the results once, with their own question. It is therefore possible for a small player to become a source with a single well-answered question, which is not the case at all in the same way at the top of the organic search results.
In demand
What is Google AI Overview?
It is an AI-generated summary that appears at the top of Google’s search results, which answers the query directly and refers to a few sources. In Finnish, it is labelled ‘AI summary’. When it appears on the page, the usual search results are moved below it.
How often does the AI Overview appear in Finnish?
In my own analysis on 10 September 2026, it appeared in 24 out of 40 searches, or 60 per cent. The distribution was stark: 70 per cent in information-seeking searches, and 38 per cent in purchase-intent searches. The searches were selected from the perspective of a small Finnish business owner, so the figure cannot be generalised to all searches.
Is the top result in a search always the source of an AI-generated answer?
No. In my analysis, the top organic result was the source for 58 per cent of the searches in which the AI-generated answer appeared. A top position helps but isn’t enough: of all sources, 76 per cent were in the top ten organic results and 24 per cent were outside them.
Is search engine optimisation still worth it if artificial intelligence handles the search?
It’s worth it, for two specific reasons. Three out of four sources of AI-generated responses were also at the top of organic search results, meaning that search engine visibility is a prerequisite for AI visibility. And in searches with purchase intent, AI-generated results appeared in only 38 per cent of cases — it is precisely those searches that bring in customers that are still decided by the standard results page.
How do I become a source for AI-generated answers?
Based on the analysis, two things stood out. The page must answer the question directly and in independent sentences — the AI generates a sentence that can be understood without the preceding paragraph. And it must answer precisely the question that was searched for: the pages that made it into the results from outside the top rankings were in searches where the top result in the list did not answer the question at all.
Can this measurement be repeated?
The searches were carried out via Google’s search results API: Finland, Finnish language, mobile, the first 20 organic results, and for each one, the AI response’s source list and the top ten organic results were analysed. The list of search terms was finalised before the measurement and nothing was removed from it afterwards. Results vary from day to day, so the exact figures will not be repeated as such — though the order of magnitude should remain the same.
Would you like to know what your search terms look like?
The average tells you nothing about your specific situation — the difference was between 0 and 5 external sources within the same dataset. I’ll measure my own search terms in the same way and tell you what has happened to them.
Sources and methodology
- Data: own measurement on 10 September 2026. 40 searches, Finland, Finnish language, mobile, 20 organic results per search. The search term list was finalised prior to the measurement.
- What was recorded: the existence of AI-generated results, their source domains, and how many of them appeared in the top ten organic search results.
- Search demand: ‘google ai overview’ and ‘google ai summary’ appear in Google’s own search suggestions from the eleventh character onwards (measured on 10 September 2026).
- Related topics: What is GEO optimisation and AI optimisation: what’s being sold and what’s actually true.