Which agencies AI names in eight markets: 33,480 answers from six models
When a buyer asks an AI assistant which agency to hire, the answer names a handful of companies. We asked six models the same buyer questions in eight markets and counted who they name. Each market has its own page with the top 20 per question set, the split between answers from memory and answers with web search, and the names we could not verify.
Method: SEO questions in Poland, Estonia, Latvia, Lithuania, Finland, Sweden, Denmark and Norway, and advertising, media and PR questions in Poland, Estonia, Finland and Sweden, asked in each market's language and in English (Poland's SEO questions in Polish). Six models answered through their APIs (GPT-5.6, Claude Sonnet 5, Gemini 3.7 Flash, Grok 4.5, Mistral Large and Perplexity sonar), 15 times per question and model: 33,480 answers, collected from 7 to 9 September 2026. Five of the models answer from memory; Perplexity searches the web.
Rankfor.AI (opens in a new tab) publishes this study, sells software and a managed service, and does not rank itself.
The SEO agencies AI names most, by market
The three companies named most often in answers to the local-language SEO questions. Each market page has the full top 20 for every question set measured there.
Poland PL
Polish questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Delante | 64.2% |
| 2 | Widoczni | 54.5% |
| 3 | Bluerank | 43.9% |
Estonia EE
Estonian questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Convertal | 35.3% |
| 2 | SEO.ee † | 25.3% |
| 3 | Nobel Digital | 21.0% |
Latvia LV
Latvian questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | SEO.lv | 32.9% |
| 2 | Infinitum | 29.1% |
| 3 | iMarketings.lv * | 28.3% |
Lithuania LT
Lithuanian questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Market Rats | 26.1% |
| 2 | SEO.lt † | 23.2% |
| 3 | Digital Star | 19.2% |
Finland FI
Finnish questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Suomen Digimarkkinointi | 39.2% |
| 2 | Hopkins | 32.9% |
| 3 | Tulos Helsinki | 24.0% |
Sweden SE
Swedish questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Pineberry | 62.4% |
| 2 | Brath | 50.4% |
| 3 | Topdog | 30.4% |
Denmark DK
Danish questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Bonzer | 42.1% |
| 2 | Obsidian Digital | 35.0% |
| 3 | Webamp | 34.1% |
Norway NO
Norwegian questions, 1,080 answers
| Rank | Agency | Share of answers |
|---|---|---|
| 1 | Synlighet | 52.7% |
| 2 | Bonzer | 35.0% |
| 3 | Netlife | 22.4% |
Share of answers that contain the company name, counted once per answer. These are counts of names in AI answers. They do not measure quality of work, client results or market share.
† The name appears in answers and the company behind it is not established.
* Counted from the answer text because the name registry had no entry for it.
Which model wrote the links that led nowhere
In answers about SEO, advertising, media and PR agencies, Mistral Large wrote 8,590 of the 8,615 agency links that pointed to hosts with no DNS record (99.7%). In the SEO answers, counted in the table below, it wrote 5,117 of the 5,124 dead links, and its share ran from 99.5% in Estonia to 100% in six of the eight markets.
| Market | Dead links in SEO answers | Written by Mistral | Mistral share of agency links in SEO answers |
|---|---|---|---|
| Poland | 78 | 78 (100.0%) | 100.0% |
| Estonia | 638 | 635 (99.5%) | 95.8% |
| Latvia | 1,498 | 1,494 (99.7%) | 96.5% |
| Lithuania | 814 | 814 (100.0%) | 98.2% |
| Finland | 909 | 909 (100.0%) | 95.7% |
| Sweden | 402 | 402 (100.0%) | 95.1% |
| Denmark | 315 | 315 (100.0%) | 96.6% |
| Norway | 470 | 470 (100.0%) | 94.7% |
Mistral also wrote 22,929 of the 23,805 inline agency links in the SEO answers (96.3%). Grok, Claude and Perplexity wrote none, so the dead-link rate describes Mistral and tells nothing about them. Perplexity lists its web sources apart from the answer text, and those source lists were not DNS-checked.
See what AI tells your buyers about you
We help buyers find useful answers about your business when they ask AI what to choose. For the audit, we run the questions your buyers ask, in your buyers' language and in English, across the AI assistants they use. You see where your company is named, which sources the answers rely on, and what to publish first. Dmitrij Żatuchin, founder of Rankfor.AI, answers every request himself within one working day.
What the audit shows
- Where your company appears in the answers, and next to whom
- The pages AI cites in your market
- What to publish first on your own site
Questions people ask
Which AI models were tested?
GPT-5.6, Claude Sonnet 5, Gemini 3.7 Flash, Grok 4.5, Mistral Large and Perplexity sonar, all through their APIs. The first five answered from memory with web search off. Perplexity searched the web before answering. Every question went to every model 15 times.
Which markets and agency types are covered?
SEO agencies in Poland, Estonia, Latvia, Lithuania, Finland, Sweden, Denmark and Norway. Advertising, media and PR agencies in Poland, Estonia, Finland and Sweden.
Why do some AI answers link to websites that do not exist?
Almost all of those links came from one model. In answers about SEO, advertising, media and PR agencies, Mistral Large wrote 8,590 of the 8,615 agency links that pointed to hosts with no DNS record (99.7%). In the SEO answers it also wrote 96.3% of all inline agency links.
Do these rankings measure agency quality?
No. They count how often AI answers name a company. They do not measure quality of work, client results, prices or market share.
Measure your own company the same way
The method behind these pages, and the three numbers worth reporting, are on the answer engine optimization page. The Dice Roller runs one buyer question several times in one model, free and with no account. With your own API keys it runs the same question in Gemini, GPT and Grok.
Answer engine optimization Answer Trail (opens in a new tab)