Blog
-
6
min read

Lithuanian AI Visibility Study 2026: What ChatGPT, Perplexity and Gemini Actually Cite in 504 Answers

We asked three AI assistants 504 questions about brands in six Lithuanian sectors. Banks and telecoms are named in 84–89 percent of answers; real estate, health and B2B mostly get no brand at all. Company websites are only 9 percent of Perplexity's citations.
By
Aurimas Paulius Girčys
-
Jul 20, 2026

Published in Lithuanian in two parts on 16 August 2026; English summary 2 September 2026. Author: Aurimas Paulius Girčys, CEO, APG Agency. Full data, prompts and an Excel workbook with all 504 answers are available on request.

Short answer: in Lithuania, AI assistants already have favourites in banking, telecoms and e-commerce, and almost none in real estate, health and B2B services. The sources they cite are mostly not the brands' own websites: registries, directories, Reddit, media and comparison sites decide who gets named. For any Baltic brand, that means the way into AI answers runs through third-party sources first and your own site second.

Method

  • Six sectors: e-commerce, finance, telecoms, real estate, health and B2B services; 60 brands tracked.
  • 12 buyer prompts per sector in Lithuanian, plus English variants for comparison: "best X in Lithuania", "which X to choose", "how much does X cost", "is X reliable".
  • Three assistants via API: Perplexity Sonar (three runs per prompt), Gemini 3.5 Flash with search (three runs), GPT-5.5 with search (one run). 504 answers in total, 1,273 citations classified by source type in part one alone.
  • What we counted: whether a brand is named, in what position, which URLs are cited, and how much the answer changes between repeated runs.

Finding 1: some sectors have AI incumbents already

SectorMost-named brandShare of answers naming itAnswers with no brand at all
TelecomsTelia89%low
FinanceSwedbank and SEB84%low
E-commercePigu69%low
Real estatescatteredunder 30%19–30%
Healthscatteredunder 30%19–30%
B2B servicesscatteredunder 30%19–30%

In telecoms, finance and e-commerce the assistants converge on one or two national leaders, which makes those categories hard to enter and easy to defend. In real estate, health and B2B a fifth to a third of answers name nobody, and the rest spread across many small brands. Those three sectors are open: a brand that gets into the cited sources this year can become the default answer.

Finding 2: the assistants cite different worlds

AssistantTop cited sources (part 1, e-commerce and finance)Share of citations that are brand websites
Perplexityreddit.com (86 of 1,273 citations), comparison and review sites, media19%
ChatGPT (GPT-5.5)brand websites, then media and directories40%
GeminiGoogle-ranked pages, directories, mediabetween the two

In part two, covering telecoms, real estate, health and B2B, Perplexity's top sources were rekvizitai.vz.lt (the company registry, 104 citations), Reddit (93) and medicina.lt (81). Company websites were 9 percent of Perplexity citations and 20 percent of ChatGPT citations. The practical rule: to be named by Perplexity you must be present in registries, directories and community threads; to be named by ChatGPT your own pages must answer the question directly.

Finding 3: language changes the answer

The same question in English returned on average 5.5 brands; in Lithuanian, 3.8. Small-language markets are less crowded inside AI answers, which is good news for a Lithuanian, Latvian or Estonian brand: fewer competitors are named, so each mention carries more weight, and getting in is cheaper than in English.

Finding 4: one run is not a measurement

Repeating the same prompt produced answers that overlapped by only 33–43 percent. A single screenshot of ChatGPT naming or not naming a brand proves nothing. AI visibility has to be measured as a share across repeated runs and reported per engine, which is how we run the AI visibility audit and the monthly monitoring inside GEO services.

What this means for a Baltic brand

  1. Check the registries first. If rekvizitai, the Latvian or Estonian company registers, or a directory lists an old address or the wrong services, that error is what Perplexity repeats.
  2. Get into the lists. "Best X" questions are answered from comparisons, directories and media. Being absent from the six lists an assistant cites for your category means being absent from the answer.
  3. Answer the question on your own site. ChatGPT cites brand pages when they contain a direct answer, a price and a date. A homepage tagline is not cited.
  4. Measure with a fixed prompt set, monthly. 15–30 questions, three or four engines, repeated runs, per-engine share.
  5. Consider the paid layer where organic is slow. Since August 2026 a sponsored card can sit under the ChatGPT answer in all three Baltic markets; see ChatGPT Ads management.

Limitations

One market, one month, three assistants, 60 brands. GPT-5.5 was run once per prompt because of cost, so its per-brand shares carry more noise than Perplexity's and Gemini's. Sector definitions and brand lists are ours. We will repeat the study quarterly and publish the deltas.

Related: Free AI visibility check · GEO pricing 2026 · Lithuanian originals: part 1, part 2.

Ready to Accelerate Your Business?
  • Holistic Digital Solutions
  • ROI-Focused Campaigns
  • Data-Driven Decision Making
  • Strategic Content Creation
  • Cross-Channel Synergy
Contact us
HERE IS WHAT WE DO

APG Media HUB

APG Media offers holistic, growth-focused digital marketing solutions for businesses with a purpose to grow.

As part of our commitment to staying at the forefront of the industry, our range of services is continually expanding. Our core services include:

Your Digital Marketing Partner