Skip to content
WebSEM
— FOR THE PRESS

Press data: findings from Websem's AI-visibility studies

Every major finding from Websem's studies, written as a self-contained sentence with its measurement date and the dataset DOI. 5 studies, all with raw data downloadable under CC BY 4.0: you may quote and reuse them with attribution.

How to cite us

The attribution line, with the DOI of the study a finding comes from, sits under each result. Websem is an AEO and GEO agency in Bucharest, founded in 2013; the studies' author is Dan Cristian Alexandrescu.

  1. Episode 5 · 10 Sep 2026

    Can an AI agent find you?

    Zero Agent Cards out of 126 responding domains, a mean ADO Score of 17 out of 100, and the interoperability that does exist comes from Shopify and WordPress plugins, not from strategy.

    130 Romanian domains machine-probed on 12 signals — A2A Agent Card, MCP discovery, llms.txt, JSON-LD, Wikidata — on the ADO Score framework (Exista.io, 2026).

    Attribution: According to a study by the AEO agency Websem (DOI: 10.5281/zenodo.22664431)

    DOI: 10.5281/zenodo.22664431 · licence CC BY 4.0 · full study

  2. Episode 4 · 9 Aug 2026

    How many Romanian sites publish llms.txt — and does it matter?

    64% publish llms.txt, so it is already the dominant practice — yet the most-cited domain carries a plugin-generated file with none of the summary line the specification asks for.

    87 Romanian domains that AI engines actually cite, checked at /llms.txt and /robots.txt in a single run.

    Attribution: According to a study by the AEO agency Websem (DOI: 10.5281/zenodo.22111520)

    DOI: 10.5281/zenodo.22111520 · licence CC BY 4.0 · full study

  3. Episode 3 · 6 Aug 2026

    Who sells the laptop when the customer asks an AI?

    86 of 87 answers name a major retailer, so this market is decided by position rather than presence: eMAG takes first place in 66 of its 82 mentions.

    18 brand-free purchase questions across 5 AI engines in a single complete run — 87 answers analysed, 378 mentions, 13 tracked retailers.

    Attribution: According to a study by the AEO agency Websem (DOI: 10.5281/zenodo.21821185)

    DOI: 10.5281/zenodo.21821185 · licence CC BY 4.0 · full study

  4. Episode 2 · 31 Jul 2026

    Who sells the book when the customer asks an AI?

    58% of answers name no major brand at all, and some brands are cited as sources without ever being named — Librarul: 14 citations, a single mention.

    18 brand-free purchase questions across 5 AI engines over two weekly runs — 180 answers, 247 mentions, 21 tracked brands.

    Attribution: According to a study by the AEO agency Websem (DOI: 10.5281/zenodo.21736211)

    DOI: 10.5281/zenodo.21736211 · licence CC BY 4.0 · full study

  5. Episode 1 · 14 Jul 2026

    The AI visibility of Romania's luxury jewelry

    Average pairwise agreement between models is 18%, and no brand appears in all ten lists.

    One identical prompt put to 10 public LLMs — 29 distinct brands across 50 available positions.

    Attribution: According to a study by the AEO agency Websem (DOI: 10.5281/zenodo.21724399)

    DOI: 10.5281/zenodo.21724399 · licence CC BY 4.0 · full study

Press contact

For interviews, methodology questions or access to the raw data: office@websem.ro · 0760 604 849. The studies' shared methodology is published on the AI Visibility Index.