AI-engine visibility of Romania's book market: 18 purchase questions across 5 AI engines, July 2026

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

Raw data from an original study measuring how brands in Romania's book market appear inside the answers of five AI engines. Method. Eighteen real purchase questions were put to ChatGPT, Google Gemini, Perplexity, Google AI Mode and Google AI Overviews, in Romanian, from Romania, across two complete weekly runs (20 and 27 July 2026). No question contained a brand name — every brand that appears was named by the engine on its own. The result is 180 analysed answers (36 per engine) and 247 brand mentions across 21 tracked brands, with the position of each mention, the domains cited as sources, and sentiment recorded throughout. Monitoring was automated via the LLM Pulse platform, with no session history and no personalisation. Key findings. 58% of answers name none of the 21 tracked brands. Only 75 of 180 answers mention at least one; of those, just 19 recommend a single brand. Libris is the market's default answer: 37 mentions (20.6% of answers), and when it appears it is the first brand recommended in 65% of cases. 44% of all mentions are publishers rather than bookstores, even though every question is about buying books. Reddit is the second most-cited source in the whole market (34 answers), after carturesti.ro (46) — above Facebook, YouTube and any media publication. Some brands are cited but never named: librarul.ro is cited as a source in 14 answers and named once; targulcartii.ro has 18 citations and 4 mentions. The reverse also exists: Humanitas has 13 mentions and zero citations of humanitas.ro. Rankings move weekly. Between the two runs, elefant.ro went from 7.0% to 10.9% share of voice, Editura Corint quadrupled its mentions, and Editura Trei halved its own. Zero negative sentiment across all 247 mentions. Contents. Three CSV files (one row per brand; one row per question-brand pair; top 40 of the 422 cited domains), four charts as SVG, a data dictionary, and a starter notebook that verifies the arithmetic and rebuilds two of the charts. Two denominators. Mention rates are computed against 180 (all answers) and per-engine rates against 36. Citation percentages are computed against 172 — only the answers that actually contain citations. Mixing the two produces wrong numbers. Reproducibility. Per-engine rates are n/36 rounded to one decimal, so the underlying counts reconstruct exactly: ChatGPT 33, Gemini 78, Perplexity 30, AI Mode 56, AI Overviews 50 — summing to 247, which equals both the brand-level total and the sum of the two weekly runs (128 + 119). Limits. Two runs over two weeks; 18 questions rather than the full purchase universe; 21 tracked brands rather than the whole market; roughly 10 answers per question, so single-question differences are noise; mentions are not conversions; models update continuously. Responsibility note. All brands named in this dataset were named by the AI engines, not by the authors. The material is market research and does not evaluate the quality, services or commercial performance of any brand. Trademarks belong to their rightful holders and are used for identification within a factual analysis. Corrections: office@websem.ro Canonical version of the analysis: https://websem.ro/resurse/aeo/studiu-piata-carte-romania This is episode 2 of a series measuring how Romanian markets appear in AI answers. Episode 1 covered luxury jewelry (DOI 10.5281/zenodo.21724400).

Links and sources

Comments