Research
Field research on how machines see businesses — measured, dated, and honest about its limits. Small samples, real probes, no inflated numbers.
What this is
AVE Studio publishes repeated measurements of what AI answer engines cite and recommend when people ask a category's real buying questions. Each release is a dated run against a fixed, versioned panel — not a one-off query, and not a leaderboard. What we publish is what we observed, with the sample it came from.
The instrument, honestly
- Engines measured
- Claude, ChatGPT and Gemini, plus Google AI Overviews — the AI surface of Google Search, reported as a Google surface rather than as a fourth independent model.
- Versioned panels
- The prompt panel for a release is fixed and version-stamped before the run. A panel change starts a new version rather than editing a published one, so two releases are comparable or explicitly are not.
- Dated runs
- Every figure carries the date it was measured. AI answers move; a number without a date is not a measurement.
- Observation language
- We report what was observed, with its sampling range. We do not report a single combined score, and we do not claim that any change we made caused any change we measured.
Releases
- Model Mindshare Report— CEE · 2026 Q3 —
The terms these releases use
Findable but forgettable: are Hungarian companies ready for AI search?
Most Hungarian firms are findable but not “known”: retrieval is the healthy part, recall and authority are the gaps — and two-thirds of the sites block a basic crawler.