Gemini AI Mode test across 5 model versions: recommended entities stay the same
gaganghotra_ · x · 2026-11-13
Testing the same "best yumcha in Melbourne CBD" query across Google AI Mode versions (3.6 Flash, 3.7 Flash, 4 Argon, 4 Hydrogen, 4 Carbon), the author found the answer wording changes with each model but the recommended entities — Secret Kitchen Chinatown and Shark Fin Inn — remain the top two every time.
Takeaway: in AI Mode, the index appears to have more influence over which entities get recommended than the model itself; the model's "taste" only shapes the wording, not entity selection. A practically useful observation for SEO/GEO practitioners.
More from Apps
- AI job-application agent charges per application: $19-$99/mo for 600-4,500 submits — wavefnx · 2026-11-13
- How to build a math-tutoring agent: 5-step teaching chain beats generic LLMs — yangyi · 2026-11-13
- Dev claims the best voice mode and computer use he's used, built on OpenAI Realtime 2 and Claude — RileyRalmuto · 2026-11-13
- Voice Cursor launches Viki beta, a voice assistant that sees your screen and acts on it — Scobleizer · 2026-11-13
- Solo dev ships logic-puzzle roguelike built with Claude that screenshot-tests itself — TheCodingTutor · 2026-11-13
- One prompt to test how "open" an open-source model really is — alex_verem · 2026-11-13