CHILLNN AIO Lab Conducts Large-Scale Survey on 'Characteristics of Restaurants Recommended by AI'
CHILLNN's AIO Lab has conducted a large-scale survey in five major Japanese cities on how generative AI recommends restaurants. The study found that review volume and inclusion in summary articles are key, while star ratings showed no clear correlation.
📋 Article Processing Timeline
- 📰 Published: May 21, 2026 at 20:00
- 🔍 Collected: May 21, 2026 at 11:31
- 🤖 AI Analyzed: May 22, 2026 at 03:15 (15h 43m after Collected)
The use of generative AI such as ChatGPT and Gemini to search for restaurants is rapidly increasing. With the integration of Gemini into Google Maps (the AskMaps feature), AI-driven local search is expected to expand further.
However, for restaurant operators, the mechanism of "how to be recommended by AI" remains unclear. While there is prior research overseas, guidelines for AIO (AI Optimization) measures in the Japanese market have not been clarified. Therefore, this survey explored the characteristics of restaurants easily recommended by AI, specifically for the Japanese market.
## Overview of the Initiative
For approximately one month (April 2 to May 5), different prompts were executed against three types of AI—ChatGPT, Perplexity, and Google AI Overview—across five cities (Sapporo, Shinjuku, Kyoto, Osaka, and Fukuoka), and the restaurants recommended were aggregated on a large scale. The main findings are as follows:
- Restaurants recommended by AI show a positive correlation with the number of Google reviews, but not with the review score (star rating).
- There is no clear correlation between the number of Google search results for "restaurant name x area" (web exposure) and the frequency of AI recommendation.
- The frequency of appearance in "summary articles" referenced by AI significantly affects whether or not it is recommended.
However, the survey prompts used were limited to general phrases such as "recommended restaurants in XX." For prompts that include more personal conditions such as anniversaries, private rooms, or specific genres, there is a possibility that unique, small-scale shops other than famous stores could be recommended by AI. CHILLNN AIO Lab will continue to investigate and disseminate practical knowledge on recommendation trends with personal prompts and AIO measures in the restaurant sector.
The survey results are available for free on the AIO Lab media page.
However, for restaurant operators, the mechanism of "how to be recommended by AI" remains unclear. While there is prior research overseas, guidelines for AIO (AI Optimization) measures in the Japanese market have not been clarified. Therefore, this survey explored the characteristics of restaurants easily recommended by AI, specifically for the Japanese market.
## Overview of the Initiative
For approximately one month (April 2 to May 5), different prompts were executed against three types of AI—ChatGPT, Perplexity, and Google AI Overview—across five cities (Sapporo, Shinjuku, Kyoto, Osaka, and Fukuoka), and the restaurants recommended were aggregated on a large scale. The main findings are as follows:
- Restaurants recommended by AI show a positive correlation with the number of Google reviews, but not with the review score (star rating).
- There is no clear correlation between the number of Google search results for "restaurant name x area" (web exposure) and the frequency of AI recommendation.
- The frequency of appearance in "summary articles" referenced by AI significantly affects whether or not it is recommended.
However, the survey prompts used were limited to general phrases such as "recommended restaurants in XX." For prompts that include more personal conditions such as anniversaries, private rooms, or specific genres, there is a possibility that unique, small-scale shops other than famous stores could be recommended by AI. CHILLNN AIO Lab will continue to investigate and disseminate practical knowledge on recommendation trends with personal prompts and AIO measures in the restaurant sector.
The survey results are available for free on the AIO Lab media page.
FAQ
AIに推奨される飲食店の特徴としてどのような要素が挙げられますか？
Googleの口コミ数（レビュー数）と正の相関があり、またAIが参照する「まとめ記事」への登場頻度が推奨の有無に大きく影響を与えます。
AIの推奨頻度とレビュースコアには相関がありますか？
調査結果によると、AIの推奨頻度とレビュースコア（星の数）には明確な相関は見られませんでした。
どのような条件でAIは飲食店を推薦しますか？
今回の調査では一般的なプロンプトを用いましたが、記念日や個室などパーソナルな条件を含むプロンプトでは、有名店以外の小規模店舗が推奨される可能性があります。
調査はどの都市で行われましたか？
札幌、新宿、京都、大阪、福岡の5都市で約1ヶ月間（4月2日〜5月5日）調査が実施されました。
CHILLNNの事業内容は？
宿泊予約システムの開発・提供や、宿泊施設のマーケティング支援を行っています。