City Reputation Reports (CRR) Methodology

This project follows the "mirror record principle": we consider AI-generated content as a "mirror" of global digital information within algorithmic models. This study does not conduct field research on cities themselves, but rather audits the clarity, accuracy, and narrative preferences of that "mirror."

• Objectivity: Only record the original information output by AI, without adding researchers' personal preferences.
• Neutrality: Do not rank, score, or judge cities as better or worse.
• Verifiability: Every report must include the original, immutable share link generated by the AI platform.

The audit observes five core dimensions of a city's presence in the digital environment:

1. Food narratives: AI's cognition of local cuisine, culinary reputation, and representative flavors.
2. Heritage and tradition: AI's descriptions of a city's historical heritage and intangible cultural preservation programs.
3. Festive rhythms: AI's records of recurring celebrations, events, and their cultural influence.
4. Landmark visuals: AI's weight assignment to physical architecture and natural landscapes.
5. Arts and creativity: AI's qualitative characterization of modern art, creative industries, and cultural atmosphere.

1. Prompting stage

A standardized prompting framework is used to ensure that different cities are observed under the same passive conditions on the same platform.

• Single-platform audit: Conduct deep multi-turn conversations on a specific platform (such as ChatGPT or Claude).
• Prompt protocol: Use a "neutral extraction prompt." Example: "Please describe the international reputation of [city name] in the dimension of [dimension], and list the core facts cited."

2. Evidence archiving — the core step

• Generate link: After the exchange is complete, create the platform's official share link.
• Immutability: Once generated, the link content is locked. The report treats this link as the sole evidence source, ensuring readers can trace and verify it.
• Original record: If the AI output contains errors or hallucinations, the report records that "existence" as-is without manual correction.

3. Data extraction and classification

Researchers extract the following objective elements from the original conversation:

• High-frequency keywords: Terms repeatedly mentioned by AI when describing the city.
• Cited sources: References explicitly named by AI (if any).
• Information gaps: Important known urban assets that AI failed to cover (i.e., "silent narratives").

4. Report writing

The report structure strictly follows these elements:

• Platform identifier: Clearly state the AI platform and model version used in the audit.
• Narrative summary: Objectively summarize how the platform describes the city.
• Evidence anchor: Insert the immutable original conversation link.
• Observed findings: Describe the patterns present in the AI narrative (for example: when 90% of the city's food description centers on a single dish).

1. Specialized audit
   • Definition: A scan of city dimensions conducted on a single AI platform.
   • Purpose: To present the city profile produced under a specific algorithmic logic, helping to understand that model's "first impression" of the city.
   • Publishing form: Regular briefings and reports.
2. Customized / cross-platform audit
   • Definition: A horizontal comparison of narratives about the same city across multiple platforms (ChatGPT, Claude, Gemini, Perplexity, etc.).
   • Purpose: To identify cross-platform consensus and divergence, and detect systemic information bias.
   • Applicable scenarios: Professional advisory work, internal government briefing, or in-depth cultural research.

1. Dynamism: AI models and their databases continue to evolve; each report only captures the narrative state at the moment of audit.
2. Non-exhaustiveness: AI descriptions do not represent the full reality of a city, only its projection within digital corpora.
3. Non-official nature: This project is an independent study and does not represent the position of AI platforms or city authorities.