Yelp AI Business Analyzer

• 

  4. User asked a question, system processing.

• 

  3. The user taps to ask a question.

• 

  2. List of Businesses

• 

  1. Home page

• 

  5. Conversation showing User question and AI Assistant answer.

• 

  6. Show options to Print, Download or Share the conversation.

Yelp_Business_Analyzer

I built this app with small business owners in mind who don’t have access to a data analytics team but still want meaningful insight into how their business is perceived. I initially envisioned a more analytics-driven product, where a business owner could ask voice-based questions like “Graph my star rating from 2022 to 2025” and receive charts or trends in response. I set up the infrastructure for this approach early on.

As the project evolved, I discovered a limitation. The Yelp AI API cannot reliably provide concrete, date-specific review or rating data. While the AI can sound confident when answering year-based questions, those answers are not guaranteed to be correct. This forced a significant rethink of the product. In light of this limitation, I added guardrails to prevent direct answers to date-specific questions and reframed the app’s purpose.

The project shifted from providing analytics to providing common themes and overall sentiments that customers are relaying through their reviews. I now view this app as a tool for the tired business owner who wants a quick, honest snapshot of what customers are saying by asking questions like “What do customers complain about most?” or “What do people say about prices?” I kept the UI simple and uncluttered, with a share option for business owners who want to pass the insights along to their team. I’d like to explore voice-activated charts and more quantitative data in the future, this felt like a realistic starting point.

Throughout development, I learned a great deal about both technical workflow and AI-assisted coding. I learned about tools such as ngrok to test mobile phone layout without needing to deploy. I relied on Claude to generate code, however, I was actively involved in outlining the architecture, debugging issues, testing edge cases, and deploying the app. I’m not a programmer by training, and the result isn’t perfect, but after many long days of iteration, I’m proud to have built a working, thoughtful product that respects both the user and the data it relies on.

Built With

• ai
• api
• css3
• express.js
• fusion
• html5
• javascript
• multer
• node.js
• openai
• railway
• react.js
• tts
• vercel
• vite
• whisper
• yelp