PlateRank AI — Restaurant AI Search Optimisation
Home → Projects → PlateRank AI
AGENTIC AI · RESTAURANT VISIBILITY · BUSINESS ANALYTICS
I built PlateRank AI, an AI Search Optimisation (AISO) agent for one restaurant. It helps restaurant owners understand when AI assistants recommend their business, compare visibility against competitors, identify gaps, draft evidence-backed improvements, and measure changes through follow-up checks.
Built for the NYU SPS × Pulse Foundry hackathon in October 2026, the live case study focuses on Café Athena, a Greek and Mediterranean restaurant in Pacific Beach, San Diego.
System
AI Search Optimisation agent
Stack
Python · FastAPI · React · Vite · Tailwind CSS
Focus
AI visibility, evidence, competitors, and repeat measurement
The Problem
Customers increasingly ask AI assistants where to eat. A restaurant that is not recommended may lose potential customers before they open a traditional search page. Owners also need to understand why competitors appear in answers and whether optimisation efforts improve visibility.
What I Built
PlateRank combines seven product areas: Where You Rank measures mention rate, average position, share of voice, question type, and source links; Website Housekeeping drafts evidence-backed recommendations for FAQs, metadata, crawler instructions, structured data, llms.txt, About content, and landing pages; Social Media finds and verifies local creators, scores candidates, and drafts briefs and outreach; Business Value analyses verified review quotes about food, service, ambience, and value; Data Farming finds relevant Reddit communities, drafts owner-reviewed posts, and evaluates them with Promoter and Skeptic agents; Unified To-Do List ranks recommendations across modules; Did It Improve? reruns selected or all question categories and compares them with an earlier check.
Where You Rank
Measures mention rate, average position, share of voice, and results by question type while retaining source links.
Website Housekeeping
Reads website content and drafts evidence-backed improvements for discoverability.
Social Media
Finds local creators, verifies profiles, scores candidates, and drafts content briefs and outreach.
Business Value
Analyses verified review quotes and surfaces complaints and rarely mentioned areas.
Data Farming
Finds relevant communities and drafts posts for owner review using Promoter and Skeptic evaluation.
Live Case Study: Café Athena
One directional check of 36 questions using Claude and web search on October 23, 2026: Café Athena 83% mention rate with average position 1.2; Spitz – Pacific Beach 56%; Slowly Restaurant & Lounge 17%; Mr Shawarma 11%; Shawarma Shack 6%. These are not guaranteed or permanent rankings and must be rechecked.
Evidence, Safety & Trust
An evidence ledger classifies information as verified, unverified, or rejected. Only verified information feeds scores and drafted recommendations. Results are labelled LIVE, CACHED, SAMPLE, or FAILED; failed answers are excluded. The system does not invent reviews, ratings, creators, prices, or competitors, and unknown facts are flagged for owner confirmation. Recommendations are never published or sent automatically. The system does not certify dishes as allergy-safe or claim same-day ranking certainty.
Engineering Architecture
Backend: FastAPI and Python. Frontend: React, Vite, and Tailwind CSS. AI: Anthropic Claude with web search and web fetch. Data sources include OpenStreetMap, website page reads, verified review information, and sourced creator/community discovery. Testing includes offline checks with simulated AI and web responses, plus an optional live test.
Python · FastAPI · React · Vite · Tailwind CSS · Anthropic Claude · Web Search · OpenStreetMap