Restaurant intelligence that makes AI answers actually useful.
Already informing 63,000+ monthly AI dining queries across 25+ platforms. Structured, dimensional, evidence-backed.
MCP-protocol compliant. Point your agent at the endpoint and start querying.
The demand signal is clear
These are Seemor's numbers with zero marketing spend and no platform partnerships. Every retrieval below was initiated by an AI platform that discovered the corpus on its own.
The quality gap
Same query, same day. "A quiet dinner in Mayfair."
Generic AI Output
60+ secondsA calm, serene fine dining restaurant in the heart of Mayfair.
Elegant French dining with a relaxed, sophisticated atmosphere.
A civilised Michelin-starred Italian restaurant, not rowdy at all.
Intimate wine bar and restaurant with a refined, quiet setting.
Upscale Indian cuisine in a relaxed Mayfair townhouse.
Unranked. No way to know which is best for your specific need.
Noise described subjectively: "calm," "serene," "civilised." No measurement.
Ends with: "If you tell me your budget, I can narrow it down."
With Seemor Intelligence
8 secondsElegant chef-driven Mayfair townhouse with standout service and an acclaimed bread program. Lunch is strong value; dinner tasting is a premium splurge.
Intimate nine-seat omakase with aged-fish artistry and warm chef interaction. Pricey but memorable if you enjoy refined Japanese cuisine.
Destination fine dining with polished service and creative tasting menus. Superb for celebrations, less ideal for conservative palates.
Inventive fully vegan sushi in a serene courtyard setting. Excellent flavors and creative presentation, but expect premium pricing and occasional slow service at peak times.
Elegant Michelin-starred Italian with polished service and refined tasting menus. Lunch is strong value; dinner has occasional seasoning and pacing inconsistencies.
How it works
Trust the words, not the rating. Seemor reads reviews to determine what they are actually saying, then scores every restaurant across 37 quality dimensions.
Ingest reviews at scale
For every restaurant, Seemor reads dozens of reviews with full reviewer metadata. Not star ratings. The actual words. What people loved, what frustrated them, what surprised them. The rating is a number someone picked. The review is what they actually experienced.
Extract menus and metadata
A dedicated pipeline discovers and parses restaurant menus across PDFs, images, embedded widgets, and platform-specific formats. Menu data powers dietary scoring, price calibration, dish recommendations, and catches gaps that reviews alone miss.
Analyze across 37 dimensions
Every restaurant is scored on food quality, service speed, noise level, formality, kid-friendliness, cuisine authenticity, value, reservation difficulty, portion size, dietary options, and 27 more dimensions. Each score is backed by specific evidence extracted from review text.
Detect review authenticity
Not all reviews are equal. Reviewer credibility is weighted: a long-time local regular counts for more than a first-time tourist. Pattern analysis identifies review manipulation: solicited reviews, gaming patterns, fake review campaigns, and tourist-trap inflation. An authenticity assessment is built into every restaurant profile. A 4.8 with manipulated reviews is not the same as a 4.8 with genuine ones.
Grade deterministically
A deterministic formula converts dimensional scores into letter grades (A+ through F). The same restaurant gets the same grade every time. LLM scoring feeds the formula; the formula produces the grade. This separation makes the system auditable and reproducible.
Why this data is different
Raw reviews require an LLM to guess. Structured intelligence delivers verified answers.
Structural independence
Review platforms monetize the restaurants they rate. The moment a platform tells users "this restaurant is gaming reviews" or gives a place a C+, there is a revenue consequence. Seemor has no restaurant relationships, no ad revenue, no business model to protect. Every grade is editorial, not commercial.
Calibration, not just a model
172 prompt iterations. Dozens of real-world validation failures, each one adding a dimension the models could not see. Kid-friendliness scoring that accounts for actual menus, not just "staff were nice to children." Authenticity detection trained on real review manipulation patterns. Noise levels verified across hundreds of reviews, not keyword counts. This is domain expertise encoded into a system, not a prompt passed to GPT.
37 dimensions, pre-computed
Every analyzed restaurant has a complete intelligence profile: food quality, service speed, noise level, formality, authenticity, kid-friendliness, value, reservation difficulty, dietary options, and 28 more. All scored, all stored, all queryable. An AI platform using raw Yelp reviews would need to re-analyze every restaurant on every query. Seemor has already done the work.
A dataset that improves with every use
Hundreds of active users rely on the consumer product for real dining decisions. Every concierge query tests the scoring system. Every user confirmation or correction refines a specific dimension. Every restaurant visit validates or challenges the analysis. This creates a compounding data advantage no competitor can replicate by crawling reviews, because the refinement signal comes from the product itself.
Tool catalog
MCP-protocol-compliant tools that replace HTML scraping with structured intelligence. Free tier returns more structured data than Yelp's paid API.
Two tools are live today. The remaining tools package existing production infrastructure (the analysis engine, concierge pipeline, and comparison system) for MCP delivery. The underlying systems are production-hardened with months of consumer usage.
lookup_restaurantLiveFreeSingle restaurant by ID. Returns grade, TL;DR, cuisine, hours, neighborhood, amenities.
search_restaurantsLiveFreeSpatial search by coordinates with optional cuisine, price, and grade filters. Sorted by grade or distance.
recommendComing Soon$80/1K callsNatural-language dining query in, ranked and reasoned results out. Occasion-aware, dimensionally-scored, evidence-backed.
ask_about_restaurantComing Soon$50/1K callsNatural-language Q&A against a restaurant's full analysis profile. 'Is it loud?' 'Good for kids?' 'Worth the price?'
explore_areaComing SoonFreeNeighborhood-level dining intelligence. Restaurant counts, grade distributions, top picks, dining character.
compare_restaurantsComing Soon$50/1K callsOccasion-aware side-by-side comparison of 2-4 restaurants with a recommendation and reasoning.
Data coverage
Analysis depth varies by region. Major cities (London, Paris, Rome, NYC, Barcelona) have 70%+ analyzed coverage. Expansion is pipeline-driven and continuous.
700,000+
Restaurants catalogued
Name, address, coordinates, cuisine, price, hours across 134 regions
55,000+
Fully analyzed
37-dimension profiles with letter grades, evidence, and dimensional scores
26 countries
134 regions worldwide
London, NYC, Paris, Rome, Tokyo, Sydney, Bangkok, Seoul, and 126 more
Unanalyzed restaurants return basic data (name, address, cuisine, price) with a coverage_level flag. Analysis expansion is prioritized by API query volume.
Pricing
Field-level pricing, following the Google Places API model. Pay for the depth of intelligence you need.
Free
$0/ rate-limited
Replace HTML scraping with structured access
- -Name address coordinates
- -Grade + grade label
- -TL;DR summary
- -Cuisine tags price level
- -Hours amenities
- -Neighborhood city
Standard
$20/ per 1,000 calls
Curated intelligence beyond the crawl surface
- -Everything in Free
- -Full narrative summary
- -Occasion fit assessment
- -Menu must-tries + signature dishes
- -Pro tips from review analysis
- -Cost per person estimates
- -Grade reasoning
- -Dietary overview
Premium
$50/ per 1,000 calls
Dimensional intelligence from the scoring engine
- -Everything in Standard
- -Noise level assessment
- -Formality + service style
- -Cuisine authenticity
- -Local vs tourist clientele
- -Review reliability
- -Value assessment
- -Reservation difficulty
- -Natural-language Q&A
Concierge
$80/ per 1,000 calls
Full recommendation pipeline
- -Natural-language query input
- -Ranked results with reasoning
- -Occasion-aware scoring
- -Evidence-backed match rationale
- -Candidate evaluation context
Enterprise
Custom volume pricing, SLAs, dedicated support, and enhanced field access.
For reference: Yelp charges $25/1,000 calls for raw reviews and ratings. Seemor's free tier delivers structured intelligence including letter grades and TL;DR summaries at no cost.
Request access
Provide your details and intended use case. Access requests are reviewed promptly.
About
Seemor is built by Ryan Fuller, co-founder of VoloMetrix (now Microsoft Viva Insights), former Corporate Vice President at Microsoft, and author of 10 articles in the Harvard Business Review on data-driven decision making.
At VoloMetrix, Ryan built a structured data platform that turned noisy workplace collaboration data into organizational intelligence. The approach here is the same: messy, unstructured data transformed into calibrated, dimensional intelligence through domain expertise and real-world validation.
The restaurant vertical is built on 172 prompt iterations, real-world dining validation across dozens of countries, and a deterministic grading formula that separates LLM scoring from grade calculation. The methodology is transferable to any domain where decisions matter and existing information is noisy.
Seemor is a Larracos Labs venture. Bootstrapped. Independent.