AllerScore distills thousands of reviews and sentiment signals into a single 1–10 safety score per allergen, per restaurant. Filter by what matters. Explore what's safe.

Every restaurant gets two scores: an AI-driven score distilled from Claude Sonnet 4.5's training on real public allergy sentiment, and a community score aggregated from diners who've actually eaten there.
We prompt Claude Sonnet 4.5 with each restaurant's name, city, and cuisine. The model draws on its training across the open web — review sites, food blogs, forums — to infer how safely each restaurant has historically handled the 8 major allergens. If Claude doesn't have reliable knowledge of a restaurant, it flags it as "Limited Data" rather than fabricating.
The AI returns a 1–10 safety score for each of the 8 allergens (peanut, tree nut, gluten, dairy, egg, shellfish, soy, sesame) plus a concise safety summary citing the specific ingredients, techniques, and cuisine traits that drive the score.
Diners submit their own per-allergen ratings and reviews. Community scores are kept entirely separate from AI scores so you can see where real-world experience agrees with or diverges from the public record.
Toggle the allergens you care about, set a minimum safety threshold, and the map rerenders with only restaurants that meet your bar. Use the tabs on each restaurant page to compare AI and community scores side-by-side.
Made with Emergent