Yelp

We wrapped Yelp so you can work local-business data the same way you work companies and people: search businesses in a place, resolve each to a full page, page through their reviews. Ratings, review counts, opening hours, service lists, and phone numbers come back normalized (E.164 phones, ISO country codes), so the output drops straight into your own models.

What it does

• Business search. Terms plus a location ("Austin, TX", a ZIP, or a full address); 10 businesses per page, sorted by relevance, rating, or review count.
• Business page. The full listing for one business: hours by day, active services, price level, address, website, images, review highlights.
• Reviews. Up to 49 reviews per page with text, rating, date, author details, and photos; filterable by star rating, keywords, language, and six sort orders.

The three compose: search gives you placeIds, the page and reviews operations take them from there. If you already know the business, the page operation also accepts the URL slug from a Yelp listing link.

Inputs and outputs

Search takes location (required), up to 5 categories or search terms, a sortBy, and a nextPageToken. Each business comes back with placeId, name, categories, rating, reviewCount, phoneNumber, priceLevel (1 to 4), neighborhood, and URLs.

Business page takes a placeId (business ID or listing slug) and returns the whole listing: openingHours per day, services with an active flag, directionsUrl, isClaimed, neighborhoods, and more.

Reviews take a placeId plus optional sortBy (relevance, newestFirst, oldestFirst, highestRated, lowestRated, elitesFirst), ratings (1 to 5 stars, up to 5 values), up to 5 keywords (one matching phrase), languageCode, and a nextPageToken. Each review carries text, rating, ISO date, authorName, authorLocation, authorIsElite, feedback counts, and photo URLs. The response also reports totalReviewCount.

Pagination, scopes, and caps

Pages are fixed size: 10 businesses for search, 49 reviews for reviews; the business page is a single response. The nextPageToken wraps your position and is scoped to the operation that produced it: a search token cannot paginate reviews and vice versa. A token that does not parse returns a 400 rather than silently restarting.

The 5-term caps on search categories and review keywords exist because the engine takes them as a single search-box string; past a handful of terms, results collapse toward empty intersections.

Using it effectively

• Start with search, then fan out. One search page seeds 10 placeIds for detail and review pulls; that is the cheapest unit of local-market coverage we offer.
• Sort reviews to fit the job. newestFirst for monitoring, lowestRated with keywords like "refund" for complaint mining, elitesFirst for the most trusted takes.
• Use ratings filters to split sentiment. Two calls, ratings [4,5] and ratings [1,2,3], give you clean positive and negative corpora for the same business.
• Slugs work where we have seen the business. The page operation takes slugs directly; the reviews operation resolves a slug only if search or a page pull has touched that business before. A raw business ID always works everywhere.
• priceLevel can be null. It parses Yelp's $ symbols; listings with non-symbol price strings return null instead of a wrong number.

Use cases

• Local-market scans: dentists, garages, restaurants in a city, enriched with review volume and ratings.
• Reputation monitoring: newest reviews across a chain's locations, keyword-filtered.
• Lead enrichment for local sellers: hours, services, and a normalized phone number per location.

Operations

Each call costs 2 credits; a search or reviews page that comes back empty is refunded, and a business page that cannot be found is refunded. All three run at 120 requests per minute. See Billing for the full pricing table, and Hotels and Google Maps for the rest of the local-business family.