Bing Images Scraper

Search Bing Images and export normalized image records from Bing's structured result-card metadata. The actor uses the website search interface, scrolls the result grid, and keeps only records that contain a full image URL, thumbnail URL, title, and source page URL.

Input

• mode: single for one query or batch for multiple queries.
• query: the primary search query; required in single mode.
• queries: up to 20 additional deduplicated queries in batch mode.
• maxItems: maximum valid records, from 1 to 1,000. Default: 100.
• maxConcurrency: independent query pages processed in parallel, from 1 to 10.
• imageSize: any, small, medium, large, or wallpaper.
• safeSearch: strict, moderate, or off.
• market: Bing locale such as en-US or en-GB.
• maxScrolls and maxStallRounds: infinite-scroll stopping controls.
• maxRequestRetries, navigationTimeoutSecs, requestHandlerTimeoutSecs, and resultsWaitSecs: bounded reliability controls.
• proxyConfiguration: Apify Proxy or user-provided proxy configuration.
• debug: sanitized scroll-progress logging.

{
  "mode": "single",
  "query": "artificial intelligence",
  "maxItems": 15,
  "imageSize": "any",
  "safeSearch": "moderate",
  "market": "en-US",
  "proxyConfiguration": { "useApifyProxy": false }
}

Output

The 45-field output contract includes:

• Identity and ordering: stable id, imageId, contentId, thumbnailId, cacheKey, md5, position, page, and queryPosition.
• Image data: imageUrl, thumbnailUrl, previewImage, width, height, aspectRatio, sizePixels, fileFormat, mimeType, and isAnimated.
• Listing text: title, altText, description, and caption.
• Source data: hostPageUrl, sourceUrl, hostPageDomain, sourceName, sourceDomain, pageTitle, and domain.
• Search context: searchQuery, query, searchUrl, market, country, language, engineId, searchMetadata, and scrapedAt.

Malformed metadata, results without a source page, duplicate IDs/URLs, challenge pages, and invalid URLs are never pushed. Some fields remain optional because Bing does not expose every identifier on every card.

Extraction strategy

The actor parses the JSON stored in Bing's .iusc[m] result cards. Payloads are parsed defensively and mapped field-by-field; raw metadata is not stored. DOM values provide title/alt fallbacks and dimensions when needed. Crawlee sessions and consistent browser fingerprints are enabled, and proxy input is honored.

Local verification

npm ci
npm test
npm run validate:dataset
apify validate-schema
npm run start
apify run --purge --input-file INPUT.json

Additional test inputs are available under test/inputs/ for minimal, filtered, multiple-scroll, and unlikely-query runs. Bing may return fallback images for an unlikely query instead of an empty page; those are retained only when they are genuine structured image results.