Perplexity Search Images Scraper

Search Perplexity's public Images view with one query or a batch of queries. The Actor captures Perplexity's structured media_items response when available, uses the rendered page as a fallback, normalizes the records, and removes duplicates across the run.

Input

• mode: single or batch.
• query: one image-search query. In batch mode it is included first when supplied.
• queries: additional queries for batch mode.
• maxItems: global record limit across all queries (1–500).
• maxPages: image-result batches loaded per query (1–5).
• includeInlineAnswerImages: include usable images rendered in the answer as a DOM fallback.
• language and country: align locale and browser fingerprint settings.
• maxConcurrency, retry, navigation, result-wait, and handler timeout controls.
• proxyConfiguration: optional Apify or custom proxy configuration.

Example single query:

{
  "mode": "single",
  "query": "northern lights photography",
  "maxItems": 10,
  "maxPages": 1,
  "language": "en-US",
  "country": "US",
  "proxyConfiguration": {
    "useApifyProxy": false
  }
}

Example batch:

{
  "mode": "batch",
  "queries": [
    "northern lights photography",
    "Eiffel Tower night photography"
  ],
  "maxItems": 20,
  "maxConcurrency": 1
}

Output

Every stored record includes a stable id, position, query, image URL, thumbnail URL, displayed card title, source-page URL/domain when present, result type, page, search engine, and scrape timestamp. Source-provided dimensions, thumbnail dimensions, file metadata, creator/license fields, and extraction provenance are included when available. The Actor does not invent an original image URL or card title when the page does not expose one.

Unavailable optional fields are omitted instead of stored as null or empty strings. Raw API payloads, cookies, tokens, authorization data, and challenge pages are never stored.

Chrome observation

The supplied Chrome observation showed the public signed-out search session with Answer, Links, and Images tabs. After the sign-in prompt was dismissed without logging in, the Images view displayed 14 cards with title labels and source-page links, including Cats Protection, National Geographic, and Britannica pages. Those visible fields are the DOM fallback's extraction targets; this was a browser observation, not an Actor run or dataset sample. A current “Sign up below” gate can cover the page, so the Actor stops with TARGET_BLOCKED while that gate is present.

Proxy guidance

Direct access is the current working default. Perplexity may reject some proxy exit pools; a US Residential test returned HTTP 403 while a direct Apify Cloud run succeeded. Proxy support remains available for users whose target geography or network requires it. Keep maxConcurrency at 1 for anonymous public searches unless higher concurrency has been validated for your environment.

Failure behavior

• HTTP status and content type are validated before HTML or JSON processing.
• Malformed and unexpected API payloads are rejected safely.
• Explicit Cloudflare challenges, CAPTCHA pages, and sign-in prompts (including Perplexity's “Sign up below” gate) are classified as TARGET_BLOCKED, are not retried, and do not create dataset records.
• A genuine “No images match this query” response completes with an empty dataset.
• Temporary browser, proxy, rate-limit, and server failures use bounded retries.

The Actor does not solve CAPTCHAs or bypass authentication and other access controls.