Link Preview Metadata (OG + JSON-LD): Bulk URL Extractor avatar

Link Preview Metadata (OG + JSON-LD): Bulk URL Extractor

Pricing

from $0.95 / 1,000 preview extracteds

Go to Apify Store
Link Preview Metadata (OG + JSON-LD): Bulk URL Extractor

Link Preview Metadata (OG + JSON-LD): Bulk URL Extractor

Bulk link preview data for any list of URLs: Open Graph, Twitter Card, favicon and touch icon, og:image size, canonical, oEmbed, JSON-LD types and a 0-100 preview quality score with fix hints. Plain HTTP, failed URLs never charged.

Pricing

from $0.95 / 1,000 preview extracteds

Rating

0.0

(0)

Developer

Detroit Web Agent - Open Data

Detroit Web Agent - Open Data

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

2 days ago

Last modified

Features

Limited permissions

Share

Give it a list of URLs and get back, for each one, what a link preview card would show: title, description, image with its size, site name and favicon. You also get every Open Graph and Twitter Card tag, the canonical URL, oEmbed links, JSON-LD types and a 0-100 preview quality score with fix hints. It uses plain HTTP: no browser, no login, no proxy needed for most sites.

What data you get

  • Preview card: title, description, image (absolute URL), image width, height and alt, site name, favicon, apple-touch-icon
  • Tags: all og:* and twitter:* tags, canonical, <title>, meta description, lang, theme color, author, published and modified time
  • Discovery: oEmbed JSON/XML links, RSS/Atom feeds, web app manifest, JSON-LD @types
  • Fetch facts: final URL after redirects, redirect chain, HTTP status, content type, response time
  • Quality: previewQuality (0-100), letter grade, missing tags and plain-English fixes

Use cases

  • Unfurl links pasted into a chat, CRM, newsletter or bookmark app, with the image size known up front
  • Check landing pages before a campaign so none of them share as a blank card
  • Enrich a list of article or company URLs with canonical URL, language, dates and feeds

How it works

  1. Each URL is fetched over HTTPS with a normal desktop browser user agent. Redirects are followed (up to 10) and recorded.
  2. Only HTML answers are parsed. PDFs, images and other files are detected from the headers and skipped without downloading them.
  3. Pages are read up to 2 MB; larger pages are parsed from the first 2 MB (htmlTruncated tells you).
  4. Tags are read in the same fallback order link unfurlers use: Open Graph first, then Twitter Card, then plain HTML and JSON-LD.
  5. Relative image, icon and canonical URLs are turned into absolute URLs.
  6. Every URL gets a row, including failures, so you can match the output back to your list with inputIndex. Only successful previews are charged.

Input

FieldWhat it doesDefault
urlsPublic page URLs, one per line (up to 5,000). URLs without https:// get it added; duplicates are removed.required
maxUrlsUpper limit of URLs processed in one run.1000
requestTimeoutSecsTime limit per URL, redirects included (3-60 seconds).15
maxConcurrencyPages fetched at the same time (1-25).10
includeRawTagsKeep the full openGraph, twitterCard and icons objects. Turn off for a flat CSV.true
proxyConfigurationOptional Apify Proxy or your own proxies, for sites that refuse cloud IPs.off

Example input

{
"urls": ["https://www.python.org", "https://developer.mozilla.org/en-US/docs/Web/HTML", "apify.com"],
"requestTimeoutSecs": 15,
"maxConcurrency": 10
}

Output

One dataset row per input URL. The Previews view shows the card fields and score, SEO tags shows title, description, canonical, language and JSON-LD types, and Fetch status shows status codes, errors and what was charged. Download as JSON, CSV, Excel or HTML, or read it through the API. A run summary (counts, errors by code, charged events) is saved as the OUTPUT record in the key-value store.

Sample output

Trimmed row for https://www.python.org (the raw openGraph, twitterCard, icons and ogImages objects are left out here):

{
"inputUrl": "https://www.python.org/",
"url": "https://www.python.org/",
"statusCode": 200,
"ok": true,
"billed": true,
"title": "Welcome to Python.org",
"description": "The official home of the Python Programming Language",
"image": "https://www.python.org/static/opengraph-icon-200x200.png",
"imageWidth": null,
"siteName": "Python.org",
"favicon": "https://www.python.org/static/favicon.ico",
"faviconSource": "declared",
"appleTouchIcon": "https://www.python.org/static/apple-touch-icon-144x144-precomposed.png",
"ogType": "website",
"lang": "en",
"feedUrls": ["https://blog.python.org/feeds/posts/default?alt=rss", "https://www.python.org/downloads/feed.rss"],
"jsonLdTypes": ["WebSite", "SearchAction"],
"previewQuality": 78,
"previewGrade": "B",
"missingTags": ["og:image:width/height", "og:image:alt", "twitter:card", "canonical"],
"hints": ["Add og:image:width and og:image:height so Facebook and LinkedIn render the image on the very first share instead of after a background scrape.", "..."],
"fetchedAt": "2026-11-08T06:01:46.070Z"
}

A failed URL looks like {"inputUrl": "https://example.com/missing", "ok": false, "error": "http-404", "billed": false, ...}. Error codes include timeout, dns-not-found, connection-refused, tls-error, http-403, http-404, http-500 (any status as http-<code>), not-html, empty-body, too-many-redirects and blocked-non-public-url.

Preview quality score

The score adds up points for the tags that decide how a shared link looks: og:title (12), og:description (8), og:image (18), declared image size (4), https image (3), image alt (3), og:url (4), og:type (4), og:site_name (4), twitter:card (10), a 10-70 character title (8), a 50-170 character meta description (6), canonical (5), a declared favicon (5), apple-touch-icon (3) and html lang (3). Grades: A 90+, B 75+, C 60+, D 40+, F below 40. Every missing item comes with a one-line fix in hints.

Pricing

Pay per event, no monthly fee:

  • Preview extracted: $0.00095 per URL that returned an HTML page and was parsed ($0.95 per 1,000).
  • Actor start: $0.00005 per run.

URLs that time out, fail DNS, answer 4xx/5xx, are not HTML, or are private/local addresses are listed in the dataset but never charged. If no input URL is usable the run stops at once with a clear message. Set "Maximum cost per run" in the run options to cap spend; the actor stops cleanly at that limit.

Call it from code

Python:

from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("transparent_meteorite/link-preview-metadata-extractor").call(run_input={"urls": ["https://www.python.org", "https://github.com"]})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
print(row["url"], row["title"], row["image"])

cURL (starts a run and waits for the rows):

$curl -X POST "https://api.apify.com/v2/acts/transparent_meteorite~link-preview-metadata-extractor/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" -H "Content-Type: application/json" -d '{"urls": ["https://www.python.org"]}'

The actor also works as a tool for AI agents through the Apify MCP server (https://mcp.apify.com), using the same input.

FAQ

Does it run JavaScript? No. It reads the HTML the server sends, which is also what Slack, LinkedIn, X and most unfurlers read. Pages that only add their tags with client-side JavaScript will show the tags present in the server HTML.

Why did a site return 403? Some sites refuse requests from cloud data centers. Turn on Apify Proxy in proxyConfiguration and retry those URLs. The actor does not bypass captchas or logins; pages behind them come back as errors and are not charged.

Is the favicon always real? faviconSource says declared when the page links an icon and default-guess when the actor fell back to /favicon.ico on the site root.

Can I schedule it? Yes. Save a task with your URL list and add a schedule; each run writes a new dataset you can send to a webhook, Make, Zapier or n8n.

Limits

  • Up to 5,000 URLs per run; 2 MB of HTML per page; 10 redirects per URL.
  • Only public http(s) URLs: localhost, private IP ranges and non-web schemes are refused.
  • Tags added only by client-side JavaScript are not seen.
  • A page answering 202 or another non-200 success code (often a bot check) is reported as an error, not parsed.