Wikidata Entity Audit for Brands

Check how complete and well sourced your brand's Wikidata item is, and get a list of what to fix. Give the Actor your website domain, a Wikidata item ID or a name. It finds the item, scores it from 0 to 100 and lists each gap in plain words:

• labels and descriptions missing in the languages you care about (a "mul" default label counts for every language)
• key statements missing: instance of, official website, logo, and for organizations inception date, country, headquarters, industry and founders
• statements that cite no source, or only say "imported from Wikimedia project"
• Wikipedia articles: how many, and which of your languages have none
• Google Knowledge Graph ID (P2671), LEI, OpenCorporates, Crunchbase and official social accounts
• other Wikidata items that claim the same official website (duplicates, or a product and its company that should be linked)

Each row also carries a ready sameAs list and, for organizations, a schema.org Organization JSON-LD snippet built from the item.

Why use it

Search engines and AI assistants read Wikidata to decide what a brand is, which site and profiles belong to it, and what to show in knowledge panels. An item with no sources, no logo, a stale http:// website or a duplicate is a common reason for a wrong or missing panel. This Actor gives entity-SEO, PR and brand teams the checklist in one run, for one brand or a few hundred.

The Actor only reads. It never edits Wikidata. Fix the gaps yourself on wikidata.org, following Wikidata's rules on notability, sources and conflict of interest.

Input

• Website domains: e.g. cloudflare.com. Every item whose official website (P856) is the home page of the domain (http or https, with or without www, with or without a trailing slash) is audited. Up to 10 items per domain.
• Wikidata item IDs: e.g. Q4778915 or https://www.wikidata.org/wiki/Q4778915.
• Names: e.g. Notion. Searched on Wikidata; the best match is audited. Check the matched item in the results.
• Languages (default en, at most 10): languages to check labels, descriptions and Wikipedia articles in.
• Items per name (default 1, at most 5): how many search matches to audit per name.

At least one domain, item or name is required, and at most 1,000 in total.

Output

One dataset row per audited item. Real example from a run on raspberrypi.com with languages en, de (2026-11-01), shortened:

{
  "status": "audited",
  "input": "raspberrypi.com",
  "matchedBy": "website",
  "matchedWebsite": "https://raspberrypi.com",
  "qid": "Q65518082",
  "label": "Raspberry Pi Ltd",
  "entityType": "organization",
  "instanceOf": ["private company limited by shares"],
  "score": 68,
  "issues": [
    "No label in de",
    "No description in de",
    "No logo image (P154)",
    "No Wikipedia article in en, de",
    "No Google Knowledge Graph ID (P2671)",
    "No industry (P452)",
    "No \"founded by\" (P112)",
    "2 other item(s) list the same official website: Q245 (Raspberry Pi), Q129710524 (Raspberry Pi Holdings)"
  ],
  "inception": "2012-09-10",
  "country": ["United Kingdom"],
  "headquarters": ["Cambridge"],
  "wikipediaArticleCount": 1,
  "identifiers": { "lei": "894500E8FB2SGSN63230", "openCorporates": "gb/08207441" },
  "socialProfiles": { "linkedin": "https://www.linkedin.com/company/raspberry-pi-trading-limited/" },
  "referenceableStatements": 14,
  "sourcedStatements": 7,
  "unsourcedStatements": 7,
  "sameAs": ["https://www.wikidata.org/wiki/Q65518082", "https://search.gleif.org/#/record/894500E8FB2SGSN63230", "..."],
  "jsonLd": "{ \"@context\": \"https://schema.org\", \"@type\": \"Organization\", ... }",
  "lastModified": "..."
}

In the same run, the product item Raspberry Pi (Q245) scored 94 and the holding company Q129710524 scored 57.

Other fields: wikidataUrl, description, officialWebsites, logoUrl, industry, wikipediaUrls (your languages), missingWikipediaLanguages, missingLabels, missingDescriptions, aliasCount (in your languages), statementCount, wikiImportOnlyStatements, otherItemsWithSameWebsite, checkedAt.

Inputs that produce no audit get a free row with status not-found (no item has that home page, the ID does not exist, or nothing matches the name), skipped-person (the item is a person, see below) or error (the lookup failed), and the reason in issues.

The OUTPUT record in the key-value store holds the counts, the average score and the ten lowest-scoring items.

How the score works

Each check has a weight. The score is the weighted share of checks passed. Language checks and the source check give partial credit.

Check	Weight
Label in each of your languages	2
Description in each of your languages	2
Instance of (P31)	2
Official website (P856)	2
Official website uses https (only if one is listed)	1
Logo (P154); for items that are not organizations, a logo or an image (P18)	1
At least one Wikipedia article	2
A Wikipedia article in each of your languages (only if there is at least one article)	1
Google Knowledge Graph ID (P2671)	1
At least one official social account	1
Statements cite a real source (full marks at 50% or more)	2
Organizations only: inception, country, headquarters, industry, founded by	1 each

"Statements" here means the ones Wikidata expects to carry a reference. External IDs, URLs, media files and "instance of"/"subclass of" are not counted. A reference counts as a real source if it has anything besides "imported from Wikimedia project", "Wikimedia import URL", "retrieved", "based on heuristic" or "inferred from". Deprecated statements are ignored. Duplicate items are listed as an issue but do not change the score.

An item counts as an organization if its "instance of" class is within four subclass steps of organization, business, company or enterprise.

How it works

1. Domains are matched against official-website statements with one query to the Wikidata Query Service per 25 domains. Names use Wikidata's wbsearchentities search.
2. Items are fetched with the Wikidata API (wbgetentities), 20 at a time.
3. One more query finds other items that claim each audited item's website, and one classifies organizations.
4. Labels for countries, classes and other linked items are fetched in your first language.

Requests are paced, identify the Actor in the User-Agent, and are retried when Wikidata reports it is busy.

Data collected

Only public Wikidata content about the items you ask for. People are never audited: an item whose "instance of" is human is skipped with a skipped-person row that carries only the item ID, and people are dropped from name matches. Founders are checked for presence only; their names are not returned.

Pricing

Pay per event: each audited item is one entity-audited event. Not-found, skipped and failed inputs are free. A domain can match several items, and each of them is audited and charged. Use the run's maximum cost setting to cap spend. The run stops cleanly when the cap is reached.

Limitations

• Domains match only official-website values that are the home page. An item that lists https://www.example.com/en/ is not found by domain; add its ID under item IDs.
• The score is a completeness and sourcing checklist. It is not Google's view of your entity, and a high score does not guarantee a knowledge panel.
• Name search returns Wikidata's best text match, which can be a different thing with the same name.
• Wikidata is edited constantly. Results reflect the item at the time of the run (lastModified shows its last edit).
• If the Query Service is down, organization detection falls back to the item's direct "instance of" values; OUTPUT.organizationDetection says which was used.

Support

Open an issue on the Actor's Issues tab. Breakages are fixed within 48 hours. See also our Wikipedia Brand Mention Finder, which finds the Wikipedia articles that mention your brand and whether they link to you.

This Actor is built and maintained by an AI agent working for its publisher. A human owns the account and is responsible for it. It is not affiliated with or endorsed by the Wikimedia Foundation. Wikidata and Wikipedia are trademarks of the Wikimedia Foundation.