◀ All articles

Authority

Wikidata, Crunchbase and Wikipedia: the knowledge-graph triangle

July 3, 2026 · 4 min read

Three records that turn a website into a recognised entity, in the order worth pursuing them, with the notability rules stated plainly.


Three records do more than any other off-site work to turn a website into a recognised entity: a Wikidata item, a Crunchbase profile, and — if you qualify, which you probably do not — a Wikipedia article. They are structured, they are machine-read at scale, and they are the records other systems reconcile against.

They are also governed by rules that punish self-promotion, so the order and the manner matter.

Crunchbase — start here

The easiest of the three, and the one nearly every company can have.

Crunchbase is a structured company database read by a great many downstream tools. A complete profile gives a machine your founding date, location, size, funding, category and website in fields rather than prose.

Getting one. Create or claim the profile. Fill in every field you can support: legal name, founding year, headquarters, employee range, industry categories, a description that matches your schema description, and the canonical website URL. Add funding if you have raised; leave it empty rather than vague if you have not.

The one rule. Everything must match your other sources exactly. A Crunchbase founding date of 2021 and an About page saying "since 2020" is a contradiction you personally created, and it lowers confidence in both.

Then add the profile URL to your sameAs array so the connection is declared. Why that matters.

Wikidata — the underrated one

Wikidata is a free, structured knowledge base. Every item has a Q-number, properties with values, and references. It is queryable, it is dumped in bulk, and it is consumed by search engines, research pipelines and training corpora alike. It is, in a real sense, the closest thing to a public entity registry the web has.

Its notability bar is much lower than Wikipedia's. An item needs to describe a clearly identifiable entity and be supported by at least one serious, independent reference — not "be famous."

Doing it properly.

  1. Search first. An item may already exist. Improving an existing item is better than creating a duplicate, and duplicates get merged by editors who will notice who made them.
  2. Disclose your connection. Wikidata has a conflict-of-interest expectation. Declare it on your user page. Editors are far more tolerant of a disclosed employee adding accurate structured data than of an undisclosed one.
  3. Reference everything. Each statement wants a source, and the source should be independent where possible — a press article, a regulatory filing, a trade publication. Self-sourced statements are tolerated for uncontroversial facts like a website URL and disliked for claims.
  4. Stick to facts with a defined property. instance of (business), inception, headquarters location, official website, industry, founded by. There is no property for "leading provider," and trying to express one is how you get reverted.

What gets you reverted: marketing language in the description, unsourced claims, creating an item for a company with no independent coverage at all, and editing without disclosure. A revert is not fatal, but a pattern of them attracts attention you do not want.

Wikipedia — do not chase this

The notability bar is genuinely high: significant coverage in multiple reliable, independent secondary sources, where "independent" excludes your press releases, your funding announcements repackaged by a trade blog, and anything you paid for.

Most companies do not qualify. That is fine and normal.

If you do qualify, do not write it yourself. Editing an article about your own company is a conflict of interest and, done badly, produces an article that gets deleted along with a permanent record of the attempt. Use the Articles for Creation process, disclose your connection, and let an independent editor write it.

If you do not qualify, do the Wikidata item instead. It carries a surprising amount of the same benefit for machine readability, and the door is actually open.

Paid Wikipedia editing services: some are legitimate and disclosed; most are not, and Wikipedia actively hunts undisclosed paid editing. The failure mode is an article deleted plus a noticeboard thread with your company name in the title, which is the opposite of what you were buying.

The order, and the honest expectation

Crunchbase this week. Wikidata this month. Wikipedia only if a journalist has already written about you twice without being asked.

And calibrate: none of these produces a visible jump in a visibility check next week. What they do is make you resolvable — they give a machine trying to work out whether you are a real company something structured and independent to check against. That shows up as fewer hedges and more confident mentions, over months, and it is one of the few things in this field that compounds.

See how your own site scores

One scan checks your homepage, robots.txt, llms.txt, About page and JSON-LD, then hands you the copy-paste fixes. Free, no account needed for the first run.

Keep reading