Wikidata is an openly licensed structured database that feeds multiple training and retrieval pipelines. It is one of the few places where you can add a machine-readable fact about your company for free.
The licence is the reason it travels. All structured data there is released under CC0, which means anyone can ingest it without attribution or permission.
Most advice about Wikidata is either “get an entry” or “you cannot, you are not notable enough.” Both are wrong in the same way: they treat it as a directory listing rather than as a data source that other systems copy.
What is Wikidata, structurally?
A collection of items, each with an identifier and a set of statements. Its introduction describes items as uniquely identified by “a Q followed by a number, such as Douglas Adams (Q42),” with statements that “consist of a property and a value,” where properties carry their own P identifiers.
So a company in Wikidata is not a page. It is a node with typed edges: inception, headquarters location, industry, founder, official website. Each edge can carry a reference pointing at where the fact came from.
That shape is why it is useful to machines and unimpressive to humans. There is no prose to read and nothing to optimise, only facts that are either present with a source or absent.
Why does the licence matter more than the traffic?
Because traffic is not the mechanism. Wikidata’s licensing page states that “All structured data in the main, property and lexeme namespaces is made available under the Creative Commons CC0 License (Public domain),” and CC0 carries no attribution requirement at all.
A dataset that can be copied wholesale, without permission and without crediting anyone, gets copied wholesale. That is the entire reason a small structured database has outsized reach: it is the least legally complicated source of factual triples on the open web.
Nobody visits your Wikidata item. Systems ingest it, and that is the point.
This also sets expectations correctly. Adding an item will not produce a traffic line in your analytics, and judging it that way will make it look like a failure. Its payoff is that machines have a sourced answer where previously they had inference, which is the accuracy argument in what is an entity in AI search.
Does your company actually qualify?
Possibly, and the bar is not a Wikipedia article. Wikidata’s notability policy accepts an item if it meets at least one of three conditions, and the second is the relevant one for most companies: it refers to “an instance of a clearly identifiable conceptual or material entity that can be described using serious and publicly available references.”
- What it means
- The item links to a page on Wikipedia or another Wikimedia project.
- Who clears it
- Companies with an existing Wikipedia article. The narrowest route and the one most people assume is the only one.
- What it means
- A real, identifiable thing describable using serious and publicly available references.
- Who clears it
- Most operating companies with a registration, coverage or filings. This is the practical route.
- What it means
- The item is needed to make statements on other items more useful.
- Who clears it
- Companies already referenced as a value elsewhere, for example as an employer or a developer of a product.
The word doing the work is references. An item asserting facts with no sources is the kind that gets deleted, and rightly. If you cannot point at a filing, a registry, or independent coverage for a claim, do not add that claim.
How do you add or correct an item?
Five steps, and the discipline is entirely in the sourcing.
sameAs array in your Organization schema, so the assertion runs in both directions.On that last step, schema.org defines sameAs as the “URL of a reference Web page that unambiguously indicates the item’s identity. E.g. the URL of the item’s Wikipedia page, Wikidata entry, or official website,” and Google’s organization structured data documentation says the markup helps it “disambiguate your organization in search results.” A Wikidata Q-number is the cleanest value that property can hold.
Which properties are worth filling in first?
The ones that answer questions buyers and machines actually ask, and that will not need revisiting. Inception, headquarters location, country, industry, official website, and instance of business or enterprise. Six statements with references is a complete, defensible item.
Founder and chief executive are worth adding when they are documented, because they create the second hop that makes a graph useful: a question about your founder can be answered without any statement directly connecting your company to the answer.
Leave out anything that changes quarterly. Headcount, funding round and revenue all go stale, and a stale sourced fact is worse than an absent one because it is confidently wrong. The properties to prioritise are the ones you would still be happy to see quoted in three years.
What should you not do?
Do not write marketing copy into it. There is nowhere to put it, and attempting it is the fastest way to have the whole item scrutinised. The structure only accepts typed facts, which is the same reason positioning language does not survive into a graph, covered in what is a knowledge graph.
Do not assert anything you cannot reference. Do not create a second item because you could not find the first. And do not edit anonymously while pretending to be a disinterested party; the edit history is public and permanent, and a company caught quietly editing its own facts without sources loses the item and the argument.
Stick to what is documented and the conflict of interest resolves itself. Facts with references are welcome regardless of who adds them, because they are checkable by anyone, and an edit nobody needs to trust is an edit nobody needs to revert.
The same logic applies to correcting a wrong fact about yourself, which is the most common reason a company needs Wikidata at all. Change the statement, attach the source that proves it, and leave the edit summary explaining what you fixed. That is a stronger position than arguing about it later.
How do you tell whether it worked?
Ask engines attribute questions and watch for consistency. Founding year, headquarters, founder, industry. A resolved company gets the same answer from every engine and gets it without hedging.
Expect a slow clock. Retrieval-side changes can appear within days, but a structured fact influencing what a model knows travels through training pipelines and may not surface until a later model generation. That split between the two response times is covered in mention rate vs citation rate. Set the item up, point your markup at it, and check the attribute answers quarterly rather than weekly.