Most B2B software companies do not exist in Wikidata, which means they are structurally invisible to the knowledge graphs that Google and Gemini draw from — regardless of how good their SEO or their G2 reviews are. In a spot check of 20 venture-backed B2B SaaS companies run for this article on August 19, 2026, only 6 resolved to a correct company item through Wikidata’s own entity-search API. The rest returned nothing, or worse, resolved to an unrelated item entirely — a yoga posture for “Asana,” a muscle disorder for “Rippling.”
This is one layer of the broader question this site tracks under how AI assistants decide which brands to recommend: before an assistant can recommend a brand, it has to correctly identify which entity that brand even refers to.
What Wikidata actually is, and why AI engines care
Wikidata is the Wikimedia Foundation’s structured, machine-readable database of “items” — one per person, place, organization, or concept, each with a permanent identifier (a “Q-number”) and a set of sourced statements. As of the current Wikidata:Statistics page, the project holds roughly 122.5 million items; separately, Wikidata:News logged the creation of the site’s 140-millionth item on May 31, 2026 — the two counters track different things (current live items versus cumulative items ever created, net of deletions and merges), but both point to a graph that is still growing fast.
That graph matters to AI systems for a specific, documented reason. When Google shut down Freebase, it migrated the data into Wikidata: reporting from the time (Search Engine Land, 2014) confirmed Freebase would go read-only in March 2015 and close in 2016, with its data folded into Wikidata as one of the successor sources for Google’s Knowledge Graph. Wikidata has been a foundational input to that graph ever since, and Google’s Gemini, along with other assistants, draws on Wikidata-backed entity data to populate knowledge-panel-style answers. Schema.org’s sameAs property is the mechanical link: pointing a site’s Organization markup at its Wikidata Q-number is one of the cleanest, least ambiguous entity signals a machine can consume — cleaner than a brand name alone, which (as the spot check below shows) is frequently ambiguous.
The notability bar B2B brands actually have to clear
Wikidata’s own policy, Wikidata:Notability, states an item is acceptable if it meets at least one of three criteria: (1) it already carries a valid sitelink to a page on Wikipedia or another Wikimedia project; (2) it refers to “a clearly identifiable conceptual or material entity that can be described using serious and publicly available references”; or (3) it fulfills a structural need, such as being required to make statements on other items more useful. For a private B2B software company, criterion 2 is the realistic path, and “serious and publicly available references” in practice means a funding announcement, an SEC filing, a Crunchbase profile backed by press coverage, or trade-press reporting — not a press release on the company’s own site, and not a G2 or Capterra listing alone.
This is a materially lower bar than English Wikipedia’s notability standard, which generally requires sustained, independent secondary-source coverage sufficient to support an entire encyclopedia article. A company can clear Wikidata’s bar — and get a Q-number, structured statements, and a sameAs target — years before it could plausibly survive a Wikipedia deletion discussion. That gap is exactly where most funded B2B software companies currently sit: real, referenced, but absent from the graph.
The spot check: 20 B2B SaaS companies, 6 confirmed
To see how that plays out in practice, this article queried Wikidata’s public wbsearchentities API (the same lookup AI tooling and Wikidata-aware plugins use to resolve a name to an item) for 20 venture-backed B2B SaaS companies on August 19, 2026. The results:
| Company | Wikidata API result | Verdict |
|---|---|---|
| Notion | Q60747998 — “productivity software” | Correct item |
| Airtable | Q23016614 — “cloud collaboration service” | Correct item |
| monday.com | Q65073369 — “team & project management software” | Correct item |
| Gong.io | Q109592043 | Correct item (thin description) |
| Chargebee | Q141009096 — “subscription-billing company” | Correct item |
| Klaviyo | Q106631196 — “B2C CRM/marketing automation” | Correct item |
| Asana | Q466797 — yoga postures | Wrong item (name collision) |
| Rippling | Q18966134 — muscle disease | Wrong item (name collision) |
| Deel | Matched “volume” (book series term) | Wrong item (name collision) |
| Attio | Matched “80” (the number) | Wrong item (name collision) |
| Amplitude | Matched an unrelated 1999 journal article | Wrong item (name collision) |
| Ada (Ada Support) | Matched an unrelated 1994 journal article | Wrong item (name collision) |
| Ramp, Vanta, Drata, Ashby, Ironclad, Braze, Pendo, Metronome | No match returned | No item found |
Six of 20 — 30% — resolved cleanly. The other 70% either had no item at all or, more instructively, had a name that collides with an unrelated dictionary word or existing entity, meaning an automated entity lookup surfaces the wrong thing entirely rather than simply coming up empty. That distinction matters: “no item” is a visibility gap; “wrong item” is a disambiguation hazard that structured sameAs markup can’t paper over, because there is nothing correct on the other end of the link.
Why this specific gap matters for AI visibility
Separately from Wikidata coverage, the most recent large-scale citation study — “The State of AI Citations 2026,” published by 5W on June 2, 2026, based on more than 680 million tracked citations across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — found Wikipedia supplies roughly 47.9% of ChatGPT’s top-10 source share, while Reddit supplies about 46.7% of Perplexity’s. Only around 11% of domains are cited by both platforms, meaning no single content or entity strategy wins across every engine. Wikidata coverage doesn’t show up as a citation in that study — assistants don’t footnote a Wikidata Q-number the way they footnote a URL — but it is the backend plumbing behind the knowledge-panel-style facts (founding date, headquarters, category, ownership) that those citations often get checked against. A brand without a Wikidata item is more likely to have those facts assembled, unverified, from whatever secondary sources the model can find — which is a direct, documented path to the kind of entity confusion this site has covered under why AI chatbots hallucinate about brands.
Closing the gap without gaming it
The realistic sequence, in order: first, generate at least one serious, independent, publicly available reference — a funding round covered by trade press, an SEC/EDGAR filing, or comparable third-party reporting; self-published company blog posts and G2/Capterra listings don’t satisfy Wikidata:Notability criterion 2 on their own. Second, create the item with sourced statements (industry, founding date, headquarters, legal name) rather than marketing copy — Wikidata rejects promotional language and requires citations on contested claims. Third, add the Organization schema markup on the company’s own site with a sameAs pointing at the new Q-number, closing the loop between the brand’s own structured data and the external graph. Fourth, disambiguate deliberately: if the company name collides with an existing Wikidata item (as six of the names above do), the item description and aliases need to make the correct entity resolvable, not just present. None of this is a ranking hack; it’s the same evidentiary bar this site applies to entity consistency work generally — sourced, verifiable, and third-party.
FAQ
Does having a Wikidata item guarantee an AI assistant will mention my brand correctly?
No. Wikidata is one input among many that knowledge graphs and some assistants draw on; the 5W citation data above shows assistants lean heavily on Wikipedia and Reddit for citations, not Wikidata directly. A Wikidata item reduces the odds of entity confusion and gives structured facts a sourced home, but it doesn’t control what a model chooses to say.
Can any company just create its own Wikidata item?
Anyone can attempt it, but the item will be deleted if it doesn’t meet Wikidata:Notability — in practice, at least one serious, independent, publicly available reference such as press coverage of a funding round or a regulatory filing. Self-published sources and Wikidata edits made with an obvious conflict of interest are routinely challenged by volunteer editors.
Is a Wikidata item the same thing as a Wikipedia article?
No, and that’s the point of this piece. Wikidata’s bar (criterion 2: a describable entity with serious public references) is substantially lower than Wikipedia’s sustained-coverage notability standard. A company can be legitimately item-worthy on Wikidata for years before it has enough independent coverage to survive as a standalone Wikipedia article.
Last updated August 19, 2026. This page is refreshed as benchmarks and scores move.