Skip to content
WikiBusines

Wikipedia for AI startups · notability index 2026

Wikipedia page for an AI startup: where hype does not replace sources

In IT, SaaS, software, AI and cybersecurity, 63% of topics in our 2018–2026 assessments have low confidence of notability, 36% medium, 1% high. For AI startups our expert assessment is: "depends on sources". A page is realistic where independent press has written about the company or the model, not where there is a funding round, a valuation or a first place on a leaderboard.

Notability audit · from €490 Create a page · from €2,400

// 63% low confidence in the industry: our scoring pipeline, not the market · prices from the 11.2026 price list · paid edits are disclosed on behalf of WikiBusines

en.wikipedia.org/wiki/Company_Nindex 2026 · 63% low
ArticleTalkRead

Company N

From Wikipedia

Company N is an artificial intelligence lab that develops its own language models. The models are available through an API and a chat app.[1][2]

Type
AI lab
Products
own models, API
Founded
2021
  1. 1
    WiredTier 1editorial analysis of the model and the company
    ✓yes
  2. 2
    MIT Technology ReviewTier 2analysis of the safety approach
    ✓yes
  3. 3
    Ars TechnicaTier 1model test and critique of its limits
    ✓yes
  4. 4
    TechCrunchTier 2routine news item on a funding round, no depth
    ✗no
  5. 5
    arXivpreprint, not peer-reviewed
    ✗no
  6. 6
    Company leaderboardbenchmark from the company itself
    ✗no

significant independent sources: 3 of 6

passes: 3 significant independent

ArticleTalkRead

Company N

From Wikipedia

Company N is an app with an AI assistant for writing résumés and letters. It runs on a third-party provider's models; basic features are free.[1]

Type
AI product
Basis
third-party models
Model
freemium
  1. 1
    The VergeTier 2editorial article: app test
    ✓yes
  2. 2
    TechCrunchTier 2funding round news, retold release
    ✗no
  3. 3
    VentureBeatTier 3announcement of a new feature
    ✗no
  4. 4
    PR NewswireAvoidlaunch press release
    ✗no
  5. 5
    MediumAvoidfounder's blog
    ✗no
  6. 6
    Forbes ContributorAvoidcontributor column
    ✗no

significant independent sources: 1 of 6

fails: 1 significant independent

ArticleTalkRead

Company N

From Wikipedia

Company N develops an open language model and a library for running it. The model weights and code are published under an open license.[1][2]

Type
open source
Product
model, library
License
open
  1. 1
    Ars TechnicaTier 1analysis of the model, license and community
    ✓yes
  2. 2
    The RegisterTier 1technical article on a licensing dispute
    ✓yes
  3. 3
    GitHubrepository stars and forks
    ✗no
  4. 4
    Hugging Facemodel downloads
    ✗no
  5. 5
    VentureBeatTier 3release announcement
    ✗no
  6. 6
    arXivpreprint with results
    ✗no

significant independent sources: 2 of 6

borderline: 2 significant independent

ArticleTalkRead

Company N

From Wikipedia

Company N develops accelerators for training neural networks and a cloud platform for them. It sells compute by the hour and under contract.[1][2]

Type
AI infrastructure
Product
chips, cloud
Customers
labs
  1. 1
    ReutersTier 1news report on a chip supply deal
    ✓yes
  2. 2
    BloombergTier 1feature on the company's data centers
    ✓yes
  3. 3
    CNBCTier 1CEO comment on air
    ✗no
  4. 4
    TechCrunchTier 2funding round announcement
    ✗no
  5. 5
    PR NewswireAvoidpartnership press release
    ✗no
  6. 6
    Vendor testbenchmark from the company itself
    ✗no

significant independent sources: 2 of 6

borderline: 2 significant independent

auto
63%of topics with low confidence in the field
4–5weeks per page
90days of monitoring and guarantee
€2,400English Wikipedia page from

Why a page

Why an AI startup needs a Wikipedia page: trust, hiring and AI answers

A lot is written about AI companies, and it varies widely: from analyses to advertising. When a buyer, a researcher or an assistant checks a company, Wikipedia is usually on the first screen. A neutral article with independent sources answers "what is this company and its model" before someone else's hype or rumor does.

What a page gives an AI startup

01

Enterprise buyers and security

Procurement and security teams vet a model provider before a pilot: who is behind it, what is known about data and incidents. An article with sources works as a check you cannot fake on your own website.

Why it matters → 02

Hiring researchers

Researchers and engineers google a lab before a conversation. An article with history, products and criticism shows a mature team, not just a blog with releases.

Why it matters → 03

AI assistants and the Knowledge Panel

When ChatGPT, Gemini and Perplexity answer about AI companies, they rely on encyclopedic sources, and Wikipedia is among the first. Google's Knowledge Panel takes its facts from the same place.

Why it matters → 04

Investors and partners

Due diligence starts with a search. An article calmly answers who this is and what the press has written. To be honest: in the industry 63% of topics have low confidence, so a page is not possible for every startup.

Why it matters →

We professionally create, edit and maintain Wikipedia pages for AI startups

  1. Page creation

    An article about a company or a model: sources, infobox, structure and disclosure. For AI we first decide what the article is about, and we write it by hand from sources.

    from €2,400English · 4–5 weeks · 90 days of monitoring and guarantee

  2. Editing and updates

    New rounds, model versions, partnerships, leadership; we remove what is outdated and promotional, everything sourced.

    €700–€1,9002–3 weeks

  3. Ongoing support

    Model versions change fast, and the article gets outdated. We watch for edits and vandalism, update from new sources and respond to deletion nominations.

    from €720/year12 months of guarantee · monitoring of changes and vandalism

We always disclose paid edits on Wikipedia in the name of WikiBusines. Page protection is applied only by Wikipedia administrators.auto

Why Wikipedia matters →All services →

01Will it pass

Notability of an AI startup on Wikipedia: what counts and what does not

The rule for companies (WP:NCORP, built on the general WP:GNG) requires significant coverage in several independent reliable sources. Independence (WP:ORGIND) means material about the company, not from it. Depth (WP:CORPDEPTH) means that routine reports on funding rounds, model releases and appointments do not provide it. In AI this breaks most often: a lot is written about models, but most texts retell a release, a benchmark or a round.

Notability index 2026: IT / SaaS / software / AI / cybersecurity

  • 63% low confidence
  • 36% medium
  • 1% high

Source: WikiBusines notability index 2026 (dataset, CC BY 4.0). These are percentages for the whole industry for 2018–2026, not for AI startups alone and not Wikipedia statistics. The "depends on sources" rating for AI startups is a WikiBusines expert assessment from 11.2026, not part of the dataset.

WiredTier 1editorial analysis · staff writer byline

"Company N trained a model that really holds a long context, but independent researchers found unstable answers in tests, and the company itself gave no explanations."

Counts: independent assessment, there is criticism, people outside the company are quoted.

PR NewswireAvoidpress release · text from the company

"Company N, a leader of the new generation of AI, has released a revolutionary human-level model and raised a Series A round. The CEO announced this."

Does not count: written by the company itself, no editorial check, promotional tone.

Journal articlepeer-reviewedpeer-reviewed · about the technology

"The article describes the architecture of Company N's model and the test results. The company itself is mentioned only in the authors' affiliation."

Gray area: a source about the technology, not about the company. An article about the company needs editorial pieces.

auto

Types of AI companies: where realistic and where almost never

Click a type to highlight the closest profile in the widget at the top of the page.

WikiBusines expert assessment 11.2026, not a dataset.auto

Wikipedia for IT companies →Notability index: all 13 industries →Notability audit →

02Sources

Which outlets count for AI startups: business, technology and science press

What counts is the piece, not the outlet. The tier and WP:RSP status show how the Wikipedia community regards an outlet in general, while whether a specific text fits is decided by its author, length and independence from the company.

auto
Financial TimesTier 1generally reliable

United Kingdom · Generally reliable, especially for business news.

✓Editorial articles and investigations about an AI company, deals and regulatory decisions.

✗Sponsored content, partner sections and retold press releases.

BloombergTier 1generally reliable

Global · Generally reliable for news and business.

✓News and analysis about the company, data centers and compute contracts.

✗Bloomberg profiles: they rely on press releases, so they are only a source of facts (Tier 3).

ReutersTier 1generally reliable

United Kingdom · One of the world's leading news agencies, generally reliable.

✓News about deals, lawsuits, regulatory decisions and incidents that name the company directly.

✗Press releases posted on the platforms alongside news.

The Wall Street JournalTier 1generally reliable

United States · A leading newspaper of record, generally reliable.

✓Editorial pieces on the business model, funding and leadership of an AI company.

✗Columns by outside authors and pieces written off a press release.

The EconomistTier 1generally reliable

United Kingdom · Generally reliable. Distinguish news from analysis.

✓Analytical pieces on the company, the technology and the AI market.

✗A passing mention of the company in a piece on the industry.

CNBCTier 1

United States · Generally reliable for business and financial news.

✓Editorial reports on the company, deals and the market.

✗A CEO's comment on air and a retelling of a press release.

WiredTier 1generally reliable

United States · Generally reliable for technology and culture topics.

✓Investigations, technical analyses and tests of models and products.

✗Sponsored roundups and partner content.

Ars TechnicaTier 1generally reliable

United States · Generally reliable for articles related to science and technology.

✓In-depth analyses of models, their limits and vulnerabilities.

✗Short retellings of announcements without original analysis.

The RegisterTier 1generally reliable

United Kingdom · The Register is considered generally reliable for technology-related articles.

✓Technical articles on the company, licenses, infrastructure and incidents.

✗Short notes based on a press release.

MIT Technology ReviewTier 2

Global

✓Technical and analytical pieces on models, labs and the impact of the technology.

✗Sponsored content and roundups without editorial analysis.

The VergeTier 2generally reliable

Global

✓Product tests and editorial articles about the company.

✗Short news items that repeat a press release.

The InformationTier 2

Global · Paywalled but reliable.

✓Investigations and analysis of an AI company's business and product.

✗A mention in a market overview without a separate analysis of the company.

The New York TimesTier 1generally reliable

United States · A leading newspaper of record, generally reliable.

✓Reports and investigations on the company, lawsuits and the impact of the technology.

✗Opinion columns and mentions without analysis of the company.

The GuardianTier 1generally reliable

United Kingdom · Generally reliable. Op-eds should be treated as opinions.

✓Editorial pieces and investigations about the company and its models.

✗Opinion columns and the outlet's blogs (Tier 3).

AxiosTier 1generally reliable

United States · Generally reliable, but some consider it biased. Opinions should be attributed.

✓Editorial pieces on the company and AI policy with its own analysis.

✗Short news lines based on a press release and opinions without attribution.

Business InsiderTier 1generally reliable

United States · Reliability is situational. Culture section is generally reliable (Tier 1), other areas have no consensus.

✓Editorial articles about the company by a staff writer.

✗A mention in a "top AI startups" roundup and sponsored content.

TechCrunchTier 2no consensus

Global · Stronger for funded companies; lighter for fluff coverage.

✓Editorial pieces with original analysis and independent sources.

✗Routine funding round news (it gives no depth), CEO interviews, contributor sections and sponsored content.

VentureBeatTier 3no consensus

United States · There is no consensus on the reliability of VentureBeat.

✓Only selected editorial pieces, after checking each one.

✗Announcements of rounds and model releases, sponsored content.

CrunchbaseAvoiddeprecated

United States · In the 2019 RfC, there was consensus to deprecate Crunchbase, but also to continue allowing external links to the website.

✓An external link in the "External links" section.

✗A source in the text and proof of notability: the database lists everyone.

PR NewswireAvoidgenerally unreliable

United States · There is consensus that PR Newswire is generally unreliable, as press releases published on the site are not subject to editorial oversight.

✓Nothing: it is a press release distribution service.

✗Any piece: there is no editorial review.

Forbes ContributorAvoidgenerally unreliable

Global · User-generated content, not Forbes/Inc editorial.

✓Nothing: it is not the outlet's editorial content. Forbes staff writers are assessed separately.

✗Contributor columns, including those about your company.

MediumAvoidgenerally unreliable

Global · Medium is a blog hosting service.

✓Nothing: it is a blogging platform.

✗Founders' and researchers' blogs, the company's technical notes.

LinkedIn PulseAvoidgenerally unreliable

Global · LinkedIn is a social network.

✓Nothing: it is a social network.

✗Articles and posts by founders and employees.

TierTier 1 · reliableTier 2 · with caveatsTier 3 · situationalAvoid

WP:RSP statusgenerally reliableno consensusgenerally unreliable · deprecated

Tiers and statuses: WikiBusines' working classification based on Wikipedia consensus (WP:RSP), reviewed 11.2026. This is not an official Wikipedia list. arXiv, GitHub and Hugging Face are not on the list: they are primary sources and platforms, not outlets.

How we read a piece: five signals and a verdict

auto
WiredTier 1

Company N: what its model can do and where it goes wrong

editorial analysis · staff writer, tests and quotes from researchers outside the company

  1. ✓Independent of the company
  2. ✓Editorial piece, not contributor
  3. ✓Not sponsored
  4. ✓Substantial coverage of the company, not a mention
  5. ✓Not a press release disguised as news

Counts as an independent sourcepassed 5 of 5

TechCrunchTier 2

Company N raises a round and releases a new model

"news" · text matches the press release, only quote is from the CEO

  1. ✗Independent of the company
  2. ✓Editorial piece, not contributor
  3. ✓Not sponsored
  4. ✗Substantial coverage of the company, not a mention
  5. ✗Not a press release disguised as news

Does not count A retold release is not independent coverage.passed 2 of 5

Media list: 891 outlets →Media coverage →

03Risks

Why AI startup pages get deleted: five typical reasons

A reviewer does not assess the model. They look for promotional tone, sources without editorial review and a suspicion that the company itself is writing the article. "Revolutionary", "world's first", "human-level" and benchmarks without an independent source get a tag almost immediately.

auto

Company N

!
{{Advert}}The article is written like an advertisement and needs rewriting.

Company N created the world's first revolutionary human-level model.Outlet X described Company N's model as one of the language models for programming.[1]

no sources · ratings from the company's website

Why it gets deleted. "Revolutionary", "world's first" and "human-level" without an independent source read as advertising. Such text is deleted entirely rather than corrected.

how we do itNeutral language. We tie every assessment to an editorial piece, and we do not write claims of being first or "human-level" without a source.

Company N

!
{{Primary sources}}The article relies on primary sources.

The model scored the best result in the test and outperformed all others.[1]The result in the test is described in outlet X; the methodology was checked by independent researchers.[1]

[1] company blog · company leaderboard

Why it gets deleted. A result that the company itself measures and publishes is a primary source. Without an independent outlet that analyzed it, the claim of "best" does not stay.

how we do itWe present a benchmark only with an independent analysis and the name of the test. Own leaderboards and preprints are used as sources for individual facts.

Company N

!
{{Notability}}The sources describe the model, not the company.

Company N is known for its model, which reviews praise.[1]The model is covered in a separate article; about the company we write only what the sources contain.

[1] model test · nothing about the company

Why it gets deleted. A model review points to the notability of the model. For the company, coverage of the company itself is required: its history, leadership, deals, decisions. Open models are often written about more than those who released them.

how we do itFirst we decide what the article is about: the company or the model. We select sources for that, before submission.

Company N

!
{{Update}}Some of the information in the article is outdated.

Company N's latest model is last year's version; the latest round is A.Model versions and rounds are given with a date: "as of month, year".[1]

[1] last year's news · new version without a source

Why it gets deleted. Model versions, product names and rounds change fast. An article without updates becomes inaccurate and gets a tag, and then people edit it without sources.

how we do itOngoing support: we watch releases and rounds and update only from new editorial pieces.

Company N

!
{{AI-generated}}The text looks like it was generated by a language model.

Company N is a leading innovator striving to change the future of artificial intelligence.We write the text by hand: every sentence has a footnote to a verified source.[1]

generic phrases · some links do not open

Why it gets deleted. Reviewers reject machine text: generic phrases, invented sources, no verification. A chatbot draft raises the risk of rejection rather than lowering it.

how we do itWe write by hand from sources and open and check every footnote. Paid editing is disclosed in the name of WikiBusines.

Guarantees →Deleted Page Recovery →

04Article

An AI startup article on Wikipedia: structure, infobox, what is in it and what is not

An encyclopedic article consists of verified facts, each with a footnote to an independent source. Point the scanner at the sections: it shows which source supports which claim.

auto
en.wikipedia.org/wiki/Company_N
ArticleTalkRead

Company N

From Wikipedia

History

Company N was founded in 2021 by a group of researchers who had previously worked in a university lab.[1] In 2023 it released its first public model.[2]

  • 1The Informationcompany history
  • 2Reuterseditorial news item

Technology and products

The main product is a language model available through an API and a chat app.[3] An outlet's test notes the quality of the code and weak answers in languages with small text corpora.[4]

  • 3Wiredtechnical analysis
  • 4Ars Technicamodel test

Funding

In 2024 the company raised investment in a Series A round, the outlet reported.[5] The terms of the deal and the company's valuation were not disclosed.[6]

  • 5Reutersnews item on the round
  • 6Bloombergpiece on the deal

Criticism and safety

Researchers showed that the model can be made to reproduce fragments of training data.[7] Rights holders challenged the use of their texts in the training set.[8]

  • 7Wiredinvestigation
  • 8Reutersnews item on the lawsuit

Notes

Every claim has footnotes; we check that each leads to an independent piece. Links to the company website appear only in the "External links" section.

  • 1–88 footnotesall editorial
  • →External linkscompany website

What does not go into the article

  • ✗Slogans, mission and "building the future of AI"
  • ✗API prices and tariffs (they change)
  • ✗Results of own leaderboards and benchmarks without an independent source
  • ✗Photos and logos without a free license
  • ✗"Human-level", "world's first" and user numbers without a source

Page creation →Image checker →

05Process

From audit to page in 4–5 weeks: steps for AI startups

Five steps, and each has something specific to AI companies. After publication a 90-day window of monitoring, support and guarantee begins.

90-day windowday 90 of 90

If a published page cannot be defended after 3 attempts within this window, we refund 80%. Annual support extends the care by 12 months.

auto

How it works →Annual support →

06Pricing

How much a page costs for an AI startup: price list prices, no industry markups

AI companies have the same prices as other industries. We start with an audit: it shows whether there are enough sources and is credited toward the project.

Guarantee: 80% refund if a published page cannot be defended after 3 attempts within the 90-day window. Guarantee terms →

Price calculator →

07FAQ

Wikipedia for AI startups: frequently asked questions

Is a big round, a high valuation or a place on a leaderboard enough?

No. A round, a valuation and a position in a model ranking guarantee nothing on their own: routine reports on rounds give no depth, and databases like Crunchbase list everyone. A reviewer looks for several independent, significant pieces in different outlets that analyze the company rather than retell it.

Do preprints on arXiv and benchmarks count?

Not as a source of notability. A preprint is primary and not peer-reviewed, a leaderboard from the company itself is not independent, and GitHub stars and Hugging Face downloads say nothing about coverage. A peer-reviewed article works as a source of facts about the technology, while the company's notability is proven by editorial pieces.

Should we write a separate article about the model or about the company?

It depends on the sources. If independent outlets analyze the model itself and mention the company only in passing, the product is notable and the article will be about it (notability is not inherited from a model, investors or founders, WP:INHERITORG). So before submission we decide what the article is about and select sources for it.

Can a draft be written with ChatGPT?

We do not recommend it. Reviewers recognize machine text: generic phrases, invented or irrelevant footnotes, promotional tone. Such a draft gets a tag or a rejection, and trust for the next submission is lost. We write by hand from sources and open every footnote.

Can we skip the audit?

You can, but we do not recommend it: an audit from €490 in 3–5 days shows whether there are enough sources and is credited toward the project. Without it there is a risk of submitting a draft that gets rejected with a reviewer comment.

What if the index and the audit say "not yet"?

It means "not yet", not "never". There are routes: media coverage for independent sources, Wikidata, the SEO and LLM package and AI visibility. We return to a Wikipedia page when significant publications appear.

Full FAQ →Notability audit →

Related pages

Wikipedia for IT companiesParent page: software, cloud and the whole IT industry Wikipedia for cybersecurityNeighboring page: research counts, mentions do not Notability auditFrom €490, written conclusion Notability index13 industries, 2018–2026 assessments Media list891 outlets with tier and status

AI startup and Wikipedia: we will tell you honestly whether your company passes

Notability audit from €490: a written conclusion with a list of sources that count and a direct answer on whether a page is realistic now. The cost is credited toward the project. If the answer is "not yet", we will say so.

  1. 1
    Publications about the company and the modelLinks to analyses, reviews and investigations from the last 24 months.
  2. 2
    What the product and segment areLab, AI product, open source or infrastructure, for whom, since which year.
  3. 3
    GoalA Wikipedia page, Wikidata or AI visibility.

or write to us: team@wikibusines.com · Telegram

Notability audit · from €490 Create a page
  • Auditfee credited toward the project
  • 4–5 weekspage creation, if there are enough sources
  • 90 daysmonitoring, support and guarantee after publication