The WikiBusines Wikipedia Risk Report 2026: What Public Drafts, Deletions, and Source Patterns Reveal

Answer box: who actually gets a Wikipedia page in 2026, and how risky is it?

Most new Wikipedia submissions are declined, not published. In a public sample of 1,009 Articles for Creation (AfC) drafts from September–October 2025, Lumino Digital reported a 68% overall rejection rate, with business topics accepted only ~16% of the time and startups/technology ~6% — while arts and culture reached ~48% (Lumino Digital, AfC Approval Rates, 2025). The dominant reason for decline is notability and sourcing, not formatting. The single biggest lever you control before spending money is source quality: independent, reliable, in-depth coverage. Founder and company pages fail more often than creators because their available sources skew promotional. The honest read for 2026: a Wikipedia page is achievable for a minority of subjects, the risk is real and largely predictable from your source list, and no provider can guarantee a community outcome. Score your sources first; spend second.

TL;DR

• Rejection is the base rate, not the exception. Public AfC data shows roughly two-thirds of new drafts declined, with business at ~16% and startups ~6% accepted (Lumino Digital, 2025). Plan for risk, not certainty.
• Source quality predicts almost everything. Independence, reliability, depth, and freshness separate accepted drafts from declined ones — far more than design, tone polish, or word count.
• Risk is not uniform. Companies and founders carry higher risk than established creators and cultural subjects, because their coverage is more promotional and less independent.
• AI-generated drafts are now a hard fail. Wikipedia's Articles for Creation process states that drafts "generated entirely by LLMs will be rejected" — and reviewers increasingly flag AI-pattern text.
• You can measure your own risk before paying anyone. The Source Strength Index below turns a messy source list into a single 0–100 score and a go / fix / wait decision.

The Source Strength Index: a public-data risk model you can run yourself

Wikipedia decisions feel opaque, but the inputs that drive them are surprisingly legible. After reading hundreds of public declined drafts and Articles for Deletion (AfD) discussions, one pattern repeats: outcomes track the source package, not the subject's self-belief. So we built a scoring instrument around exactly the signals reviewers cite, and we are publishing it openly.

The Source Strength Index (SSI) is a 0–100 score that estimates whether a body of sources is strong enough to support a stand-alone article under Wikipedia's notability rules. It scores six factors that map directly to policy language, then sorts the result into a clear decision band. It is deliberately conservative: it is designed to talk you out of a doomed submission as readily as into a viable one.

The six SSI factors:

1. Independence — Is the coverage genuinely independent of the subject, or is it reprinted PR, sponsored content, founder interviews, and the company's own channels? Wikipedia's organisations guideline requires "the notice of reliable sources unrelated to the organization or product" (WP:NCORP, "verifiable evidence that the organization or product has attracted the notice of reliable sources unrelated to the organization or product").
2. Reliability — Do the outlets have "a reputation for fact-checking and accuracy" (WP:RS, "Articles should be based on reliable, independent, published sources with a reputation for fact-checking and accuracy")? Trade blogs and pay-to-publish outlets score low; established newspapers of record score high.
3. Depth — Does each source address the subject "directly and in detail" (the General Notability Guideline's bar for "significant coverage"), or is it a passing mention, a funding-round line item, or a directory entry?
4. Independence of authorship / freshness — Is coverage spread across time and authors, or clustered in one campaign week? A single launch-week press splash is fragile; sustained coverage over years is durable.
5. Verifiability — Can a stranger check each claim against the source (WP:V, "verifiability means that people can check that facts or claims correspond to reliable sources")? Paywalled-but-real counts; vanished or unverifiable links do not.
6. Topic-type fit — Companies and founders are held to stricter, more skeptical standards than creators or cultural subjects. The same source list is worth fewer points for a startup than for a museum.

Each factor is scored 0–5, weighted, and summed to 100. The full scorecard and weights are in the framework table below. The decision bands:

• 70–100 — Defensible. Sources plausibly clear the notability bar. Proceed with neutral drafting and disclosure.
• 45–69 — Fixable. A real article may be possible after closing specific gaps (usually independence and depth). Audit, then decide.
• 0–44 — Not yet. Submitting now mainly buys a public decline and a harder second attempt. Build coverage first.

The SSI is a risk-reduction tool, not a guarantee engine. A high score lowers your odds of a decline; it cannot promise a community will agree, because no method can (see what we will not promise). Below, we apply this lens to the public 2026 evidence.

Methodology: public data first, our own figures labelled indicative, nothing fabricated

This report's backbone is public, citable evidence. Where our own numbers add context, we include them as clearly-labelled indicative estimates — directional, rounded, never invented precision — and we keep them visibly separate from the attributed third-party data. A risk report that dressed up guesses as hard statistics would be exactly the kind of thing this report warns you about; an honestly-labelled indicative estimate, marked as such every time, is a different thing.

What we used:

• Public benchmark datasets. The headline approval and rejection figures come from Lumino Digital's published Articles for Creation Approval Rates analysis of 1,009 AfC submissions sampled in September–October 2025 (overall ~32% accepted / 68% declined; business ~16%; startups/technology ~6%; arts and culture ~48%; median review time ~30–32 days). These are Lumino's numbers, attributed as such — not measured by WikiBusines.
• Wikipedia's own statistics and process pages. Wikipedia maintains public statistics, and AfC/AfD are fully public processes. Anyone can read Wikipedia:Statistics, browse declined drafts, and read deletion debates verbatim.
• Public AfD discussions. Every Articles for Deletion debate is archived and readable, including the closing rationale. AfD runs "for at least seven days, after which a decision may be reached based on community consensus" (WP:AFD). We read patterns in why business and biography articles are nominated and what arguments win.
• Declined-draft patterns. AfC decline reasons are templated and public; the recurring themes (notability, independent sourcing, promotional tone) are observable without any private data.
• Public source packages. For anonymised pattern analysis, we classify the types of sources attached to public drafts (press release vs. national newspaper vs. trade blog vs. database listing) — never the identities of specific subjects.
• Source-type classification. A consistent taxonomy (independent journalism, regulatory/official records, trade press, owned/PR, user-generated, database/aggregator) applied to the above.

What we did NOT do:

• We did not fabricate any statistic. Every third-party number is attributed to a named public source; the few figures that are our own are flagged indicative and are directional estimates from our practice, not invented precision.
• We did not name clients, reveal operational secrets, or publish anything that identifies a private individual.
• We did not present third-party figures (Lumino's, Wikipedia's) as WikiBusines's own measurements.

A note on the WikiBusines figures. Where this report cites our own numbers, they are indicative operational estimates — directional figures drawn from our practice across 16 language editions and 23 in-house editors, rounded and offered for context, not a formally audited dataset. We label them indicative wherever they appear, and we will replace them with audited figures in a future revision. Third-party figures (Lumino, Wikipedia) are always attributed to their source and never blended with ours. We would rather show you an honestly-labelled estimate than either a comforting fiction or a silent gap.

A note on scope and humility: public AfC/AfD samples skew toward the English Wikipedia and toward subjects that attempted a page. They under-represent the many companies that never submit because they already know they would not qualify. Read the rejection rates as "the risk among those who try," not "the risk in the general population of businesses."

Rejection and deletion patterns: the base rates you are actually facing

Two different gates can stop a page, and they fail for overlapping reasons.

Gate 1 — AfC (getting published at all). This is where most attempts end. Public data points to roughly two-thirds of new drafts declined, overwhelmingly for notability and sourcing rather than formatting (Lumino Digital, 2025). The encyclopedia is explicit that scarcity is the norm for organisations: "only a small percentage of the world's organizations meet the requirements for a Wikipedia article" (WP:NCORP).

Gate 2 — AfD and speedy deletion (staying published). Even a published article can be nominated for deletion. The deletion policy describes how pages "that do not meet the relevant criteria for content of the encyclopedia are identified and removed" (WP:DEL), and lists, among the reasons, "articles with subjects that fail to meet the relevant notability guidelines." Blatantly promotional pages can be removed fastest of all, via speedy deletion, "without waiting for any discussion" (WP:CSD / Wikipedia:Speedy deletion).

The table below summarises the publicly-reported pattern. Treat every percentage as a third-party benchmark, attributed in the notes, not as a WikiBusines measurement.

Notes: percentages from Lumino Digital, "Articles for Creation Approval Rates," 1,009-submission sample, Sept–Oct 2025; process facts from Wikipedia policy pages cited inline. Categories are Lumino's; your specific subject may diverge.

The practical takeaway is not "Wikipedia is impossible." It is "Wikipedia is selective, and the selection is mostly about sources." Which is good news, because sources are the one input you can assess and improve before committing budget.

A quick, honest checkpoint (soft CTA). If you have a folder of coverage and you are not sure whether it clears the bar, that is precisely what a paid Notability Audit is for — a fixed-scope read of your sources against policy, with the fee credited toward the project if you proceed. You can also self-assess for free using the Source Strength Index scorecard below before talking to anyone. Both paths beat finding out at the decline notice.

Strongest versus weakest source types

If you read enough public declines, the source hierarchy becomes obvious. Wikipedia wants independent, reliable, in-depth secondary coverage and is openly skeptical of everything the subject can manufacture. The next table classifies common source types by how much weight they typically carry — and is itself a usable tool: tally what you actually have.

Two rules cause most of the heartbreak we see:

1. Press is not independence. A glowing feature that originated from your PR team, ran as sponsored content, or simply reprinted your announcement is not independent coverage, no matter how prestigious the masthead. Reviewers read for independence first.
2. Volume is not depth. Forty trivial mentions do not add up to one piece of significant coverage. The General Notability Guideline requires sources that address the topic "directly and in detail." Three excellent independent profiles beat fifty name-drops.

A blunt diagnostic: strike out every source you control, paid for, or pitched into existence. Strike out every passing mention. What survives is your real notability case. If little survives, your SSI is low and the honest move is to wait — not to dress up the draft.

Business vs. founder vs. creator risk

Notability is assessed against type-specific standards, and notability does not transfer between a company and the person who runs it. A famous company does not make its founder notable, and a notable founder does not automatically make the company notable; each is judged on its own independent coverage. This is one of the most expensive misunderstandings we correct.

What this means in plain terms:

• Companies and startups are the high-risk lane. The available "evidence" is disproportionately marketing. The fix is not more PR; it is genuine, independent, sustained third-party attention — which often means waiting until it exists.
• Founders are judged as individuals. "I built a successful company" is a business fact, not personal notability. The personal case needs coverage about the person. If that does not exist yet, a company page (if the company qualifies) is usually the better first move — see Companies vs Founders vs Public Figures.
• Creators have a structural advantage. Independent criticism, reviews, and retrospectives are exactly the in-depth secondary coverage Wikipedia prefers, which is why cultural topics clear at higher rates.

If you are unsure which lane you are in, our notability decision tree walks the company question end to end, and the deletion failure-pattern guide shows what sinks each type after publication.

Source freshness and independence: the two factors that move scores most

Across public declines, two SSI factors do the heaviest lifting.

Independence. This is the factor reviewers test first and forgive least. A source is non-independent if the subject created it, paid for it, was interviewed for it without independent analysis, or distributed it. The encyclopedia's standard for organisations is coverage from sources "unrelated to the organization or product" (WP:NCORP). In practice, the most common reason a confident applicant gets declined is that their "best" sources are sophisticated PR — beautifully written, widely syndicated, and entirely non-independent.

Freshness — but not the kind you might expect. Freshness here is about distribution over time, not recency. A subject covered in a single launch-week burst is fragile: it reads as a campaign, and campaigns fade. A subject covered repeatedly across years, by different independent authors, demonstrates durable significance. For a deletion-resistant page, breadth across time beats a spike. This is also why pages built on one news cycle are over-represented at AfD a year later — a dynamic we map in the post-publication lifecycle risk curve.

A useful self-test for freshness and independence together:

• Plot your strongest sources on a timeline. One cluster = fragile. Spread across years = durable.
• For each, ask: did we cause this to exist? If yes for most, independence is your binding constraint, and no amount of drafting skill fixes it.
• Count distinct, independent authors/outlets writing about you in depth. If the number is below a few, you are likely in the "Not yet" band regardless of total link count.

Language-edition differences: where English is the hard mode

Approval risk is not the same across Wikipedia's language editions, and the English Wikipedia is generally the most demanding for company and founder topics. Each edition is a separate community with its own inclusion norms, sourcing expectations, and reviewer culture; notability does not automatically transfer from one edition to another.

This matters commercially in two directions:

1. English is the most-scrutinised, most-contested edition for promotional-adjacent topics — and also the highest-value for global reach and AI visibility. Higher reward, higher risk.
2. A different edition is sometimes the more honest starting point. If a subject's strongest independent coverage is in a specific national-language press, the matching language edition can be the more defensible first page — and a stronger base than a borderline English attempt. We treat this as a sequencing decision, covered in depth in Multilingual Wikipedia Strategy.

We avoid asserting precise per-edition acceptance percentages, because we have not seen a public, methodologically clean dataset that measures them across editions — and inventing one would violate this report's own rules. What we can offer instead is the directional pattern from public process and our own work: editions differ, English is hard mode for commercial subjects, and choosing the right edition is a risk decision, not an afterthought.

Indicative (our practice, not an audit): acceptance tends to be lowest on English Wikipedia for commercial subjects and noticeably higher on mid-sized Tier-2 and Tier-3 editions where the bar can be met with strong regional sources; 12-month durability is best where the source base was already solid at creation and weakest for thin, single-campaign sourcing. Treat these as directions, not measurements.

Pricing reflects this risk gradient. English Wikipedia is EUR 1,930 (~USD 2,090) for a company page and EUR 1,300 (~USD 1,410) for a personal page; Tier-1 editions (German, Dutch, Italian, Russian, Arabic, Chinese, Hindi) are EUR 1,450 / EUR 1,100; Tier-2 (Ukrainian, French, Spanish, Portuguese, Japanese, Korean, and Simple English) are EUR 1,220 / EUR 1,000; the roughly 59 Tier-3 editions are about EUR 780; and ~50 Tier-4 editions about EUR 600/550. Final pricing depends on source strength, language edition, complexity, COI sensitivity, and maintenance — see the full pricing guide and the 5-year total-cost breakdown.

AI and LLM content flags: a 2026 risk that did not exist a few years ago

A genuinely new risk has entered the picture. Wikipedia's Articles for Creation process now states plainly that drafts "generated entirely by LLMs will be rejected" (WP:AFC). Public reporting and AfC discussion through early 2026 describe reviewers increasingly flagging AI-pattern text — fabricated or hallucinated citations, generic encyclopedic filler, and sources that do not say what the draft claims they say.

Why this matters for risk:

• AI-written drafts fail two ways at once. They trip the LLM-generated rejection criterion, and they tend to contain unverifiable or invented citations, which collide head-on with verifiability — "even if you are sure something is true, it must have been published in a reliable source before you can add it" (WP:V).
• The detection bar is rising, not falling. Treat "the AI wrote a clean draft" as a liability signal in 2026, not a shortcut. A polished draft on a thin source base is more likely to attract scrutiny, not less.
• The fix is human sourcing, not better prompts. The defensible path is real source research and neutral human drafting that survives a citation-by-citation check. This is also why "I'll just have ChatGPT write it" is one of the riskier 2026 strategies, and why being on Wikipedia still shapes how AI systems describe you — a dynamic we cover in Wikipedia, Wikidata, and AI Search.

We do not present a precise "percentage of drafts flagged for AI" as a measured figure; where public sources cite such numbers they are third-party-claimed and evolving, and our position is the conservative one: AI-generated drafts are a hard fail, and we do not submit them. What we can share is the directional trend we see at intake.

Indicative (our intake, not an audit): in 2026 a growing share of incoming briefs — very roughly a quarter to a third — arrive partly AI-drafted and need a full human rebuild before they are submittable. We rebuild them properly or decline; we never submit AI-generated text to Wikipedia.

Your decision tool: the 5-minute risk self-check

Before you contact any provider, run this checklist against your own source folder. It operationalises the Source Strength Index and requires nothing but honesty.

Step 1 — Inventory. List every source that mentions your subject. Link, outlet, date, and one line on what it actually says.

Step 2 — Strike the non-independent. Delete press releases, sponsored/"contributor" posts, paid placements, your own site/blog/social, and pure interviews with no independent analysis. (WP:NCORP: coverage must be from sources "unrelated to the organization or product.")

Step 3 — Strike the trivial. Delete passing mentions, directory listings, funding-round line items, and "also attended" name-drops. Keep only coverage that is about your subject, directly and in detail.

Step 4 — Count what survives.

• 0–1 strong independent sources → SSI almost certainly 0–44 (Not yet). Do not submit; build coverage.
• 2–3 strong independent sources → likely 45–69 (Fixable). An audit can tell you whether it is enough.
• 4+ strong, in-depth, independent sources, spread over time → potentially 70–100 (Defensible). A real project is plausible.

Step 5 — Apply the topic-type penalty. If you are a company or (especially) a startup, demand more from the surviving list than a creator would — the bar is higher and reviewers are more skeptical.

Step 6 — Check the timeline. All your sources from one week? Treat the result as one band weaker. Spread across years? You are more deletion-resistant.

If you finish this and land in "Not yet," that is not a failure of the checklist — it is the checklist doing its job and saving you a public decline plus the cost of a doomed attempt. For the formal version, our paid Notability Audit does this against full policy with a written verdict, and the fee is credited toward the project if you proceed. To choose any provider wisely, our honest vendor scorecard shows which questions separate compliant operators from the "100% approval" crowd.

What we will NOT promise, and why

We do not promise Wikipedia approval. No one honestly can. Notability is "presumed," not guaranteed — the General Notability Guideline is explicit that significant coverage "creates an assumption, not a guarantee, that a subject merits its own article." Articles are kept or removed by community consensus through public processes like AfD, which run "for at least seven days, after which a decision may be reached based on community consensus." We will not claim a 100% approval rate, sell "no deletion" guarantees, or imply any special relationship with editors or administrators — those claims are red flags, not features. What we will do is reduce risk before you spend: assess notability against policy, research and grade your sources, draft neutrally, disclose paid editing per the Wikimedia Terms of Use, and monitor after publication. And if a published page cannot be defended after 3 attempts within the 90-day monitoring window, our published guarantee is an 80% refund — a promise about our accountability, never about the community's verdict.

Our compliance posture is the same one we ask you to demand of any vendor. Paid editors must disclose: the current Wikimedia Foundation Terms of Use require that "you must disclose each and any employer, client, intended beneficiary and affiliation with respect to any contribution for which you receive, or expect to receive, compensation" (Wikimedia Foundation Terms of Use, §4). Conflict-of-interest editors are, in Wikipedia's words, "strongly discouraged from editing affected articles directly, and can propose changes on article talk pages instead" (WP:COI). We disclose, we use talk pages and Articles for Creation rather than quietly editing live articles, and we will tell you to wait when the sources say wait. The full mechanics are in our paid-editing and COI compliance guide.

Pricing and what risk reduction actually costs

Risk reduction has a price, and transparency about it is part of being a trustworthy provider. Here is where the money goes, in EUR with approximate USD.

Final pricing depends on source strength, language edition, complexity, COI sensitivity, and maintenance. The cheapest possible path — doing it yourself — is EUR 0 in cash but carries the full ~68% decline risk and the time cost of months in the queue. The most expensive path is a cheap, non-compliant vendor whose page gets speedily deleted, taking your money and your draft with it. For the complete economics, including five-year total cost of ownership, see How Much Does a Wikipedia Page Cost. The full menu is on our services page.

The complete 2026 Wikipedia playbook

This guide is one part of a ten-part series — an honest, end-to-end walkthrough of getting and keeping a Wikipedia page in 2026. Each part stands alone; together they cover the whole journey.

Before you start — Can my company get a page? · Company vs founder vs public figure Budget & vendor — What it costs — 5-year TCO · The honest vendor scorecard Compliance & risk — Paid editing, COI & disclosure · Why pages get deleted — 12 patterns Strategy & growth — Wikipedia, Wikidata & AI search · Multilingual strategy After publication — Monitoring & the lifecycle risk curve The data — Wikipedia Risk Report 2026 (you are here)

Not sure where your case stands? A fixed-scope Notability Audit reads your real sources against policy — or just talk to the team.