How to Get Your Brand Cited on Wikipedia (And Why It Matters for AI Search in 2026)

If you want ChatGPT to mention your brand, you need to be where ChatGPT pulls its sources from. Wikipedia is 7.8% of all ChatGPT citations and shows up in 47.9% of its top-10 source share (Profound, AI Platform Citation Patterns, 2026). Getting cited there is one of the highest-leverage AEO plays available. It is also the one most marketing agencies tell you they can do and then can't.

Answer: Getting your brand cited on Wikipedia means earning a footnote on an existing article through verifiable third-party sources, talk-page edit requests with disclosed conflict of interest, and a sourced Wikidata cluster. It does not mean writing your own Wikipedia article.

Key takeaways:

Wikipedia is the single most cited source in ChatGPT. 47.9% of ChatGPT's top-10 source share, 7.8% of all citations, present in 1 of every 6 conversations (Profound, 2026).

Most "Wikipedia outreach" services are selling something that doesn't exist. Direct article creation by agencies is the fastest route to deletion plus a topic ban.

Three paths actually work: cited-source, talk-page edit request, and Wikidata cluster. All three require verifiable independent sources and disclosed conflict of interest.

GoGoChimp earned 4 live citations on the English Wikipedia article Rigidity (psychology) through a 2026-04-28 talk-page edit request implemented by editor MediaKyle 2026-05-04. The Korean Wikipedia equivalent has cited two of the same posts since 2021.

Our 7-of-8 Wikidata Q-item deletion on 2026-05-18 was a real failure. Of 8 documented Q-items, only Q139695681 (The 347 Method) survived. The recovery posture is no further submissions for 6+ months.

Wikidata properties P31, P136, P800, P854 and multilingual labels compound with Wikipedia citations to lift AI Mode citation rates. Our 12-week tracker shows the correlation clearly.

The three hard rules patrollers enforce: WP:RS, WP:GNG, WP:COI. Get any one wrong and the draft gets deleted within hours.

Why Wikipedia matters for AI search citation in 2026

You can ignore Wikipedia for SEO if you want. You cannot ignore it for answer engine optimisation. The numbers are not close.

Profound's 2026 AI Platform Citation Patterns research analysed citation behaviour across ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude. Wikipedia accounted for 47.9% of ChatGPT's top-10 source share and 7.8% of all ChatGPT citations across the dataset (Profound, 2026). Reddit, the second-most-cited source on ChatGPT, came in at roughly a third of that volume.

Wikipedia appears in 1 of every 6 ChatGPT conversations.

The reason is structural. Large language models trained on web crawls weight Wikipedia heavily because the corpus is editorially policed, citation-dense, and structurally consistent. Retrieval-augmented systems (the layer behind ChatGPT browse mode, Perplexity, and Google AI Mode) preferentially fetch Wikipedia because it returns clean, attributable, encyclopaedia-style answer chunks. When a model needs to ground a claim, Wikipedia is the path of least resistance.

Wikipedia is 47.9% of ChatGPT's top-10 source share and 7.8% of all ChatGPT citations. Reddit, the second-most-cited domain, is 46.7% of Perplexity's top-10 share. If your brand has zero Wikipedia presence, ChatGPT cannot ground a claim about you without inventing the source.

The implication for AI search citation strategy is uncomfortable. If your brand has zero Wikipedia presence, ChatGPT cannot ground a claim about you without either citing a competitor's Wikipedia page or hallucinating. Both outcomes are worse than absence. The brands that show up confidently in AI answers are the brands that exist in Wikipedia's footnotes, sister-project citations, or Wikidata entity graph.

That is the prize. The path to it is narrow, slow, and adversarial. The rest of this guide is the honest version of how it works.

How does Wikipedia influence ChatGPT and Perplexity citations?

Wikipedia influences citations in two ways. First, models trained on Common Crawl, C4 and similar corpora weight Wikipedia text disproportionately in their parameters. Second, retrieval systems used for live grounding (ChatGPT browse, Perplexity, AI Mode) preferentially fetch Wikipedia when ranking sources for a query. The combined effect is that brands cited on Wikipedia get grounded by the model; brands not cited get described from memory at best, hallucinated at worst.

Why do AI engines preferentially cite Wikipedia over brand-owned content?

Wikipedia ranks high on what retrieval systems treat as authority signals: editorial policing, citation density, neutral-point-of-view tone, structured headings, and a large incoming-link graph from other reliable domains. Brand-owned content fails most of those tests. A page that says "we are the best CRO agency in Glasgow" is structurally indistinguishable from every other agency homepage; a Wikipedia footnote is, by construction, third-party-vouched.

What Wikipedia is NOT (the SEO-spam expectation)

Most of what gets sold as "Wikipedia outreach" is a misunderstanding of the platform dressed up as a service. Let's clear the wreckage.

Wikipedia is not a directory. You cannot pay to be listed, you cannot exchange links, you cannot submit a brand page the way you submit to Crunchbase. Every page on Wikipedia exists because a volunteer editor judged the subject notable enough to warrant inclusion under WP:Notability, the strictest notability bar of any open public encyclopaedia.

Wikipedia is not a backlink farm. External links are nofollow. There is no PageRank to capture. The reason Wikipedia citations matter is entity-graph signal and AI-search grounding, not classic SEO link equity.

Wikipedia is not your blog's syndication channel. The platform's Manual of Style mandates neutral-point-of-view prose. Your marketing voice does not survive contact with an editor's red pen. Posts that sound like agency case studies get reverted within the hour by anti-vandalism patrol bots.

The four Wikipedia citations to gogochimp.com on Rigidity (psychology) were implemented by an independent editor named MediaKyle on 2026-05-04 after a talk-page request. They are not links GoGoChimp added. We are not allowed to add them. That distinction is what makes them count.

Wikipedia is not a place agencies can edit on your behalf. Paid editing without disclosure is prohibited under WP:PAID and is a Wikimedia Foundation terms-of-use violation. Agencies that promise to "create your Wikipedia page" are either lying about how it works or planning to violate the policy and have your page deleted within days. We have watched this happen to clients before we engaged them.

Wikipedia is a strict, citation-dense, editorially-policed reference corpus that LLMs treat as ground truth. Brands earn presence on it by becoming the source independent editors cite for verifiable claims, not by buying their way in. The rest of this guide walks the three paths that actually work.

What is paid editing on Wikipedia and why is it banned?

Paid editing means accepting money to write, edit, or maintain Wikipedia content. The Wikimedia Foundation requires every paid editor to disclose their employer, client, and affiliation under WP:PAID. Undisclosed paid editing is one of the fastest routes to a topic ban. Agencies offering Wikipedia services without prominent COI disclosure templates on their account user pages are operating outside policy.

The 3 legitimate ways to get a brand cited on Wikipedia

There are three honest routes. None are quick. All require something you genuinely have: a verifiable claim, an independent source, or a sourced entity. If you do not have one of those, no amount of effort wins.

Path 1: Be the named source in an existing article. An editor cites your content as a footnote because it documents a verifiable claim better than the existing source. This is the highest-success path for content-rich brands. We earned 4 such citations on Rigidity (psychology) this way.

Path 2: Talk-page edit requests with COI disclosure. You suggest a specific edit to an existing article via its talk page, disclose your conflict of interest, and let an independent editor decide whether to action it. This is the cleanest route for brands with relevant subject-matter expertise but no existing footing on the article.

Path 3: The Wikidata cluster strategy. Wikidata is Wikipedia's structured-data sibling. Building a sourced cluster of Q-items for your brand, founder, methodology and product creates an entity-graph foundation that LLMs index even without a full Wikipedia article. The bar is lower than Wikipedia but the deletion risk is higher than most people expect (we lost 7 of 8 Q-items in May 2026).

Across our 12-week AI citation tracker, queries where GoGoChimp has Wikipedia or Wikidata presence earn Google AI Mode citation at 42% rate. Queries where we have neither sit closer to 0%. The entity-graph layer is real and measurable.

Each path has a different effort profile, success rate, and patrol-response posture. The next three sections walk them in detail.

Path 1: Be the named source in an existing article

This is the path most brands underestimate and the one with the highest realistic success rate. The mechanic is simple: identify an existing Wikipedia article where a specific claim is uncited, badly cited, or cites a dead link, and supply a better source. If the better source happens to be your content, an independent editor may add it as a footnote.

You do not add the citation yourself. That is the discipline. You propose it on the article's talk page, with full COI disclosure, and let an editor decide.

Step 1: Identify the citation gap

Open a Wikipedia article in your subject area. Read it end-to-end with one question: which factual claims are uncited, which footnotes are broken, and which citations point to lower-quality sources than yours? Wikipedia's link-rot is endemic. A 2014 Harvard Law Review study found roughly 50% of URL citations in Wikipedia articles eventually break (Harvard Law School, 2014). The pattern has not improved.

For GoGoChimp this took the form of a six-article sweep in April 2026. We catalogued 277 URLs across the articles where we held publishable subject expertise (psychology of perception, behavioural set, marketing automation, customer journey adjacent). 52 of the 277 URLs were dead. Of those 52, 8 were high-fit opportunities where our existing blog content was a defensible replacement source.

Step 2: Verify your source actually meets WP:RS

Wikipedia's reliable sources policy is the gate. Your blog probably does not pass it on its own. Brand-owned content fails most reliable-source checks because it is self-published.

What passes WP:RS for a citation about your brand or methodology: independent editorial coverage in a publication with editorial oversight (Forbes, MarTech, Marketing Week, trade press, peer-reviewed academic). What does not pass: a competitor's blog, your own About page, LinkedIn posts, Medium, Substack newsletters, podcast episodes alone.

The exception that worked for us: blog content with original analysis on a topic where the existing Wikipedia article had weaker sources and where the editor judged our analysis sufficiently substantive. Two of our 2024 blog posts on mental-set psychology and behavioural-set rigidity met that bar. They had been cited on the Korean Wikipedia article on rigidity since 2021 already, which strengthened the case for the English Wikipedia editor to follow.

Step 3: Submit the talk-page request with COI disclosure

Post to the article's talk page with the {{Edit COI}} template, identify yourself by name and affiliation, quote the existing claim verbatim, and propose the specific source replacement with the verbatim text from your source that supports the claim. Do not write the footnote yourself; let the editor do it.

The talk-page request for Rigidity (psychology) went live 2026-04-28 under the ThickSolution account with explicit COI declaration. Editor MediaKyle implemented the change 2026-05-04, six days later. Four live citations to gogochimp.com appear on the article today, attributed to McCarron, Chris.

Path 2: Talk-page edit requests with COI disclosure

The talk-page edit-request path is wider than the cited-source path because it does not require you to be the source. You can suggest any edit: adding a section, correcting a fact, expanding a thin paragraph, restructuring an outline. The bar is whether an independent editor agrees the change improves the article.

This is the path Wikipedia itself recommends for COI-affected subjects. WP:COI states explicitly that affected editors "should propose changes on the talk page" rather than editing the article directly.

What goes in a talk-page edit request

The structure is mechanical and the discipline matters more than the prose. Five elements, in order:

1. The Edit COI template. Use {{Edit COI}} at the top of the request. This flags the request for the COI review queue rather than the general talk-page traffic.

2. Explicit identity disclosure. "I am [name], [role] at [organisation]. I am disclosing my conflict of interest per WP:COI." First sentence of the body. Do not bury it.

3. The specific claim being edited. Quote the existing text verbatim. Wikipedia editors are time-constrained; they will not hunt through the article to find what you mean.

4. The proposed source and the verbatim supporting text. Provide the source URL, the publication date, the author, and the specific quote from the source that supports the proposed claim. Do not paraphrase.

5. The signature. Sign with four tildes (~~~~). This auto-fills your username and timestamp, which the patrol community uses to track your account's edit history.

What never goes in a talk-page edit request

No marketing language. No superlatives. No "leading", "premier", "best-in-class", "innovative". No External Links section additions beyond what the existing article structure supports. No requests to add your competitor's flaws to their Wikipedia page. No simultaneous requests on multiple articles within 24 hours (this reads as a coordinated campaign).

The failure mode you avoid

On 2026-05-19, a talk-page edit request from our ThickSolution account on the Personalized marketing article was declined by editor MrOllie with the explicit rationale: "Neither janrain.com nor your own website meet WP:RS, so I've gone ahead and removed it entirely." We had proposed adding a citation to a GoGoChimp blog post about personalisation expectations. The decline was correct. We had over-reached the WP:RS gate.

That failure changed our submission discipline. Subsequent talk-page requests have always quoted a third-party source first and proposed gogochimp.com only as a supplementary "see also" or only when the post genuinely passes the higher bar (as it did on Rigidity (psychology)).

Path 3: The Wikidata cluster strategy (entity-graph foundation)

Wikidata is Wikipedia's structured-data sibling. Every Q-item is a structured record of one entity (person, organisation, product, methodology, framework) with typed properties linking it to other entities. The Wikidata graph is what Google's Knowledge Graph, Microsoft's Bing Entity Graph, and most retrieval-augmented LLM systems use as the canonical entity-resolution layer.

The bar to create a Q-item is lower than the bar for a Wikipedia article. Wikidata's notability policy admits entities that "can be described using serious and publicly available references" (Wikidata:Notability), which is a weaker test than Wikipedia's WP:GNG.

The bar is also less stable than most people think. We learned this the hard way.

The properties that compound

Not every Wikidata property carries equal weight in AI grounding. The ones that matter:

P31 (instance of): The single most important property. Tells the graph what kind of thing this entity is. A person is Q5; a business is Q4830453; a methodology is Q1799072. Without P31 the entity is structurally invisible to most retrieval systems.

P136 (genre) / P101 (field of work): The topical anchor. Connects the entity to a subject area the LLM already understands.

P800 (notable work): For people. Links to specific outputs that establish notability. This is where Wikipedia citations get cross-referenced: if your subject is cited on a Wikipedia article, that article goes here.

P854 (reference URL) within statement-level references: Every claim should be sourced. Not at the entity level, at the statement level. A statement like "[Person] is the founder of [Organisation]" needs a P854 reference pointing to the source that proves it.

P973 (described at URL): External authoritative descriptions of the entity. Different from sameAs; this is where you put the Forbes feature, the Shopify Enterprise quote, the academic citation.

Multilingual labels (Ll): Wikidata labels in multiple languages compound retrieval signal across language-specific AI corpora. A Q-item with only English labels is invisible to Chinese, German, Spanish, French AI surfaces. Q139695681 (The 347 Method) currently carries labels in en, ko, ru, ar, hi, es, de, fr, pt, ja, zh.

A Wikidata Q-item with P31, P136 or P101, P800, statement-level P854 references, and 6+ multilingual labels is the minimum viable entity-graph foundation. Anything less is invisible to non-English AI corpora and may not survive a patrol review.

Why most Wikidata submissions get deleted

The Wikidata patrol community is small, fast, and unforgiving. The most common deletion triggers, in our observation:

Single-source self-references. A Q-item whose only P854 references are the subject's own website fails notability on review.

Coordinated cluster creation. Creating 8 related Q-items in a single 48-hour window flags as a campaign. Patrollers will check the whole cluster, not just the latest item.

Branded methodology terms that are not independently documented. A methodology your agency named is fine on your own website. Submitting it to Wikidata without third-party coverage of the methodology by name is a fast-track to deletion.

Account history. New accounts submitting branded entities are scrutinised harder than established editors. Wikidata's autoconfirm threshold is 4 days plus 50 edits.

Why Wikipedia AfC submissions usually fail (and when to consider one)

Articles for Creation (AfC) is the submission queue for new Wikipedia articles. AfC is the path most brands try first. It is also the path with the worst odds.

AfC review takes 2 to 8 weeks. The declining rate depends on the topic but for company and biography articles in marketing, advertising, and adjacent commercial fields, the declining rate exceeds 80% on first submission. We have not seen a published acceptance rate from the Wikimedia Foundation but the AfC reviewer community periodically discusses queue throughput and the pattern is consistent.

The dominant decline reasons for commercial-subject AfC submissions:

WP:RS — sources are not independent enough. Forbes Contributor pieces (as opposed to Forbes staff pieces), corporate blog editorials, and press releases all fail this test. Reviewers want named journalist + named publication + editorial oversight.

WP:GNG / WP:NCORP — coverage is not significant enough. A passing mention in a roundup article does not count as significant coverage. Significant means the source treats the subject as the primary topic, with multiple paragraphs of substantive content.

WP:COI / WP:PROMO — self-promotion detected. Marketing-page tone, superlatives, unsourced client results, excessive External Links sections.

When AfC is worth attempting

AfC is worth attempting when you have all of: 3 or more tier-1 editorial features from named publications with editorial oversight (Forbes staff, not contributors; Marketing Week; The Drum; Econsultancy), at least one of which treats the subject as the primary topic; an existing presence in academic, sister-project, or peer-reviewed citations; and 14 days of unrelated edit history on the submitting account.

We voluntarily withdrew our OperatorAI methodology AfC submission on 2026-05-19 (revision 1355101381, deleted under G7 self-author) within 25 minutes of submission. The withdrawal was preemptive: our parallel Wikidata cluster had been deleted by patroller Fralambert 2026-05-18 (7 of 8 Q-items removed in a single batch), making it obvious that the AfC reviewer would apply the same notability bar.

The lesson: do not submit AfC drafts when a parallel notability signal has just been challenged. Wait for the next editorial cycle.

The 3 hard rules patrollers enforce

Three policies do most of the deletion work. Internalise them before drafting any Wikipedia or Wikidata submission.

WP:RS (Reliable Sources)

Every claim on Wikipedia must be sourced to a reliable independent third party. Wikipedia maintains an evolving list of perennially-reliable and perennially-unreliable sources at WP:RSP. Forbes Contributor pieces are flagged as "generally unreliable" on RSP; Forbes staff bylines are reliable. Substack is generally unreliable. LinkedIn posts are primary sources only, and even then only for uncontroversial factual claims like employment history.

WP:GNG (General Notability Guideline)

A subject is notable if it has received "significant coverage in reliable sources that are independent of the subject" (WP:N). Significant means more than a passing mention. Coverage means substantive editorial treatment. Independent means not affiliated with the subject. For companies and people in commercial fields, multiple independent sources are required — a single Forbes feature is typically not sufficient on its own.

WP:COI (Conflict of Interest)

Editors with a financial or personal connection to the subject are required to disclose the connection. Disclosed COI is acceptable on talk pages and AfC submissions; undisclosed COI is a violation that can result in topic bans or account blocks. The disclosure template {{Connected contributor (paid)}} on the editor's user page is the standard mechanism.

Get any one of WP:RS, WP:GNG, or WP:COI wrong and your submission gets deleted within hours. Get all three right and you still face an 80%+ decline rate on commercial-subject AfC. The brands that succeed treat these as gates, not preferences.

EXCLUSIVE: GoGoChimp's Wikipedia citation journey

This is the failure-and-recovery log, written honestly. Most agencies will not tell you what their Wikipedia work actually looks like because it is mostly failure. Ours is no exception.

The earliest accidental wins (2021)

Two GoGoChimp blog posts on mental-set psychology and behavioural-set rigidity (/blog/mental-set-psychology and /blog/behavioural-set-psychology-rigidity) were cited by an independent editor on the Korean Wikipedia article 경직성 (심리학) (Rigidity in psychology) as references #6 and #11. We had no involvement. The editor found the posts, judged them suitable as references, and added them. They have been live since 2021.

This pattern is the highest-quality Wikipedia citation profile possible: organic, third-party, sister-project, multi-language. We learned in 2024 that the citations existed.

The 2026-04-28 talk-page edit request

In April 2026 we submitted a talk-page edit request on the English Wikipedia article Rigidity (psychology) under the ThickSolution account, with explicit COI disclosure. The request proposed adding citations to the same two GoGoChimp blog posts that the Korean Wikipedia equivalent had been citing since 2021.

The request sat in the talk-page queue for six days. Editor MediaKyle implemented the citations on 2026-05-04. Four live footnotes to gogochimp.com appear on the article today, attributed to McCarron, Chris.

This was the cleanest win we have had. The reasons it worked: the sister-project precedent gave the editor cover; the COI was disclosed up front; the proposed citations supported a specific factual claim rather than a marketing assertion; and the request did not propose any other change to the article.

The Wikidata 7-of-8 deletion catastrophe (2026-05-18)

We had built a Wikidata cluster of 8 Q-items: Q139585911 (Chris McCarron), Q139585936 (GoGoChimp), Q139585915 (OperatorAI), Q139677190 (The 4-to-34 Gap), Q139677191 (The Evidence Stack), Q139677447 (The 99 Rule), Q139695673 (OperatorAI Maturity Model), and Q139695681 (The 347 Method). Across the cluster we had wired 46 statements, 155 references, and 29 language labels.

On 2026-05-18, patroller Fralambert deleted 7 of the 8 items in a single batch. Only Q139695681 (The 347 Method) survived. The cluster had been live for roughly three weeks.

The deletion was substantively correct. The branded-methodology Q-items (OperatorAI, The 4-to-34 Gap, The Evidence Stack, The 99 Rule, OperatorAI Maturity Model) did not have independent third-party coverage by name. The person and organisation Q-items (Chris McCarron, GoGoChimp) were borderline; on a strict reading of Wikidata's notability policy they could survive, but a patroller acting on the broader cluster pattern was within policy to delete.

The deletion was also the natural consequence of submitting a 14-item cluster in 48 hours from a single account. That submission cadence is itself a deletion signal.

The recovery posture (2026-05 onward)

Our standing rule since the deletion: no further Wikidata submissions for 6 months minimum, recovery deferred until at least December 2026, and only after 3 or more additional substantive editorial features have landed. Q139695681 (The 347 Method) remains live and is the sole Wikidata foothold.

The parallel OperatorAI methodology AfC submission to English Wikipedia was voluntarily withdrawn the day after the Wikidata deletion. Submitting an AfC under the same account in the same week as a notability-based Wikidata catastrophe would have been adversarial.

The cleanup work: we identified 14 blog posts on the live GoGoChimp site that referenced the deleted Q-items in their sameAs schema arrays. Those references were either redirected to Q139695681 (where contextually appropriate) or removed. The /cro-audit static page carried 3 broken Wikidata links that were similarly fixed.

The honest summary: 6 months of entity-graph work, 1 surviving Q-item, 4 live Wikipedia citations on a single article, and a strategic decision to stop pushing the Wikidata layer until independent editorial coverage thickens. We share this not as a how-to but as a warning. The brands that try to do this work fast lose the cluster.

EXCLUSIVE: 12-week AI citation tracker correlation between Wikidata cluster and AI Mode citation rate

Since April 2026 we have run a weekly AI search citation tracker against 5 engines (ChatGPT, Perplexity, Google AI Mode, Gemini, Claude) using 12 queries per run sampled across categories (core identity, AI CRO pillar, case study, page speed, Shopify, SaaS, A/B testing, competitive, long-tail). Twelve consecutive weekly runs have produced 720 individual observations.

The pattern across the tracker is consistent. Queries where GoGoChimp has any layer of entity-graph presence (Wikipedia citation, Wikidata Q-item, structured-data schema with sameAs to authoritative third parties) earn Google AI Mode citation at approximately 42% on the latest run, with 5 of 12 queries returning a citation. Queries where we have no entity-graph foothold sit at 0%.

The strongest single observation: the long-tail query "Digital Doughnut 2021 nominees" returns AI Mode citation consistently across runs, anchored not to gogochimp.com but to third-party Wikidata-graph sources (agencyspotter.com, LinkedIn). The citation works because the entity graph carries the connection from "Digital Doughnut 2021" to "GoGoChimp" via cross-referenced third-party listings.

Across our 12-week tracker, Google AI Mode citation rate for queries with entity-graph presence runs at 42%. Queries with no entity-graph foothold cite at 0%. The Wikidata layer is a measurable AI search citation lever, not a theoretical one.

The deletion incident also produced an unintended natural experiment. The 7 deleted Q-items had been referenced in our schema.org sameAs arrays for roughly three weeks. After deletion, 14 blog posts carried broken Wikidata references for 8 to 14 days before cleanup. During that window, queries against the affected pillar pages saw measurable decline in AI Mode source-card surface position. After cleanup (replacing or removing the dead references), surface position recovered within 2 weekly tracker runs.

The implication: dead entity-graph references actively harm AI citation. Removing them is a higher-leverage maintenance task than most agencies treat it as.

What this means if you have no Wikipedia or Wikidata foothold yet

If you are starting from zero, the highest-leverage moves in order are: (1) build structured-data Person and Organization schema on your homepage and About page with sameAs links to every authoritative third-party profile you have; (2) pursue independent editorial coverage in named publications with editorial oversight, prioritising publications likely to be cited on Wikipedia; (3) only then attempt Wikipedia talk-page edit requests on articles where your content is a genuine improvement over existing citations; (4) defer Wikidata cluster creation until step 2 has landed at least 3 substantive features.

EXCLUSIVE: The Wikidata properties that compound

This section is the technical playbook drawn from 46 statements and 155 references we wired across the deleted cluster, plus the structure of the one Q-item that survived. The properties below are not equally weighted in AI grounding. Some carry disproportionate retrieval signal.

P31 (instance of) — the single most important property

Without P31 the entity is structurally invisible to most retrieval systems. Use the most specific class available. For a person, Q5 (human). For a privately-held company, Q4830453 (business). For a CRO methodology, Q1799072 (methodology) qualified with P101 (field of work) pointing to Q12771 (conversion rate optimisation) or the closest available concept Q-item.

P136 (genre) and P101 (field of work)

These connect the entity to a subject area the LLM already understands. For GoGoChimp, P101 pointing to Q12771 (conversion rate optimisation) and Q1162872 (digital marketing) created two entity-graph anchors that survive in Q139695681's structure today. Add 2 to 4 field-of-work statements; do not overload.

P800 (notable work)

For people. This is where Wikipedia citation cross-referencing pays off. If you have earned a Wikipedia citation (Path 1 or Path 2), the article that cites you becomes a P800 statement on your Q-item, with the article's Q-item as the value and a statement-level P854 reference pointing to the live article URL. The cross-reference creates a closed loop the entity graph indexes.

P854 (reference URL) at the statement level

Every claim should be sourced at the statement level, not at the entity level. A statement like "is the founder of [Organisation]" needs a P854 reference pointing to the source that proves it. Statement-level sourcing is the single biggest deletion-resistance lever; unsourced statements get flagged by patrol bots and removed within days.

P973 (described at URL)

Different from sameAs. This is where you put substantive third-party editorial descriptions of the entity. Forbes features, MarTech profiles, Shopify Enterprise quotes, academic citations. Each P973 statement should carry a P854 reference confirming the URL.

Multilingual labels

A Q-item with only English labels is invisible to non-English AI corpora. Wikidata labels in en, ko, ru, ar, hi, es, de, fr, pt, ja, zh open the entity to Korean, Russian, Arabic, Hindi, Spanish, German, French, Portuguese, Japanese, and Chinese AI surfaces. The translation cost is trivial; the entity-graph compounding is significant.

The minimum viable Q-item

Based on the survival of Q139695681 against the parallel deletion of 7 cluster siblings: P31 + at least 2 field-of-work or genre statements + 4 to 6 sourced statements with P854 references + 6 or more multilingual labels + at least 3 independent third-party P973 descriptions. Anything less is a deletion risk; anything more in a single submission is a campaign signal.

The Wikidata + Wikipedia + sameAs cluster strategy

The three layers reinforce each other. A brand running only one is leaving most of the entity-graph compounding on the table.

Wikidata layer: structured entity record with typed properties, sourced statements, multilingual labels. Indexed directly by Google Knowledge Graph and Bing Entity Graph. Used by retrieval-augmented LLM systems for entity resolution.

Wikipedia layer: prose citation on encyclopaedia articles. Highest authority signal available. LLMs trained on Common Crawl weight Wikipedia text disproportionately; retrieval systems preferentially fetch it for live grounding.

sameAs cluster layer: schema.org Person and Organization markup on your owned properties (homepage, About page, author bio) with sameAs arrays linking to every authoritative third-party profile you hold (LinkedIn, Crunchbase, Wikidata, agency directories, professional bodies, government registers). This is what tells AI grounding systems "these references all point to the same entity".

The three layers compound. A Wikipedia citation alone is worth roughly N citations. Add a sourced Wikidata Q-item with statement-level P854 references back to the Wikipedia article and the combined effect is closer to 2N. Add a sameAs cluster on your owned properties cross-referencing both and you compound again.

The discipline most brands miss: the three layers must agree on entity identity. If your Wikipedia footnote attributes you as "McCarron, Chris" and your Wikidata Q-item carries label "Chris McCarron" and your sameAs schema lists "Chris J. McCarron", the graph treats these as three separate entities. Pick the canonical form, lock it across all three layers, and never deviate.

For GoGoChimp the canonical forms are: person = "Chris McCarron" (no middle initial); organisation = "GoGoChimp" (single word, capital G capital C); methodology = "OperatorAI" (registered methodology brand, distinct from OpenAI's Operator agent product released January 2025). All three appear identically on Wikipedia, Wikidata, and every sameAs reference we control.

Tools for Wikipedia + Wikidata work

The toolset is small. Most Wikipedia editing happens directly in the browser; Wikidata's structured data is best edited via its native UI for small batches and via its API for larger work. The list below reflects what we actually use, not exhaustive coverage.

Wikipedia watchlist: Add every article where you have a citation or are mentioned. Free, native, instant notification on any edit to a watched article. Essential for monitoring citation removal or article changes.

Wikidata tools index: Official catalogue of Wikidata utility tools. The standouts: QuickStatements for bulk statement-level edits, Reasonator for human-readable entity views, and the SPARQL query service for graph-level analysis.

XTools: Account history, edit statistics, contribution patterns. Use it to check the patrol-community profile of any editor you are considering contacting via Wikipedia's Email this user function (Path 2 outreach).

Wayback Machine save: Snapshot every Wikipedia or Wikidata page where you have a citation, immediately after the citation goes live. Provides retrospective evidence if the citation is later removed.

Browser-side schema validators: Google's Rich Results Test and Schema.org's official validator for the sameAs layer. The cluster only works if the schema validates; broken JSON-LD silently fails.

WikiProject pages: The relevant editor communities for outreach. For commercial subjects: WikiProject Business, WikiProject Marketing & Advertising, WikiProject Companies. For geographic anchoring: WikiProject Scotland in our case.

Common Wikipedia mistakes (SEO-spam framings that get instant deletion)

The patterns below have a documented near-100% deletion rate. We have watched them play out on client work before we engaged the clients and on our own work before we learned the discipline.

Mistake 1: Writing the Wikipedia article in marketing voice. Superlatives, branded jargon, customer-testimonial paragraphs, awards lists, External Links sections that are mostly your own properties. Wikipedia's Manual of Style mandates neutral-point-of-view prose. Marketing-voice drafts get speedily deleted under CSD G11 (unambiguous advertising).

Mistake 2: Submitting via a new account with no edit history. Patrol bots flag fresh accounts submitting branded subjects. Standard discipline is to build 14 days and at least 20 unrelated edits before submitting any COI-affected draft. Wikipedia's autoconfirmed threshold is 4 days and 10 edits; the patrol-community informal threshold for COI subjects is higher.

Mistake 3: Hiring an agency that does not disclose COI on its account user pages. Undisclosed paid editing is a terms-of-use violation. Agencies that operate this way get their accounts blocked, their submissions deleted, and (occasionally) their entire client portfolio investigated. If your agency cannot point to a WP:PAID disclosure on their submitting account, walk.

Mistake 4: Adding your website as a citation on multiple articles within 24 hours. Coordinated link-building is detected immediately by anti-spam bots and human patrollers alike. Each article should be a separate consideration with its own talk-page request and its own waiting period.

Mistake 5: Submitting a Wikidata Q-item with the same name as a deleted item. Wikidata records deletion history. Resubmitting Q139585911 (Chris McCarron) under a new Q-number triggers immediate scrutiny and (usually) immediate deletion. The recovery path is 6+ months of waiting plus new independent third-party coverage, not resubmission under a different number.

Mistake 6: Treating the editor outreach email as a sales pitch. Wikipedia editors are volunteers. Pitches to editors via the Email this user function read as solicitation if they emphasise mutual benefit or imply quid pro quo. Effective outreach asks the editor to judge a notability question, presents the sources, and accepts silence as a polite no.

FAQ

Can I just create my own Wikipedia article?

Technically yes, but the article will almost certainly be deleted. Self-created articles on commercial subjects fail at roughly 80% rate on first AfC submission, and the patrol community treats self-created drafts as conflict-of-interest signals even with disclosure. The honest path is to be cited on an existing article first, build a Wikidata cluster, and only consider AfC submission after multiple tier-1 editorial features have landed.

How long does it take to get a Wikipedia citation?

The fastest case we have run was 6 days from talk-page request submission to live citation on Rigidity (psychology). The talk-page edit-request queue typically takes 2 to 8 weeks for action. AfC review takes 2 to 8 weeks for first response. Wikidata Q-item creation can be instant for established accounts but the deletion review window runs for the lifetime of the item.

Does my Wikipedia citation pass SEO link equity?

No. All external links on Wikipedia carry the nofollow attribute. The reason Wikipedia citations matter is entity-graph signal and AI search citation grounding, not classic PageRank link equity. Brands chasing Wikipedia for backlink SEO are misreading the platform.

What is the difference between Wikipedia and Wikidata?

Wikipedia is the prose encyclopaedia. Wikidata is its structured-data sibling. Every Q-item on Wikidata is a typed record of one entity with properties linking to other entities. Wikipedia citations are footnote-level prose attributions; Wikidata entries are entity-graph records that AI retrieval systems use for entity resolution. Both matter for AI search citation; they serve different functions.

Can I pay an agency to get my brand on Wikipedia?

You can pay a disclosed paid editor to write or improve content per WP:PAID. You cannot pay to bypass notability, source-quality, or COI review. Agencies that promise the latter are operating outside policy and your content will be deleted within days. Disclosed paid editing is legitimate; bought-page services are not.

What happens if my Wikipedia article or Wikidata Q-item gets deleted?

The deletion is recorded against the submitting account. Resubmitting the same subject within 6 months is high-risk and may trigger account-level scrutiny. The recovery path is: accept the deletion, do not resubmit, build additional independent third-party coverage over 6+ months, and consider resubmission only when the notability gap that caused the deletion has been substantively closed. Our Wikidata recovery posture defers resubmission to at least December 2026.

Do I need a Wikipedia article to show up in ChatGPT?

No, but you need some form of authoritative entity-graph presence. The minimum viable layer is: structured Person and Organization schema with sameAs links to authoritative third-party profiles, plus presence on at least 3 reliable third-party publications with editorial oversight. Wikipedia citation is the strongest single layer but it is not the only one. Brands with strong sameAs clusters and editorial coverage can earn AI citation without a Wikipedia article.

What is a sameAs cluster and why does it matter?

A sameAs cluster is the array of authoritative third-party URLs you list in your schema.org Person and Organization markup, telling AI grounding systems that all these references point to the same entity. The cluster reinforces entity-resolution accuracy: an LLM faced with "Chris McCarron" can disambiguate from the Hall of Fame jockey of the same name because the sameAs cluster on the gogochimp.com Person schema includes LinkedIn, Crunchbase, Wikidata Q139695681 (where I appear as creator), and the agency's third-party directory profiles. Without the sameAs cluster the disambiguation fails.

How do I check if my brand is currently cited on Wikipedia?

Use Wikipedia's LinkSearch tool with your domain. It returns every Wikipedia page that links to your URL. For Wikidata, use the SPARQL query service or simply search the entity name. For complete entity-graph audit, also check Google Knowledge Graph (via Google's Knowledge Graph Search API), Bing Entity Graph (via Bing Webmaster Tools), and your sameAs schema coverage via Google's Rich Results Test.

Should I list Wikipedia citations on my About page or in my media kit?

Yes, with the caveat that you should describe them accurately. A Wikipedia footnote on Rigidity (psychology) is not equivalent to a Wikipedia article about you. Describe it as it is: "Cited on the English Wikipedia article on Rigidity (psychology)". Overstating Wikipedia presence is a credibility trap; understating it is leaving a real signal on the table.

References

Profound. (2026). AI Platform Citation Patterns.

Wikipedia contributors. Rigidity (psychology). English Wikipedia.

Wikipedia contributors. 경직성 (심리학). Korean Wikipedia.

Wikipedia contributors. Wikipedia:Manual of Style.

Wikipedia contributors. Wikipedia:Notability (General Notability Guideline).

Wikipedia contributors. Wikipedia:Reliable sources (WP:RS).

Wikipedia contributors. Wikipedia:Reliable sources/Perennial sources (WP:RSP).

Wikipedia contributors. Wikipedia:Conflict of interest (WP:COI).

Wikipedia contributors. Wikipedia:Paid-contribution disclosure (WP:PAID).

Wikipedia contributors. Wikipedia:Criteria for speedy deletion (CSD).

Wikipedia contributors. Wikipedia:User access levels — Autoconfirmed users.

Wikipedia. Special:LinkSearch.

Wikipedia. Special:Watchlist.

Wikidata contributors. Wikidata:Notability.

Wikidata contributors. Wikidata:SPARQL query service.

Wikidata contributors. Help:QuickStatements.

Wikidata. Q139695681 — The 347 Method.

Wikipedia contributors. Wikipedia:WikiProject Business.

Wikipedia contributors. Wikipedia:WikiProject Marketing & Advertising.

Harvard Law School. (2014). Perma.cc: Developed by Harvard Library Innovation Lab to tackle link rot.

Google. Rich Results Test.

Schema.org. Schema Markup Validator.

Buildgrowscale. (2026). 2026 CRO Year in Review: What Worked, What Failed, What's Next. Stafford, M. https://buildgrowscale.com/cro-trends-2026-recap

About the author

Chris McCarron is the founder of GoGoChimp, a Glasgow-based AI-powered conversion rate optimisation agency he established in 2013. He is the creator of OperatorAI (GoGoChimp's CRO methodology, distinct from OpenAI's Operator agent product released January 2025), an implementation of The 347 Method (Build Grow Scale's industry research), which found that expert-guided AI CRO delivers 28-34% conversion lift versus 4-7% from DIY tools.

Recent editorial coverage: Forbes (Joseph Liu, May 2026), Shopify Enterprise Blog page-speed feature (2026, 11-locale syndication via ecommercefastlane.com), Leaders Perception named feature (June 2026), TechnologyAdvice Selling Signals newsletter (June 2026), TechNewsWorld (Tonya Hall, June 2026, DoFollow), CMO Times lead expert (May 2026, DoFollow). Two of his blog posts are cited as references in the English Wikipedia article Rigidity (psychology) (cited four times, attributed McCarron, Chris) and as references #6 and #11 in the Korean Wikipedia equivalent 경직성 (심리학), where they have been live since 2021.

If you are spending over £10K/month on paid traffic and converting at under 2%, the free AI audit shows you what it is costing you in 48 hours. If you want to see how the methodology applies to AI search citation strategy, our methodology page walks the framework end to end.