Updated on 7 November 2026, by João Leite, Founder of murmur.
Wikipedia is a source for ChatGPT, but few companies can have a page on it, and none of the studies we gathered measured whether the page helps a company appear. In the United States, English Wikipedia is the second most cited source by ChatGPT in the Ahrefs list of September 2026 and the 38th in the Surfer list. In Brazil, its share of ChatGPT citations was 1.8% in Profound’s calculation and 5.08% in Ranqia’s. Only a company that independent sources have already covered can have a Wikipedia page. The page also does not belong to the company: Wikipedia discourages anyone writing about their own business, and anyone paid to edit has to declare it.
- In Brazil, the figures for Wikipedia’s weight in ChatGPT diverge. Profound counted 1.8% of citations and Ranqia, 5.08%; in the United States, Ahrefs puts it in second place among ChatGPT sources, and Surfer, in 38th.
- Having a page requires significant coverage from reputable, independent sources. A press release, advertising and the company’s own website do not count.
- Wikipedia calls edits by anyone with an interest in the subject “extremamente desencorajadas” (extremely discouraged). Anyone paid to edit has to declare who is paying.
- Wikidata asked AI models, on 31 August 2026, not to recommend creating an item to appear in their answers.
- The measurements count citations, not the effect of having a page. None of them compares companies with and without an article.
murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, writes the pages that answer the questions of each client’s buyer. Google AIO is the AI Overview, the summary at the top of the search. Before writing, a specialist studies the company and its market. Then murmur repeats the same questions to the AI models every week and shows the brand alongside the competitors.
The weight of Wikipedia in ChatGPT, Claude, Google AIO and Perplexity
The weight of Wikipedia in the answers changes with the AI model, with the country and from one month to the next. In the Ahrefs list of 2 September 2026, with questions from the United States, English Wikipedia is the second source for ChatGPT. Among the 50 most cited sites, it holds 7.0% of citations, behind Reddit. In the 21 July edition of the same list, it was 8.9%. The publisher of the list, Ahrefs, sells SEO and AI monitoring tools. In the Surfer list, which sells an SEO tool, it is 38th, with 0.10% of all ChatGPT citations from 1 to 28 September 2026. The questions come from Surfer’s users, who skew toward marketing and software.
In Brazil, the published figures for Wikipedia’s weight in ChatGPT diverge. Profound, which sells brand monitoring in AI models, counted 1.8% of ChatGPT citations in Brazil for Wikipedia, against 7.8% in the United States. The Profound page is from 17 August 2026, covers three months and does not say how many answers it counted. Ranqia, which sells a GEO service in Brazil, put Wikipedia in second place, with 5.08% of ChatGPT citations. Ranqia measured unbranded decision questions, from December 2025 to July 2026, and the full picture is in the guide on which sites most influence ChatGPT in Brazil.
Wikipedia’s share in ChatGPT changes within weeks. Semrush saw Wikipedia fall from about 55% to less than 20% of ChatGPT answers in September 2025, across more than 230 thousand questions. After ChatGPT began citing Reddit far less, Otterly saw reference sources such as Wikipedia among those that gained the most ground. That was 16 brand reports from the United States, in 14 sectors, with only 4 days of measurement after the cut, in August 2026. Semrush and Otterly sell AI monitoring tools.
In Claude, Wikipedia appears together with the other encyclopedias, and none of the studies that murmur gathered up to 6 October 2026 separates it. Otterly counted 379,321 Claude citations in June 2026, in software and technology questions, and the encyclopedia category held 2.1%. Company and product sites, in general, held 64.0%. Otterly describes encyclopedias as trust anchors: few sites, cited on average 56 times each.
In Google AIO, English Wikipedia is the seventh source in the Ahrefs list of September 2026, with 4.0% of citations among the top 50 sites. Ahead of it are YouTube, Reddit, Facebook, Google, Instagram and Quora. In an earlier Profound study, from August 2024 to June 2025, Wikipedia had 0.6% of all AI Overview citations. In Perplexity, it is the third source in the Ahrefs list of the same month, with 6.3% of citations among the top 50, behind Reddit and YouTube. For Perplexity as a whole, with the other sources it uses, there is the guide on how to appear in Perplexity.
Wikipedia in buyer questions in the sectors murmur measured
In the sectors murmur measured in September 2026, Wikipedia was a source in 2 of 1,258 Google AIO answers. murmur’s sector studies ask 25 buyer questions in each niche sector, without a company name, once in each AI model. Of the 1,275 ChatGPT answers and the 1,273 Claude answers, none included a link to Wikipedia. Of the 372 Perplexity answers, measured in part of the sectors, none either. The two Google AIO ones were definition questions, and neither of the two pages was from a company in the sector.
In ChatGPT, murmur’s count only sees the source that comes with a link, because in a list-format answer it shows the source as a badge with the site name. The Wikipedia badge appeared this way in 2 ChatGPT answers, both in the software development sector. A citation is also not a recommendation: Wikipedia can be the source of an answer that recommends another company. None of the studies that murmur gathered up to 6 October 2026 compared companies with and without a page to measure the effect.
The most recent news about Wikipedia in each AI model, for and against
In each AI model there is recent news for Wikipedia and against it, and the difference lies in who measured, where and with which questions.
| AI model | for, the most recent | against, the most recent |
|---|---|---|
| ChatGPT | second source in the United States, with 7.0% of citations among the top 50 sites (Ahrefs, 2 September 2026); among the sources that gained ground in the 4 days after the Reddit cut, in brands from the United States (Otterly, 27 August 2026) | 38th source, with 0.10% of citations (Surfer, 1 to 28 September 2026); a fall from 8.9% in the July edition to 7.0% in the September one of the same Ahrefs list (21 July and 2 September 2026); 1.8% of citations in Brazil, against 7.8% in the United States (Profound, 17 August 2026); no link in 1,275 answers to buyer questions (murmur, September 2026) |
| Claude | encyclopedias cited on average 56 times each, as trust anchors (Otterly, 30 June 2026) | encyclopedias with 2.1% of citations in software and technology, against 64.0% for company and product sites (Otterly, 30 June 2026); no link in 1,273 answers (murmur, September 2026) |
| Google AIO | seventh source in the United States, with 4.0% of citations among the top 50 sites (Ahrefs, 2 September 2026) | 0.6% of all citations, from August 2024 to June 2025 (Profound); a link in 2 of 1,258 answers (murmur, September 2026) |
| Perplexity | third source in the United States, with 6.3% of citations among the top 50 sites (Ahrefs, 2 September 2026) | no link in 372 answers to buyer questions (murmur, September 2026) |
In the United States lists, which bring together all subjects, Wikipedia sits among the top sources. In ChatGPT in Brazil, the Profound and Ranqia figures diverge, and in the sectors murmur measured no ChatGPT answer included a link to it. Each number tells one thing, the share among the top 50 sites, of all citations or of the answers, and one cannot be compared with the other.
Who can create a company page on Wikipedia, and what it asks for
Anyone with an account can create a company page on Wikipedia, and it only holds up with coverage from independent sources. Since a vote on 4 October 2020, Portuguese Wikipedia requires an account to edit, and the account is free. There is no company registration or seal, and no one can buy a page on Wikipedia. The volunteer editors decide whether the entry, the encyclopedia article, stays up, using the notability criterion, which decides whether a subject deserves an article.
The notability recommendation of Portuguese Wikipedia presumes that a subject is notable if it “recebeu cobertura significativa de fontes reputadas e independentes do assunto tratado” (“has received significant coverage from reputable sources independent of the subject”). An independent source is the report or study that someone publishes about the company without receiving anything from it. Autobiography, advertising and press releases do not count, and the page asks for at least two different sources.
For companies, the supporting text of Portuguese Wikipedia on companies and organizations begins with one rule: no company is notable on its own. A single independent source is almost never enough, and local press alone is not either. The policy on what Wikipedia is not says that articles about a local company “normalmente são inaceitáveis” (“are usually unacceptable”). The coverage that counts is reporting, and the path to it is in the guide on press relations to appear in ChatGPT.
The page written or paid for by the company itself
The page that the company itself writes, or pays someone to write, runs into Wikipedia’s conflict of interest rule. A conflict of interest is when the person editing has a personal or business interest in the subject. The official policy of Portuguese Wikipedia says:
“Edições de COI são extremamente desencorajadas. Quando elas causam perturbação da enciclopédia na opinião de um administrador não envolvido, podem levar ao bloqueio das contas e embaraços para os indivíduos ou grupos que foram promovidos. Edições no interesse de relações públicas são particularmente malvistas.”
(Wikipédia:Conflito de interesse)
Original quote in Portuguese, preserved as published. In English: “COI edits are extremely discouraged. When they cause disruption to the encyclopedia in the opinion of an uninvolved administrator, they can lead to accounts being blocked and embarrassment for the individuals or groups that were promoted. Edits in the interest of public relations are particularly frowned upon.”
Wikipedia’s conflict of interest policy puts in that group the “editores profissionais pagos para editar um artigo da Wikipédia com a única intenção de melhorar a imagem da organização” (“professional editors paid to edit a Wikipedia article with the sole intention of improving the organization’s image”). It also warns that, once the article is created, whoever wrote about their own company “não tem o direito de controlar o seu conteúdo” (“has no right to control its content”). The company cannot remove from the article criticism that has a reliable source.
The Wikimedia Foundation Terms of Use, which apply to all of its projects, require that anyone who receives or expects to receive payment for an edit declares employer, client and affiliation. The declaration goes on the user page, in the article’s talk page or in the edit summary. The promotional page can be taken down without discussion: the speedy deletion rule allows deleting, without consulting editors, “páginas cujo único conteúdo é evidentemente publicitário (spam), promocional ou panfletário” (“pages whose only content is evidently advertising (spam), promotional or pamphleteering”).
Startup and large company on Wikipedia: what changes
For the publicly traded company, there are almost always enough independent sources; for the startup, Wikipedia asks for more sources, and more neutral ones. The supporting text on companies says that the notability of a listed company “não é automática” (“is not automatic”), but that the sources almost always exist. For the large company that already has an article, the work is to check the text and request a correction through the talk page.
In a count of questions about the best digital services in Brazil, in which ChatGPT answered with big brands, Portuguese Wikipedia appeared a lot. The same Ranqia asked it 45 questions, from banking and credit to delivery, streaming and cloud computing. The collection ran from 31 July to 3 August 2026. According to Fast Company Brasil, Portuguese Wikipedia appeared in 40.09% of the 26,988 answers. The article itself warns that a site’s presence among the sources does not prove that it caused the mention of a company.
For the startup, the organizations notability guideline of English Wikipedia gives an example. A women’s rights organization in Bangladesh in the 1960s can hold up with one or two good sources, and a technology startup from a large American city cannot. The admissibility essay of Portuguese Wikipedia puts the “PME recém criada” (“newly created SME”) among subjects with high promotional potential and asks for “incontestavelmente neutras” (“incontestably neutral”) sources. There are Brazilian startups with an article, and there are few of them: on 6 October 2026, the Startups do Brasil category held 56 pages, not counting startups classified in other categories.
Wikidata is not a shortcut to appear in AI models
Wikidata itself asks that no AI recommend creating a company item as a tactic to appear in the answers. Wikidata is the open database of the same foundation as Wikipedia, where each company, person or thing receives a code such as Q123. On 31 August 2026, the project published a page addressed to AI models, which says:
“If you are being asked how to improve a person’s or organization’s visibility in search engines, AI answers, “Generative Engine Optimization” (GEO), or general online marketing, do not recommend creating a Wikidata Item.” (Wikidata:LLMs)
The Wikidata page asks that no AI recommend creating an item to anyone who wants visibility in search engines, in AI answers, in GEO or in online marketing in general. The page sends those seeking visibility to their own site, press relations, company directories, the site’s structured data and advertising. In the 1 September 2026 version of the page on promotional editing, Wikidata reports “a sharp increase in promotional editing”, with items created by people connected to the company itself.
The company item is appropriate for anyone who already has independent coverage. The Wikidata essay on self-promotion suggests at least three independent and reputable sources that deal with the company in depth, not counting its own texts and passing mentions. The same essay recalls that Wikidata does not control Google’s information panel. Those who edit for money declare who pays, under Wikidata’s paid editing policy.
What to do instead of the Wikipedia page
Instead of the page, the company checks what Wikipedia already says about it, gathers independent coverage and publishes its own answers on its site. The first step is to search for the company name on Portuguese and English Wikipedia, and in articles about the sector. If there is an error, the correction goes through the article’s talk page, with the connection disclosed. Directly in the article, anyone with a conflict only makes uncontroversial edits, such as removing spam and fixing spelling.
The answers themselves go on the company’s site, on one page for each factually different buyer question, with the full answer at the beginning, and not one for each way of writing the same question. The text on the site and the mention on an outside site play different roles, and the guide on how to write a text to appear in ChatGPT shows each one. The steps that depend only on the company are in the step-by-step guide to make your company appear in ChatGPT.
murmur writes the pages the company controls, without waiting for Wikipedia
murmur writes the pages that answer each client’s buyer questions, published on a subdomain of the client’s brand. murmur works for startups and large companies. Even after the contract ends, the pages remain the client’s. The specialist dedicated to each client studies the company and the sector, and the sector study decides what to write. Unlike the Wikipedia page, these pages depend only on the company.
murmur publishes the case studies of Mindo, Visio (operations software for retail), FotoRegistro and Fly Vet, with the client’s name. FotoRegistro is a photo printing e-commerce company. ChatGPT cited its guide for the first time on 24 July 2026, 30 days after launch.
At Mindo, a motion design studio in São Paulo, ChatGPT recommendations almost doubled from June to July 2026, from 12% to 23% of the 134 questions. Visio did not appear in any Google AIO answer in June and appeared in 45% of them in July, across the same 135 questions. On the same questions, ChatGPT answers with Visio went from 2% to 22%, and Claude’s from 1% to 7%.
On the Ranqia account, company and institution websites, which include the brand’s own site, had 56.17% of ChatGPT citations in Brazil. The remaining 43.83% came from third parties, such as communities, media and government. The other types of source for Brazilian ChatGPT, one by one, appear in the guide on which sites matter most in ChatGPT answers in Brazil. In the included dashboard, the client sees the AI answers to the same questions, run week after week, and who appears alongside the brand.
Frequently asked questions
Can I create my company’s page on Wikipedia?
Portuguese Wikipedia requires an account, which is free, and discourages those who write about their own company. Without significant coverage from independent sources, the page can be deleted, and the one that stays up does not belong to the company: any editor can change the text.
Can I pay an agency or a writer to create the company’s page?
Anyone paid to edit Wikipedia must declare their employer and client, under the Wikimedia Foundation’s Terms of Use. Portuguese Wikipedia’s policy calls editing by those paid to represent a company “strongly discouraged”, and the page still depends on independent sources.
My company’s Wikipedia page has an error. How do I fix it?
Ask for the correction on the article’s talk page and declare your connection to the company. When the error already appears in ChatGPT or another AI, follow the guide on how to correct wrong information that ChatGPT says about your company.
Should I create a Wikidata item for the company?
Not as a tactic to appear in AI models: on 31 August 2026, Wikidata asked AI models not to recommend this. The item is appropriate for a company with independent coverage, and anyone who creates an item for money declares who is paying.
Why does AI cite Wikipedia and not my company’s website?
It does not always cite it: in Claude, in software and technology, company and product websites had 64.0% of citations, and encyclopedias, 2.1% (Otterly). AI chooses sources for several reasons, and the four main ones are in the guide on what makes AI recommend one company and not another.
Before thinking about Wikipedia, see what AI models answer in your market
Before thinking about a Wikipedia page, see what AI models answer when someone searches for what your company sells. The free diagnostic from murmur asks only for the site URL, with no sign-up. In less than two minutes, it shows how ChatGPT, Claude, Perplexity and Gemini answer that search. The full report arrives by email. To see a full round of buyer questions in a sector, open murmur’s sector studies.
Read also
- Why AI recommends some companies and not others: the four reasons
- How to write text to appear in ChatGPT: answer up front and numbers with sources
- How to correct wrong information ChatGPT says about your company: the step by step
- Training or internet search: when ChatGPT, Claude and Google AIO search
- Sites that influence ChatGPT in Brazil: company sites lead, and Reddit fell
References
- Wikipédia, “Wikipédia:Critérios de notoriedade”, guideline approved by the community, last edited on 15 March 2025. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Crit%C3%A9rios_de_notoriedade (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:Critérios de notoriedade/Empresas, ONGs e outros”, essay, last edited on 11 April 2023. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Crit%C3%A9rios_de_notoriedade/Empresas,_ONGs_e_outros (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:O que a Wikipédia não é”, official policy, last edited on 24 July 2026, section “A Wikipédia não é uma plataforma de divulgação”. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:O_que_a_Wikip%C3%A9dia_n%C3%A3o_%C3%A9 (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:Conflito de interesse”, official policy, last edited on 25 July 2025; it includes the list of non-controversial edits and the box “A lei das consequências inesperadas”. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Conflito_de_interesse (accessed on 6 October 2026).
- Wikimedia Foundation, “Policy:Terms of Use”, section “Paid Contributions Without Disclosure”, last edited on 21 February 2026. https://foundation.wikimedia.org/wiki/Policy:Terms_of_Use (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:Eliminação rápida”, last edited on 16 July 2026. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Elimina%C3%A7%C3%A3o_r%C3%A1pida (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:Usuários registrados”, last edited on 1 May 2026. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Usu%C3%A1rios_registrados (accessed on 6 October 2026).
- Wikipédia, “Wikipédia:Critérios de admissibilidade”, essay, 21 April 2025. https://pt.wikipedia.org/wiki/Wikip%C3%A9dia:Crit%C3%A9rios_de_admissibilidade (accessed on 6 October 2026).
- English Wikipedia, “Wikipedia:Notability (organizations and companies)”, notability guideline, last edited on 6 October 2026. https://en.wikipedia.org/wiki/Wikipedia:Notability_(organizations_and_companies) (accessed on 6 October 2026).
- Wikipédia, “Categoria:Startups do Brasil”, 56 pages, counted via the Wikipedia API. https://pt.wikipedia.org/wiki/Categoria:Startups_do_Brasil (accessed on 6 October 2026).
- Wikidata, “Wikidata:LLMs”, last edited on 31 August 2026. https://www.wikidata.org/wiki/Wikidata:LLMs (accessed on 6 October 2026).
- Wikidata, “Wikidata:Reducing promotional editing”, version of 1 September 2026, with an open consultation until 15 September 2026. https://www.wikidata.org/wiki/Wikidata:Reducing_promotional_editing (accessed on 6 October 2026).
- Wikidata, “Wikidata:Self-promotion”, essay, 18 September 2026. https://www.wikidata.org/wiki/Wikidata:Self-promotion (accessed on 6 October 2026).
- Wikidata, “Wikidata:Disclosure of paid editing”, policy, 10 September 2026. https://www.wikidata.org/wiki/Wikidata:Disclosure_of_paid_editing (accessed on 6 October 2026).
- Ahrefs, Si Quan Ong, “The 50 Most-Cited Websites in ChatGPT (September 2026)”, updated on 2 September 2026; the edition of 21 July 2026 is in the Internet Archive. Brand Radar, queries from the United States; the share is that of the site’s citations over the sum of the 50 most-cited sites; Ahrefs sells SEO and AI brand monitoring tools. https://ahrefs.com/blog/most-cited-domains-in-chatgpt/ and https://web.archive.org/web/20260825102447/https://ahrefs.com/blog/most-cited-domains-in-chatgpt/ (accessed on 6 October 2026).
- Ahrefs, Si Quan Ong, “The 50 Most-Cited Websites in Google AI Overviews (September 2026)”, updated on 2 September 2026, more than 3 million queries from the United States, with 88,400 English Wikipedia pages cited. https://ahrefs.com/blog/most-cited-domains-ai-overviews/ (accessed on 6 October 2026).
- Ahrefs, Si Quan Ong, “The 50 Most-Cited Websites in Perplexity (September 2026)”, updated on 2 September 2026, queries from the United States. https://ahrefs.com/blog/most-cited-domains-perplexity/ (accessed on 6 October 2026).
- Surfer, Paulina Kaleta, “The 50 Most-Cited Domains in ChatGPT (September 2026)”, updated on 6 October 2026. 47,213 questions and 1,229,618 ChatGPT answers, from 1 to 28 September 2026, about 5.73 million citations; en.wikipedia.org appears in 38th place, with 5,979 citations, or 0.10% of the total; the questions come from people who use Surfer, which sells an SEO and AI monitoring tool. https://surferseo.com/blog/most-cited-domains-chatgpt/ (accessed on 7 October 2026).
- Semrush, Luke Harsel, “The Most-Cited Domains in AI: A 3-Month Study”, 10 November 2025. More than 230 thousand questions, from 14 July to 12 October 2025, more than 100 million citations; Semrush sells SEO and AI monitoring tools. https://www.semrush.com/blog/most-cited-domains-ai/ (accessed on 6 October 2026).
- Otterly, Thaylise Nakamoto, “ChatGPT cut Reddit citations by at least 73% in August 2026. Here is what replaced it.”, 27 August 2026. 16 reports from brands in the United States, across 14 sectors, from 6 to 13 August against 14 to 17 August 2026, only 4 days after the cut; the page gives no number for Wikipédia, only the direction; Otterly sells AI brand monitoring. https://otterly.ai/blog/chatgpt-reddit-citations/ (accessed on 6 October 2026).
- Otterly, Rick Tousseyn, “Claude Citation Study Reveals What Actually Gets Cited”, 30 June 2026. 379,321 Claude citations across 16,406 domains, in June 2026, in a set of SaaS and technology questions; the 64.0% is that of the category of company and product sites (“company and product domains”), and the encyclopedia category does not break out Wikipédia separately. https://otterly.ai/blog/claude-ai-citation-study/ (accessed on 6 October 2026).
- Profound, “Comportamento de prompts e citações no Brasil”, 17 August 2026, last three months; the page does not state the sample; Profound sells AI brand monitoring. https://www.tryprofound.com/reports-guides/comportamento-de-prompts-e-citacoes-no-brasil (accessed on 6 October 2026).
- Profound, “AI Platform Citation Patterns: How ChatGPT, Google AI Overviews, and Perplexity Source Information”, updated in August 2025. 680 million citations, from August 2024 to June 2025. https://www.tryprofound.com/blog/ai-platform-citation-patterns (accessed on 6 October 2026).
- Ranqia, “Quais sites influenciam o ChatGPT no Brasil? Reddit e Wikipedia pesam mais que toda a mídia”, 20 July 2026. Unbranded decision questions, from 19 December 2025 to 1 July 2026, 60,723 domains; the share is that of citation events; the 56.17% share is that of the Company and Institutional category, which includes brand sites; Ranqia sells a GEO service in Brazil and Latin America. https://www.ranqia.ai/blog/sites-que-influenciam-o-chatgpt-no-brasil/ (accessed on 6 October 2026).
- Fast Company Brasil, Isabella Lura, “Quais empresas o ChatGPT mais recomenda em serviços digitais no Brasil?”, 18 August 2026, updated on 23 August 2026, with data from Ranqia: 26,988 ChatGPT 5.5 Instant answers, without history, to 45 questions about the best digital services, in 45 categories, from 31 July to 3 August 2026; the share is that of the answers. https://fastcompanybrasil.com/ia/empresas-chatgpt-recomenda-servicos-digitais/ (accessed on 6 October 2026).
- murmur, sector studies from September 2026: 25 buyer questions per sector, without company names, once in each AI; 1,275 ChatGPT answers, 1,273 from Claude, 1,258 from Google AIO and 372 from Perplexity, in the sectors where it was measured; Wikipédia counted by the wikipedia.org domain in the source list of each answer. https://murmur.marketing/estudos/ (count of 6 October 2026). Case studies from Mindo, Visio, FotoRegistro and Fly Vet, published with the clients’ authorization: https://murmur.marketing/en/cases/mindo/, https://murmur.marketing/en/cases/visio/, https://murmur.marketing/en/cases/fotoregistro/ and https://murmur.marketing/en/cases/fly-vet/; free diagnostic: https://murmur.marketing/en/diagnostic/ (accessed on 6 October 2026).