An evidence-led media investigation / Updated 20 October 2026

Africa Check:
Who Checks the Fact-Checkers?

A Johannesburg-based organisation can publish a fact-check that travels far beyond its own website. Its relationships reach international foundations, journalists, election officials and social-media platforms. This investigation explains how that works, where the money comes from, and which claims about its influence actually survive scrutiny.

Research brief · Institutional records · Platform disclosures · Published case studies · Every disputed assertion labelled

01 / The basics

First, what exactly is Africa Check?

Africa Check is a non-profit fact-checking organisation founded in 2012. Its researchers investigate public statements and publish evidence-based assessments of whether those claims hold up. It is not a government department, a court or the owner of Facebook.

Its origins lie in a journalism initiative developed by Peter Cunliffe-Jones, with support from AFP Foundation and a base at the University of the Witwatersrand (Wits) in Johannesburg. Wits is a public research university; hosting a fact-checker means providing an institutional base, not automatically exercising editorial control over individual articles. [1]

The organisation originally had a UK corporate structure. Africa Check says control transferred to the South African Africa Check Trust in 2019, registration IT000728/2015(C). It also operates beyond South Africa, including work in Kenya, Nigeria and Senegal. [2]

The essential distinction: a fact-check is an editorial conclusion. Its real-world significance changes when other media repeat it, institutions use it for training, or a technology platform incorporates it into distribution decisions.

2012Year founded
2019Transfer of control to the South African Trust
58Organisations in the Africa Facts Network, as described in 2026
02 / How it grew

From a newsroom experiment to a continental institution

The timeline matters because the relationships did not all begin together. It is misleading to present a 2016 funding ratio, a 2024 Meta statement and a 2026 network size as if they describe one moment.

The launch in Johannesburg

Peter Cunliffe-Jones launches Africa Check with early journalism-sector backing and Wits as a host. The original proposition is simple: publicly check consequential claims made by politicians, officials and media. [1]

International standards and the social-platform era

The International Fact-Checking Network (IFCN), associated with the Poynter Institute in the United States, establishes common standards for fact-checkers. Cunliffe-Jones takes part in developing the Code of Principles. Facebook begins relying on outside fact-checkers as part of its misinformation response. [6]

The Africa Facts Network begins

Africa Check helps organise cooperation among fact-checking groups across the continent. This later becomes a much larger network, which the organisation describes as 58 members in more than 35 countries in 2026. Membership does not establish editorial control over each member. [4]

Control of the organisation moves to the South African trust

The UK-registered company remains part of the organisation's history, but Africa Check states that the South African trust becomes the controlling entity. [2]

Election-period integration becomes explicit

Meta says it works with Africa Check and AFP in South Africa. It describes reduced Feed distribution for debunked content and says Africa Check assisted training for Electoral Commission personnel. [5]

Meta changes its US programme

Meta announces an end to third-party fact-checking in the United States in favour of Community Notes. That US decision cannot be used by itself to infer the exact status of every non-US agreement in October 2026. [8]

Africa Check's current structure is the story

Its network role, public grant relationships and links to election-information projects raise a wider question: how should organisations with outsized influence be scrutinised and held accountable?

03 / Follow the money

Who pays, and what can we prove?

Funding relationships deserve scrutiny. But reporting them accurately requires distinguishing a named funder, a historical percentage, a project partnership and an allegation of editorial influence. Those are not interchangeable.

DOCUMENTED RELATIONSHIP

The funding collaborative

Africa Check's public partner information describes a pooled-support arrangement involving Open Society Foundations, Social Justice Initiative, This Day Foundation and Yellowwoods Social Investments. This is an important current funding relationship, not proof that those organisations approve its verdicts. [3]

Why readers should care: donors may shape what kinds of capacity an NGO can build through grants and priorities. Testing the influence of that funding requires grant terms, budgets, governance disclosures and decisions—not assumptions about a donor's politics.

NAMED ON PARTNER MATERIAL

Open Society Foundations

A philanthropic network founded by George Soros. Its participation in a current funding collaborative is relevant to financial transparency; it does not establish editorial instructions to staff. [3]

HISTORICAL AND REPORTED

Gates Foundation

Has supported Africa Check over earlier reporting periods. A widely repeated 13% figure is historical, not a verified 2025 or 2026 share. Do not put it in a contemporary funding pie chart. [2]

PERIOD-DEPENDENT

Luminate / Omidyar

Historically associated with fact-checking support. Its funding must be documented by year and grant before claiming a present budget share. A relationship with fact-checking infrastructure is not proof it determined any specific rating. [2]

PLATFORM RELATIONSHIP

Meta

Meta's own 2024 announcement confirms an operational relationship with Africa Check. It does not disclose enough financial information to justify the old page's undated claim that Meta supplied approximately 10% of revenue. [5]

Why the previous percentages were removed

The older page displayed Gates 13%, OSF 12%, Luminate 8%, Meta roughly 10%, Google roughly 8% and “own earnings” of 15%. These figures were presented without one consistent, identified accounting year or primary financial statement. They therefore did not constitute a reliable current budget breakdown. The replacement deliberately avoids a pie chart until a specific audited year, accounting basis and underlying document can be cited.

Still outstanding: the exact percentages in the organisation's 2025 income breakdown require direct confirmation against the underlying chart/financial report. UNVERIFIED AS EXACT SHARES

04 / The relationships

How influence moves between institutions

This is an explanatory diagram of documented relationship types. The boxes do not represent a single command structure, shared agenda or inevitable chain of censorship.

FundersGrants, pooled support, project finance
Africa CheckFact-check articles, research, training
IFCN / PoynterProfessional standards and independent assessment
Media / civil societyCitations, training and network collaboration
MetaPlatform rules and distribution actions

What is the IFCN?

The International Fact-Checking Network is a Poynter initiative that sets standards of transparency, methodology and corrections. Africa Check founder Peter Cunliffe-Jones helped develop its early Code of Principles and subsequently worked as an adviser. This is relevant professional overlap, but there is no evidence he personally certified Africa Check's individual applications. [6]

Why Wits matters

Wits Journalism provides an academic home in Johannesburg. Hosting creates an institutional connection; it does not establish that every Wits department, external researcher or university donor directs Africa Check's editorial judgements. [1]

The Africa Facts Network

Based on Africa Check's 2026 description, its network encompasses 58 organisations in more than 35 countries, with 25 IFCN signatories. This is evidence of reach and coordination. It is not evidence that all member newsrooms publish identical judgements. [4]

Journalists and election bodies

Work with newsrooms, research partners and election officials can extend a fact-checker's methods into public institutions. During the 2024 election preparations, Meta publicly described Africa Check's role in training Electoral Commission staff. [5]

05 / The direct mechanism

What happens after Africa Check rates a claim?

This is the most consequential verified part of the investigation. Meta's April 2024 South African election statement identifies Africa Check and AFP as fact-checking partners and describes what happens when partners debunk content. [5]

A claim appears

Someone publishes a statement or post—for example, a statistic about land ownership, a political speech or a claim about an election.

A fact-checker investigates

Researchers choose a specific checkable proposition, identify sources, assess the data and publish a rating. The selection of what gets checked can matter as much as the verdict.

The platform receives an assessment

Under Meta's described 2024 programme, fact-check partners could rate material that did not otherwise breach rules requiring removal.

Meta determines distribution consequences

Meta said debunked material could receive warning labels and reduced distribution in Feed. That is platform enforcement using external editorial assessments, not Africa Check directly controlling the platform algorithm. [5]

“When content is debunked by these fact checkers, we attach warning labels to the content and reduce its distribution in Feed” — Meta, South African election preparations, April 2024
IMPORTANT LIMIT

Does this prove Afrikaner content was targeted?

No. It proves a moderation mechanism existed. To show Afrikaner-specific consequences we need the original post or URL, Africa Check rating, dates, Meta notification and any reach data or appeal. The present research does not document that full chain for a named case.

Meta changed its US approach in January 2025. Regional arrangements must be checked on their own terms, rather than assuming the 2024 South African description remains unchanged worldwide. [8]

06 / Test the verdicts

The case files: where the argument meets the evidence

DOCUMENTED FACT-CHECK

Case 1 — What is the farm-murder rate?

Farm attacks and murders are serious crimes. But calculating a rate is different from counting incidents. A rate needs a numerator (the number of victims meeting a defined category) and a denominator (the population truly at risk). An often-repeated figure of 133 per 100,000 divided murders that included multiple kinds of people on farms by a much narrower estimate of full-time farmers. Those categories did not match.

Africa Check's 2017 analysis said a defensible rate was extremely difficult to calculate because the relevant population was not sufficiently defined. It did not establish that attacks were harmless or that a low alternative rate was definitive. The organisation also noted that it had removed a rate from its own earlier reporting. [7]

Investigative test: demand disaggregated victim data, clearly defined agricultural populations, consistent years and transparent arithmetic. Criticise weak methodology wherever it appears—whether from a lobby group, police, government or a fact-checker.
COUNTER-EXAMPLE

Case 2 — Africa Check also challenges the EFF

In 2024 Africa Check examined statements by the Economic Freedom Fighters (EFF), a South African political party led by Julius Malema, and rejected or could not substantiate several claims about land, unemployment, informal settlements and poverty. This matters: it contradicts a simplistic assertion that Africa Check only scrutinises Afrikaner-aligned or right-leaning narratives. [9]

Investigative test: measure a sample of Africa Check's decisions by topic, target, evidence quality and correction frequency before claiming systematic bias.
PRIMARY ARTICLE RECOVERED · 13 NOVEMBER 2018

Case 3 — Chinese police stations: what Africa Check dismissed in 2018, and what emerged later

Start with the photograph. In October 2018, pictures circulated showing senior South African police officers and Chinese representatives at the opening of a Chinese Community and Police Cooperation Centre in Port Elizabeth (now Gqeberha). Posts alleged that Chinese people were opening their own police stations in South Africa, with 13 others already in existence. This was a claim about authority and policing, not simply whether a building existed.

13 November 2018 — Africa Check's response. In a report titled “No, Chinese people aren’t opening their own police stations in South Africa”, journalist Tshiamo Malatji cited a South African Police Service (SAPS) media statement and spokesperson Captain Khaya Tonjeni, who said the Port Elizabeth centre was not a police station. The report also quoted centre head Jie Zhang describing a non-profit service offering language support and community-police cooperation. Africa Check consulted South African Institute of International Affairs researcher Cobus van Staden, who warned that poor public communication about these forums could generate understandable concern. [12]

14 November 2018 — Beijing responds separately. China's embassy in South Africa issued a statement rejecting the allegation that 13 Chinese police stations existed. It described the centres as non-profit community associations without law-enforcement authority. Important correction: the embassy's statement was issued after Africa Check's article; it should not be presented as the direct evidence the 13 November article relied upon. [13]

2022 — the international story changes. Human-rights organisation Safeguard Defenders documented a wider network of overseas Chinese police-service facilities. Later reporting put the reported count at 102 facilities across 53 countries. The locations and their actual functions required investigation; a network-wide allegation does not establish misconduct at every listed centre. [14]

April 2023 — US enforcement. The US Department of Justice charged two men over a suspected clandestine Chinese police station in Manhattan. That is serious corroboration that at least one overseas-policing concern was more than an internet rumour. It is not proof that the Port Elizabeth centre was operating unlawfully. [15]

16 February 2024 — Brookings reviews the global evidence. Researchers Emile Dirks and Diana Fu explain how apparently administrative or community premises may, in some settings, intersect with China's united-front operations and transnational repression. They also stress the limits of what is known: not every listed centre has been shown to carry out coercive activity, and some police investigations found no crime. [14]

The proper distinction: Africa Check had evidence in 2018 that the photographed Port Elizabeth facility was publicly described as a community-police cooperation centre, rather than a formal station with police powers. Later global reporting demonstrates why denials and official descriptions cannot, by themselves, settle questions about informal policing, intelligence links or foreign influence. Those later findings warrant fresh scrutiny, but they do not automatically overturn the narrow 2018 factual conclusion about the pictured centre.

What should a careful fact-checker do when authorities deny an allegation?

Ask the police and organisers—but treat their accounts as claims to verify, not proof of absence. Check the centre's registration, governance, financing, affiliations, written agreement with SAPS, access to police systems, physical operations and reports from affected residents. Distinguish a legally empowered police station from a community organisation that might facilitate informal law-enforcement activity. If subsequent evidence changes the context, append a dated review explaining what remains supported and what is newly disputed.

Editorial accountability question: Is the 2018 page adequate as a snapshot of what officials said then, or should it carry a 2024–2026 context note identifying the later international investigations and explaining why Africa Check still stands by its original conclusion? The retrieved article does not display such a contextual update. That is a fair question—not proof of deliberate deception.

FINANCIAL LEAD

Case 4 — The TRi Facts / “own earnings” question

The old page alleged that an entity called TRi Facts accounted for 15% of revenue and that a link reportedly redirected to a gambling website. Without a year-specific primary filing, contemporaneous archived link and reliable independent reporting, the amount and the implication cannot be treated as established. No allegation of impropriety is made here.

Needed next: the year-specific Africa Check accounts, legal entity identification, archived website capture and a documented opportunity for Africa Check to explain the record.
07 / The people and the standards

Who built it, and who runs it?

Peter Cunliffe-Jones

Founder of Africa Check, involved in the IFCN's foundational work and its Code of Principles. His role helps explain how the organisation became connected to the international profession. Participation in writing a standard is not proof he approved individual certification decisions. [6]

Noko Makgato

Appointed executive director in 2019. His professional experience includes South African news organisations. Scrutiny of executive leadership should focus on documented strategy, accounts, governance and published decisions—not speculation about private motivations. [10]

A related academic lead: Herman Wasserman

Professor Herman Wasserman has researched disinformation, journalism and information integrity in South Africa. The previous page claimed that his work supplied the operating definitions behind individual Africa Check verdicts. That causal claim is not established by the evidence currently in this dossier: no particular farm-attack fact-check has been traced to a named Wasserman paper or instruction. His research is relevant context, not proof that he sets Africa Check's editorial decisions. [11]

09 / Unanswered questions

What would make this investigation substantially stronger?

These are testable questions, not conclusions disguised as questions.

01 · Funding by yearWhat did each foundation, commercial client and technology platform actually pay in 2024 and 2025, and what restrictions appeared in each grant agreement?
02 · Moderation audit trailCan a specific South African post be traced from Africa Check verdict to Meta label, downranking, author notification and appeal outcome?
03 · Editorial independenceWhat policies govern recusal when trustees or directors have affiliations with donors, platforms or media partners?
04 · Topic selectionAcross a defined time period, do Africa Check's topic selection, evidence standards and correction rates differ significantly by party, viewpoint or demographic subject?
05 · Current platform statusWhich Meta partner agreements and enforcement actions remain operative specifically in South Africa after the US policy change?
06 · Corrections and responseWhich challenged verdicts were amended, which were upheld, and what published reasoning did Africa Check provide?
08 / Can a funded fact-checker be neutral?

Who funds the referee—and who examines the referee's decisions?

Fact-checking organisations ask the public to trust their independence. That trust must be earned through transparent evidence and correctable decisions, not the mere label “independent”.

First, the documented funding. Africa Check identifies Open Society Foundations as one participant in a funding collaborative with Social Justice Initiative, This Day Foundation and Yellowwoods Social Investments. Its partner list also names the Gates Foundation and the Konrad-Adenauer-Stiftung's media programme, among others. These donors have different priorities and institutional backgrounds; calling every funder “left-wing” would be inaccurate. [3]

Second, the international network. Africa Check is connected to the International Fact-Checking Network (IFCN) at Poynter, whose code sets professional standards for signatories. The IFCN's major Global Fact Check Fund disclosed Google and YouTube backing in 2024. A Poynter report says the termination of USAID-supported media programmes hurt fact-checking organisations globally, but that does not prove the IFCN itself or Africa Check received a specific USAID payment. To allege that payment we would need grant recipient records and a direct financial trail. [16] [17]

Third, the independence test. Receiving funds from an organisation with identifiable political or philanthropic preferences does not automatically make every resulting fact-check biased. Yet financial dependence can create legitimate questions about which subjects receive priority, who sits on governing bodies, whether funders influence programmes, and how complaints against the fact-checker are resolved. That is especially important when a rating has distribution consequences on a large platform.

THE QUESTIONS THAT MATTER

What would convince a sceptical reader?

Publish grant agreements and restrictions, report each significant donor by accounting year, disclose trustee and editorial conflicts, audit whether political groups face equal evidentiary standards, provide clear appeal rights, and append substantive updates to older fact-checks when new evidence changes the context.

Our conclusion: documented institutional funding makes neutrality a question worth testing; it does not establish a secretly directed political agenda. Africa Check's credibility depends on how convincingly its work survives independent scrutiny—including this 2018 Chinese-policing example.

10 / Assessment

Influence is documented. Control has to be proved.

The strongest finding is not a secret conspiracy. It is a visible institutional structure: a fact-checking NGO founded within journalism, funded through philanthropic relationships, connected to international standards, active in continental networks and used by a large platform as part of its information-ranking process.

That structure deserves hard questioning. Who sets the agenda? How are disputes resolved? Which donors fund what? What happens when a verdict is wrong? How can a small publisher or an ordinary South African challenge a consequential finding? The answers matter especially when disputed claims concern rural security, land, language or Afrikaner interests.

But public-interest investigation loses credibility the moment it upgrades inference into fact. Shared funders do not prove collusion. Training election officials does not establish election manipulation. A researcher's publications do not automatically dictate editorial ratings. A Meta moderation policy does not show a particular post was suppressed unless the case record establishes it.

The question is not whether fact-checkers should exist. The question is whether institutions with the power to influence what millions see can withstand the same scrutiny they apply to everyone else.
11 / Evidence ledger

Sources, limitations and corrections

Primary-source links are supplied wherever possible. When the document establishes a relationship but not a financial amount or causal link, the article says so. Reader corrections should identify the specific assertion and a source proving the alternative.

[1] Africa Check — Ten years: a brief historyFirst-party history; founding, launch and institutional origins.Original source ↗
[2] Africa Check — How we are fundedFirst-party corporate and funding disclosures; distinguish periods.Funding disclosure ↗
[3] Africa Check — PartnersFirst-party descriptions of current partner and funding relationships.Partner list ↗
[4] Africa Check — Africa Facts NetworkFirst-party member and geographical scope; time-sensitive counts.Network overview ↗
[5] Meta — Preparing for South Africa's 2024 electionsPrimary evidence for partner names, labelling, distribution and training.Meta statement ↗
[6] Peter Cunliffe-Jones — professional historySelf-reported IFCN standards roles; corroboration required for disputed personnel claims.Professional profile ↗
[7] Africa Check — Why calculating a farm-murder rate is near impossible2017 methodological analysis and clarification of earlier figures.Read analysis ↗
[8] Meta — More Speech and Fewer MistakesJanuary 2025 policy announcement about the United States.Policy announcement ↗
[9] Africa Check — EFF manifesto claimsEvidence that fact-checking also challenged left-wing political statements.EFF report ↗
[10] Africa Check — Our teamLeadership roles and institutional biographies.Team ↗
[11] Stellenbosch University — Information integrity researchAcademic research context; not a demonstrated editorial instruction chain.University announcement ↗
[12] Africa Check — Chinese police station claim (13 November 2018)Original South Africa fact-check, including SAPS, centre head and independent expert interviews.Original fact-check ↗
[13] Embassy of China in South Africa — Statement (14 November 2018)Primary Chinese government response published after Africa Check's article; interested-party assertion, not independent verification.Embassy statement ↗
[14] Brookings — China's overseas police stations (16 February 2024)Analysis of the reported 102 facilities, overseas government investigations and uncertainties.Brookings research ↗
[15] US Justice Department — Manhattan police-station charges (April 2023)US government prosecution; allegations in an individual case, not a legal finding about South Africa.DOJ statement ↗
[16] Poynter / IFCN — Google and YouTube-backed fund (June 2024)Primary disclosure of the Global Fact Check Fund's corporate backing.Poynter funding disclosure ↗
[17] IFCN — 2025 State of Fact-Checkers reportIndustry-wide financial pressures after USAID cuts; not evidence of an Africa Check-specific USAID grant.Industry report ↗

Editorial note, 20 October 2026: The earlier version used unsupported contemporary funding percentages, overstated a direct Wasserman-to-verdict pipeline, previously lacked the now-recovered original Chinese-police-station article and claimed total isolation of Afrikaans media without a comparative study. Those statements have been corrected or explicitly qualified here. This page does not assert wrongdoing from association alone.