Most of what gets called a "source" in marketing decks would fail Wikipedia's sourcing test before a reviewer finishes reading the byline. That gap — between what looks like coverage on a press list and what actually counts under policy — is where the majority of failed Wikipedia drafts die. This guide is a layered breakdown of the rules that govern that distinction in 2026: not a primer for one situation, but the full framework, layer by layer, in the language reviewers actually use.
It builds on our earlier reference on Wikipedia-eligible sources by going deeper into each layer — the policy foundation, the tier map, the Perennial Sources list, the discipline-specific edge cases, and the unwritten scoring conventions reviewers carry in their heads. Use it as a working reference. If you read it carefully, you should be able to evaluate your own coverage the way an experienced AfC reviewer would, and avoid spending a six-figure budget on a draft whose source base was always going to fail.
Layer 1 — The Foundational Framework
Before any tier list, perennial table, or edge case can make sense, you need the three policies that sit underneath the entire reliable-sources apparatus on Wikipedia. They are not optional reading. They are the load-bearing structure that every reviewer, deletion-discussion participant, and dispute mediator silently applies to your draft, and if you do not understand them, you will misread every situation that follows.
The three pillars of WP:RS
The reliable-sources guideline, shortcut WP:RS (formally Wikipedia:Reliable sources), defines reliability on three pillars that must all be present for a source to be treated as reliable for a given claim. The first is independence from the subject — the source must not be owned by, controlled by, financially entangled with, or substantially derived from the entity it is covering. A trade publication owned by a vendor is not independent of that vendor. A newspaper feature transparently rewritten from a press release is not independent. A "media partner" carrying client content under its masthead is not independent.
The second pillar is editorial oversight — there must be a real editorial process between the writer and publication: assignment, editing, fact-checking, corrections. Outlets that publish whatever a contributor uploads, with no gate, fail this test by definition. Self-published platforms (Medium, Substack as a platform, LinkedIn posts) fail it. Press-release wire services fail it. The presence of an "editor" in a masthead is necessary but not sufficient; what matters is whether that editor actually intervenes between submission and publication on the piece in question.
The third pillar is reputation for accuracy — the outlet must have a track record, demonstrated over time and acknowledged by other reliable outlets, of getting things right and issuing corrections when it gets them wrong. This is why young publications, even well-meaning ones, often sit in "no consensus" territory: they may be doing everything right, but they have not yet accumulated the reputational record that lets the community grade them. It is also why outlets that have visibly fabricated, retracted under pressure without correction, or repeatedly platformed disinformation get pushed down the ratings even if their writing is competent on the surface.
Why WP:V and WP:NOR form an inseparable triple with WP:RS
WP:RS does not exist alone. It is the operational guideline that implements two of Wikipedia's core content policies: WP:V (Wikipedia:Verifiability) and WP:NOR (Wikipedia:No original research). WP:V says that material added to Wikipedia must be attributable to a reliable, published source — not that it must be true in some absolute sense, but that a reader must be able to check it. WP:NOR says that Wikipedia articles must not contain analysis or synthesis that has not already been published in reliable sources — editors cannot connect dots, draw conclusions, or build arguments that the cited sources themselves did not draw.
"The threshold for inclusion in Wikipedia is verifiability, not truth—whether readers can check that material in Wikipedia has already been published by a reliable source."
That single line, paraphrased from WP:V, is the most consequential sentence in Wikipedia's content policy, and it is the one most outsiders fail to internalise. Wikipedia does not chase truth. It chases verifiable claims in reliable sources, and if those sources are wrong, the article will be wrong until the sources update. This is not a bug; it is a deliberate choice. Truth is contested, hard to verify at scale, and inviting an army of volunteer editors to determine it would destroy the project. Verifiability is checkable, neutral, and scales.
The practical consequence: when you walk into a sourcing review, you are not arguing that your claims are true. You are arguing that they are verifiable in reliable, independent, secondary sources. A founder who insists "but we really did invent it first" without a reliable-source citation for that claim is making a true-but-not-verifiable argument, and the policy is structurally indifferent to it. The corollary is liberating once you accept it: you do not have to convince anyone that you matter; you have to demonstrate that reliable sources have already covered the fact that you matter. The encyclopedia is downstream of the press, not upstream of it.
These three policies — WP:V, WP:NOR, WP:RS — interlock. WP:V demands sources; WP:RS defines which ones; WP:NOR forbids editors from filling gaps the sources did not. Together they enforce a single discipline: Wikipedia is a tertiary work assembled from secondary literature, and every claim that cannot be traced to qualifying sources must come out.
Layer 2 — The Four-Tier Map
Once the foundation is clear, the working tool is a tier map. The community does not use one official table — different reviewers carry slightly different mental versions — but the four-tier breakdown below tracks how citations are actually weighed in practice, from gold-standard down to citations that actively damage your draft by their presence.
Tier 1 — GOLD: international wires and the global broadsheets
The top tier is reserved for outlets whose editorial reputations are effectively settled across the English-language Wikipedia and most other major editions. The international wires — Reuters and the Associated Press — sit at the very top, because their entire business model is built around accuracy and corrections, and because their wire copy underwrites most of what other outlets republish. Below them, the global newspapers of record: the BBC (broadcast and online), Bloomberg, the Financial Times, the Wall Street Journal, the New York Times, the Economist, the Guardian, the Times of London. Major science journals (Nature, Science, The Lancet, NEJM) sit at the same tier for scientific claims.
What earns Tier 1 status is the combination of a deep editorial bench, an explicit corrections process, decades of WP:RSP discussion landing on "generally reliable," and a track record of being cited approvingly in deletion debates without anyone needing to argue the point. A single Tier 1 reported feature — a real piece of journalism, not a wire pickup or a contributor column — does more work on its own than three or four Tier 3 mentions stacked together. When a reviewer sees one of these mastheads, the default assumption is that the citation is good unless the piece itself is clearly an exception (an opinion column, a "sponsored" insert, a relay of a press release).
Tier 2 — SILVER: national broadsheets and high-editorial trades
Tier 2 covers outlets that are unambiguously reliable for most claims but lack the global weight of Tier 1. National broadsheets outside the English world — Le Monde, El País, Süddeutsche Zeitung, Frankfurter Allgemeine Zeitung, Corriere della Sera, the Sydney Morning Herald, the Globe and Mail — sit here for their home markets and for international stories they cover seriously. The major public broadcasters (Deutsche Welle, NHK World, France 24, CBC) sit at this tier when carrying their own reporting. Specialist outlets with strong editorial discipline — MIT Technology Review for its staff-written editorial, Wired for its features, Foreign Affairs, Foreign Policy, the Atlantic, the New Yorker for reported pieces (not personal essays) — qualify here as well.
A critical Tier 2 distinction belongs to Forbes staff journalism specifically: pieces bylined by employed Forbes staff writers, marked "Forbes Staff" or "Senior Writer," go through a real editorial process and are generally treated as reliable. They are not on the same tier as a Reuters investigation, but they count, and at WP:RSP they are explicitly distinguished from contributor pieces (which are toxic and covered in Tier 4). The same staff-vs-platform split applies to Bloomberg (staff: gold; Bloomberg Opinion: reliable for the columnist's view, not as a fact source), to Time (staff features: silver; opinion: attributed only), and to a long list of legacy mastheads with a clean editorial side and a problematic adjacent property.
Tier 3 — BRONZE: regional broadsheets, established trades, academic press
Tier 3 is the broad middle: outlets that pass WP:RS in principle but carry less weight than Tier 1–2, either because they are regional, niche, or have not accumulated enough discussion at RSP to be definitively rated. Regional broadsheets — a respected metropolitan daily, a strong national newspaper outside the global broadsheet club — sit here. Established trade press — Variety, the Hollywood Reporter, Adweek, AdAge, Billboard, Restaurant Business, the American Lawyer, Engineering News-Record — qualify, with the caveat that trade press tends toward boosterism for its industry and should be cross-referenced with a more critical outlet. University presses and reputable academic publishers belong here for non-peer-reviewed work; peer-reviewed journals from those publishers move up to Tier 1 for scientific claims.
The honest read on Tier 3 is that multiple Tier 3 sources, used together, can build a notability case, but stacking only Tier 3 invites scrutiny. The reviewer's quiet question is always whether the absence of Tier 1–2 coverage means the subject simply has not crossed into the bigger outlets' attention yet, or whether the trade-press coverage is essentially boosterism the bigger outlets have declined to pick up. The answer to that question often determines whether a draft survives.
Tier 4 — TOXIC: contributor networks, PR wires, paid placements
Tier 4 is the category we wish more buyers understood before commissioning press lists. These are not just "weak" sources — they are sources whose presence in a citation list actively damages a draft, because they signal to reviewers that the source base is thin and the writer is grasping. The headline categories: press-release wires (PR Newswire, Business Wire, GlobeNewswire, EIN Presswire) are explicit press releases with no independent reporting and never count toward notability; contributor networks (Forbes Contributor, HuffPost Contributor, Entrepreneur Contributor, Inc. Contributor, Fast Company Contributor) are self-published columns wearing a trusted logo and are explicitly treated as self-published at RSP; corporate blogs and own-media (your own newsroom, the vendor's case study about you, your CEO's Medium) are not independent.
Medium as a platform is generally unreliable regardless of writer, because Medium is publishing infrastructure, not an editorial outlet — there is no gate between the writer and publication. LinkedIn articles and posts are self-published in the same sense. Sponsored content marked or unmarked — "branded content," "partner posts," "in association with," "promoted" — is paid placement and disqualifying. Pay-to-publish portals, especially the swarm of "as featured in" outlets that take a fee for inclusion, do not count and never have. When we see a draft whose strongest "Tier 1" claim is a Forbes contributor column written by the founder, or a "Business Insider feature" that turns out to be a syndicated press release, we know the audit has not yet been done.
Layer 3 — The WP:RSP Deep Dive
Wikipedia's Perennial Sources list, shortcut WP:RSP (Wikipedia:Reliable sources/Perennial sources), is the closest thing the community has to a precedent database. It is a maintained table of outlets that come up repeatedly in disputes, each tagged with a community-consensus rating and linked back to the underlying discussions. Reviewers reach for it constantly, and any agency or in-house team taking Wikipedia seriously should be reading it directly rather than trusting summaries.
The four RSP classifications and what each one means in practice
RSP sorts entries into four formal bands. Generally reliable means the community has, through repeated discussion, concluded that the outlet has the editorial process, reputation, and track record to support most claims; reviewers treat citations from such outlets as good by default. No consensus means the community has discussed the outlet enough to note it but has not reached a stable view — usually because the outlet is good in some areas and weak in others, or because discussions land on "case by case." These are usable but expect scrutiny. Generally unreliable means the community has agreed the outlet should not be used for factual claims, though it may still be acceptable for the outlet's own views or for uncontroversial bibliographic detail. Deprecated is the strongest negative classification: the outlet has been through a formal community request for comment (RfC) and has been ruled presumptively unusable, often with an edit filter that warns editors trying to cite it.
Seven RSP discussions worth knowing
Forbes. The single most important RSP entry for business subjects. Forbes' staff-written content is generally reliable; the Forbes contributor network is generally unreliable and explicitly treated as self-published. The first thing a reviewer does with any forbes.com URL is check whether the byline reads "Forbes Staff" or "Contributor," and the discount is total: a contributor piece counts for as much as a personal blog.
HuffPost. Identical structural split. HuffPost shut down its US contributor program in 2018 after years of policy abuse — that decision is itself a useful data point in deletion discussions. Pre-2018 contributor pieces remain unreliable; post-2018 HuffPost staff content is generally reliable, with the usual caveat that the outlet leans toward opinion in some sections.
Medium. Always unreliable for notability purposes. Medium is publishing infrastructure, not an outlet, and RSP says so explicitly. The fact that a piece is on Medium tells you nothing about whether it was edited, fact-checked, or independent of the subject. Some genuinely reliable writers self-publish on Medium, but Wikipedia evaluates the platform's reliability, not the writer's individual reputation — which means even a well-known journalist's Medium post is treated as self-published.
Daily Mail. Deprecated in 2017 after a formal RfC, the first major mainstream newspaper to be deprecated on English Wikipedia. Reaffirmed in subsequent discussions. Attempts to cite dailymail.co.uk trigger a filter warning. Even flattering Daily Mail coverage of a subject is treated as zero evidence of notability and is actively stripped from articles.
The Sun / RT / Sputnik / Press TV / Breitbart. All deprecated. The Sun for fact-checking record; RT, Sputnik, and Press TV as state-controlled propaganda outlets; Breitbart for a track record of fabrication and platform-level disinformation. A draft leaning on any of these is a draft that will be rewritten or deleted.
Vice / Vice News. A live example of how RSP entries evolve. Vice was treated as generally reliable through much of its run, particularly its news arm. As the company's editorial situation deteriorated through 2022–2023 and beyond, the RSP discussion shifted, and newer Vice content is now treated with more skepticism. The lesson: an outlet's rating is not permanent, and a citation from a year a publication was solid is not the same as one from a year it was collapsing.
Substack. RSP treats Substack as a platform, not a publisher — meaning the reliability of a Substack newsletter depends on the reputation of the writer as established by their own coverage in other reliable sources. A Substack written by a Pulitzer-winning investigative journalist may carry weight; a Substack written by an anonymous figure does not. This is similar to the WP:SPS rule for self-published expert sources: usable only when the writer is an established subject-matter expert per independent reliable sources.
The deeper takeaway from working through RSP: it is not a list to memorise; it is a habit. Before any source goes into a draft, the URL should be checked against the current state of the RSP entry. Before any "as featured in" claim is taken at face value, the byline should be checked. Before any deprecated outlet is cited "because it's positive coverage," the writer should know the citation will be stripped on sight. RSP is the single highest-leverage reading anyone in this field can do.
Layer 4 — Edge Cases by Discipline
WP:RS is general by design, but real-world drafts run into specific situations where the general rules need to be specialised. The following eight categories cover the edge cases that come up most often.
Academic preprints (arXiv, SSRN, bioRxiv)
Preprints — papers posted to arXiv, SSRN, bioRxiv, ChemRxiv, and similar servers before peer review — are generally not considered reliable sources on Wikipedia. The reasoning is that the entire point of peer review is the editorial process that WP:RS requires; until a paper has cleared peer review and been published in a refereed journal, it has not had editorial scrutiny in the policy-relevant sense. Exceptions exist for established authors discussing established results, but a preprint is not safe ground, and a draft built on preprints will be challenged.
Conference proceedings
Highly field-dependent. In computer science, peer-reviewed conferences (ACM, IEEE, NeurIPS, ICML, CHI, SIGGRAPH) are often the primary publication venue and are treated as reliable on par with journals. In engineering and parts of physics, the situation is similar. In medicine, humanities, and most of the social sciences, conference proceedings carry much less weight and are not treated as reliable secondary sources for notability purposes — abstracts are quick-turn and rarely independently editorially reviewed.
Opinion and editorial
Op-eds, editorials, and signed columns are reliable for the opinion of the writer or the editorial board — attributed appropriately — but they are not reliable for factual claims about the subject of the opinion. A New York Times op-ed praising your CEO is a citation for "X argued in the NYT that…," not for any factual claim contained in the argument. Editorial-page material is also weaker for establishing notability than reported coverage.
Self-published expert sources (WP:SPS)
WP:SPS carves out a narrow exception for self-published material only when the writer is an established subject-matter expert whose work in that field has been published by independent reliable sources. A computer scientist with peer-reviewed work on a topic can be cited from their personal blog on that same topic; a marketing executive cannot be cited from theirs on whether their company is industry-leading. The expertise must come first, demonstrated by external reliable sources, before the self-published material can be used.
Industry analyst reports (Gartner, Forrester, IDC, Frost & Sullivan)
Generally reliable as primary sources for the analyst's own view and methodology. They are usable to support a discrete factual claim ("Gartner ranked X in the Leaders quadrant of its 2025 Magic Quadrant for…") but are weak for establishing notability on their own, because the analyst's view is one source, not multiple independent views, and because the analyst-vendor commercial relationship is widely understood to colour coverage. Multiple analysts converging on a finding is stronger than one.
Court records and government filings
These are primary sources — verbatim records of proceedings, filings, regulatory actions. They are usable for facts ("Company X was fined €Y by the Z regulator in 2024"), but they cannot be used as the basis for analytic claims ("Company X is widely considered a leader in compliance") that the records themselves do not make. A Wikipedia article on a company can cite a regulatory filing for a numeric fact, but the article's framing of why that fact matters must come from secondary reliable sources discussing the filing.
Social media (X, LinkedIn posts, YouTube, TikTok)
Generally not reliable for third-party claims. A tweet from a journalist asserting something about your company is not citable; the journalist's subsequent reported article on the same topic is. The narrow exception: social media posts are usable as self-published evidence about the poster themselves — "the company stated on its official X account that…" — under WP:SPS, when the account is verified and the claim is not contentious. They never establish notability.
Press releases and corporate announcements
Never count toward notability. A press release is the subject talking about itself. It can be cited for narrow factual claims about the subject — incorporation date, headquarters location, leadership names — when those claims are attributed and uncontroversial, but the moment a claim is contested or interpretive, a press release cannot support it. Reprints of press releases by news outlets ("churnalism") inherit the press release's primary character, no matter what masthead is on top.
Layer 5 — How Reviewers Actually Score Sources
Beyond the formal policies, Articles for Creation (AfC) reviewers and deletion-discussion participants apply a handful of unwritten conventions that determine which drafts survive. These are not documented in a single policy page; they are the working norms that emerge from years of reviewer practice. Knowing them is the difference between drafts that read as "policy-compliant" and drafts that read as "this person has done this before."
The three-sources rule (and why it is asymmetric)
The de-facto threshold for a defensible company or biography article is at least three independent, reliable, secondary sources providing significant coverage. This is the operational reading of the general notability guideline (WP:GNG) and of the corporate-notability guideline (WP:NCORP) as reviewers actually apply them. But the rule is asymmetric in a way the simple count obscures: one Tier 1 piece of reported journalism is worth more than three Tier 3 pieces stacked together. A Reuters investigation, a Financial Times feature, a Bloomberg profile — any one of these establishes the kind of independent, substantive coverage the rule is reaching for. Three regional-trade-press mentions, even if technically passing, signal that the subject has not yet been picked up by the outlets that anchor the notability case.
The other side of the asymmetry: lots of Tier 4 citations cannot rescue a thin Tier 1–3 base. Reviewers do not add citations in a column and pass when the total clears some threshold. They look for the strongest two or three sources and ask whether those alone make the case. If they do not, the draft fails no matter how many press-release wires and contributor columns are stacked beneath them.
The trivial-coverage filter
Significant coverage is the part of the GNG that is most often glossed and most often determinative at AfC. A passing mention — a list of attendees, a quote in an industry round-up, a name in a paragraph that is mostly about somebody else — does not contribute to notability, no matter how reliable the outlet. The unwritten test reviewers apply: can you write several encyclopedic sentences from this single source alone? If a citation can support only "X was mentioned in a Reuters piece on…", it is trivial coverage and does not count. If it can support a substantive paragraph about the subject, it qualifies.
The time-decay pattern
A single concentrated news cycle from years ago is weaker than sustained coverage over recent months. Reviewers internalise the time pattern of coverage when they read source lists: a fundraising announcement covered by ten outlets on a single day five years ago, with no follow-up since, reads as a brief news spike and is discounted accordingly. Eighteen months of sustained coverage — multiple independent pieces spread across time — reads as evidence the subject is genuinely of ongoing interest to the press, and that is what notability is meant to capture. This is also why a recent surge in coverage, without the historical depth, often produces "wait and see" outcomes at AfC.
The syndication trap
One wire story reprinted across fifty outlets is still one source. This collapses press lists faster than any other rule. A Reuters or AP story picked up by Yahoo News, MSN, dozens of regional papers, and a swarm of aggregator sites looks like a media blitz on a press list. To a reviewer, it is one citation point — the original wire — and the rest are reprints. The same applies to translated reprints, to syndication networks (Bloomberg, Bloomberg Línea, the various Bloomberg licensees), and to chains of agency placements. The diagnostic is brutal: when two "different" pieces share most of their paragraphs, they are one piece.
WP:CORPDEPTH and routine business news
For companies specifically, WP:NCORP requires more than routine business news to establish notability. Funding announcements, executive hires, product launches, acquisitions, quarterly results, and partnership announcements are explicitly identified in WP:NCORP and surrounding guidance as routine coverage that does not, on its own, support notability — because every company has these, and the press covers them by default. The notability question is whether the company has received the kind of in-depth, analytic, independent coverage that goes beyond announcement-and-relay. A profile that takes the company seriously as a subject in its own right; an investigation; a sustained analytic piece on its market position; a critical assessment of its strategy or impact. These count where funding-round write-ups do not.
The compounded effect is sobering. A company with "twenty press hits" in the last year may, once you strip out routine announcements, contributor placements, syndication duplicates, and passing mentions, be left with two or three citations of substance — and that is what determines whether the draft survives. Reviewers do this stripping by reflex. Anyone preparing a draft should do it first.
How we use this at WikiBusines
We start every project — without exception — with a source audit against the framework above. Before a draft is written, before a single talk-page disclosure is filed, before any budget is committed beyond the audit itself, the entire coverage record is pulled, deduplicated, and rated layer by layer. Each candidate citation is checked against WP:RSP, sorted into a tier, tested for independence under WP:ORGIND, classified as primary or secondary, and assessed for substantiveness against WP:CORPDEPTH. The output is a written verdict per source with reasoning attached.
If the verdict is that the source base is below a defensible three-Tier-1-equivalent threshold, we say so up front and decline the page commission. This is not a sales tactic; it is the only honest position. When a draft is built on a source base that will not survive AfC, no amount of writing craft rescues it, and the cost — financial and reputational — of pursuing it anyway falls on the client. The alternative paths we recommend in that situation: our Source Readiness Program to close the coverage gap through legitimate earned media before attempting a page; a Wikidata entity (lower notability bar, no AfC, still feeds the Knowledge Graph and AI retrieval); a Simple English Wikipedia article where the case is borderline; or a draft in another language edition where the subject's coverage is stronger.
When the source base does support a defensible draft, we use the audit as the working document for the page itself — every claim sourced to a rated citation, every uncertainty flagged before it becomes a reviewer's objection. The audit also becomes the internal record that protects the client: if leadership asks why a page did or did not launch, the audit answers the question on a single sheet, with reasoning per source.
For a full read on adjacent terrain, see the Wikipedia notability audit service (the formal audit deliverable), our guarantees page (what we will and will not promise, in the policy language we have used throughout), Wikipedia page creation (the writing-and-review workflow once the source base passes), and two related deeper reads: How to spot a fake Wikipedia agency (the warning signs that someone is not running this kind of audit) and Wikipedia COI policy explained (the disclosure framework that runs alongside the sourcing one). The layered approach in this guide is the discipline we apply on every project, and the language above is the language we expect any serious partner in this field to use.