Our Publishing Policy

Publishing Policy — publishingpolicy.org Operated by Roarke Clinton, with plans to incorporate as a 501(c)(3) nonprofit. Contact: team@publishingpolicy.org

We built a service that asks every publisher to put its standards on the public record. So we put ours here first, written in our own tool. We hold ourselves to the standard we ask of everyone else — and because we coined the word for breaking it, we accept being held to that word at least as strictly as anyone we display.

A publishing policy is a public document where an organization states the standards that govern how it creates, verifies, and distributes information — verification processes, content standards, corrections, AI disclosure, transparency, and accountability. Malpublishing is failing to live up to the commitments you make in your publishing policy. We wrote that definition. So the moment we state a commitment below, we have also defined exactly what would count as us malpublishing.

That is why this document is short. We kept it to promises a small team can actually keep today — because an over-promise we cannot honor is not a stronger policy, it is a self-inflicted one. Where something is a goal rather than a practice we already keep, we say so.

This policy covers our publishing only. It is the third side of the trust triangle we describe on /about: our Terms of Service set the legal relationship, our Privacy Policy governs your data, and this governs what and how we publish. Where those documents govern, they control.

What we publish, and where this applies

We publish this website and its marketing copy, our /about and /ai pages, our documentation, our public /llms.txt file, the example policies and templates we seed, anything we post on other platforms, and this policy.

This policy binds all of it. We do not keep a lower standard of honesty for a social post or a marketing headline than for a formal page. When we make a factual claim about the platform anywhere — including on social media — we hold it to the same accuracy standard, and we correct it if it turns out wrong.

Accuracy: how we handle facts

We check our claims about the platform — what it does, what data it collects, what runs on the site — against the current code and infrastructure. When we find a claim that is wrong or has gone stale, we correct it and log the correction. We would rather say "we're not sure" than fill a gap with confidence.

We keep facts separate from hopes. When we state an opinion, or a goal we have not reached yet, we say so, so you can tell the two apart. We credit and link the sources we rely on.

How we use AI in our own content

We use AI tools to help draft and edit our content, including parts of this policy. We tell you that here rather than hide it. Our rule: AI never has the final say, and no AI-assisted claim is published until a named human — right now, Roarke Clinton — has reviewed it and stands behind it. We never pass AI-written text off as the independent work of a real person.

We publish a framework at /ai that asks any AI publisher to pick a setting in five categories. We ran it on ourselves. Here are the settings we chose, and the standard we hold our own content to:

1. Accuracy & uncertainty → A. Always verify before asserting. If we're not sure, we say so.
2. Perspective & opinion → A. Explicitly label all. We mark opinions and goals as opinions and goals, and keep them apart from verified facts.
3. Attribution & sources → B. Cite when it matters for trust. We name our sources where the claim depends on them; we don't footnote common knowledge.
4. Content labeling → A. Always label content type. We label examples as examples and drafts as drafts.
5. Corrections & accountability → A. Proactively flag and correct. When we're wrong, we say so and fix it, as described below.

We picked B, not A, on attribution on purpose. A citation on every ordinary sentence would be theater, and we would rather tell you the truth about our standard than perform a perfect one.

What would count as us malpublishing

We did not have to guess at this. Our own platform turns commitments like the ones above into their matching failure definitions — the same way it does for every policy on the site. From commitments like ours, our engine generates, word for word:

• Publishing AI-generated or AI-assisted content without disclosing the use of AI.
• Publishing content without clearly labeling its type (fact, opinion, sponsored, etc.).
• Publishing content based on others' work without naming and linking to original sources.

This is the list you can hold us to. If you catch us doing one of these, we are malpublishing, by our own machine's definition.

Corrections and our changelog

This policy is a versioned document published on our own platform. You can see the current text, every past version, and when each version changed, on this policy's page. We do not edit it silently, and we never delete a past version to hide what we used to say — exactly what we ask of every publisher.

When we get something wrong in anything we publish:

• We acknowledge a correction request within five business days. If we are traveling and have posted an away notice, that clock starts when we are back. Either way, we will never answer a good-faith report with silence.
• We correct a confirmed factual error within seven days of confirming it.
• We log the correction in this policy's public history, saying what changed — never as a silent fix.

A change we are making in the open: our earlier auto-generated example policy promised corrections within 48 hours. This document replaces that with acknowledge-in-five-business-days and correct-in-seven, because a solo team that sometimes travels off-grid can actually keep the split version, and a promise we can keep is worth more than a tighter one we cannot. We are showing you the change instead of quietly making it — which is the whole point of this product.

Independence and commercial relationships

We are self-funded and the service is free. We do not run ads, we do not accept sponsorship, and no one can pay to be featured, ranked, certified, or given better standing. No payment ever changes how we define or apply "malpublishing." If any of that changes, we will disclose it here first, before it takes effect.

We disclose our own conflicts plainly, because we ask everyone else to:

• The examples are ours. The example policies on the platform — for useractivity.ai, roarke.io, onism.org, onistfoundation.org, and thriveguides.com — are the founder's own sites. They are real, but they are not independent third parties, and we label them as examples, not endorsements.
• Some of the tools are ours. Our analytics (UserActivity) and our feedback widget (Communications) are also the founder's products.
• We are the referee and a player. We coined "malpublishing," we run the platform that shows who follows their policy, and we publish our own policy on it. That is a real conflict. Our answer: we publish on the same public, versioned, CC-BY terms as everyone else, we accept being held to our own definition as strictly as anyone we display, and we make ourselves easy to report.

What we refuse to publish

Bright lines. We will not:

• Fabricate testimonials, user quotes, usage numbers, adoption counts, or case studies, or present an invented persona as a real person.
• Invent a policy for anyone. Organizations that have not written one are shown honestly as having no known policy — never with a standard we made up for them.
• Sell placement, ranking, a certification tier, or better standing.
• Dress up a score as a rating. Our readability score measures how easy a policy is to read. It is not a certification, a quality grade, or a truth check, and we will never present it as one.
• Silently change a published policy — ours, or anyone's on the platform. The version history shows every edit.
• Publish an AI-written claim that no human has reviewed and stands behind.
• Use pressure tactics or dark patterns to get sign-ups.

What we do not do (so you are not misled)

We are a place to disclose standards. We are not a judge of whether they are good, and not a judge of whether they are true.

• We host and display self-reported policies. We do not verify, certify, or guarantee that any policy — including our own — is accurate or actually followed. We are not a regulator, an accreditor, or a compliance auditor.
• Our certification tiers (Declared, Committed, Verified, Exemplary) measure structural facts: is a policy published, how long has it stood, is the domain verified, is there a public correction log and a recent review. They do not measure whether a policy's claims are true.
• Community reports and feedback reflect what readers perceive, not our judgment.

The value here is the public commitment and the accountability that transparency creates — not a stamp of approval from us.

Data and privacy

How we handle your data is governed by our Privacy Policy. The promise most relevant here: we do not run ads and we do not sell your personal data, so nothing you read on this site is shaped by an advertiser. If this short note and the Privacy Policy ever disagree, the Privacy Policy controls.

How to hold us accountable

If you think we have failed this policy — that we are malpublishing — tell us.

• Email team@publishingpolicy.org, or use the feedback widget on any page.
• Once our public reporting feature is live, you will be able to file a report against us on our own platform, the same way you could against anyone else. We intend to be the first organization reported there — against ourselves.
• We will acknowledge your report on the timeline above, and respond in the open: either a correction logged in this policy's history, or a plain explanation of why we believe we met the commitment.

A real person is accountable for what we publish: Roarke Clinton, operator of Publishing Policy (a personal project, with plans to incorporate as a 501(c)(3) nonprofit).

This policy is itself on the record

This document is published on our own platform, so it carries the same proof we offer every publisher:

• a stable public URL and a public version history — every past version, and when it changed — so we cannot quietly walk it back;
• a CC-BY 4.0 license and a timestamp, so anyone may quote it;
• a readability score, computed live by the same code that scores every other policy — we do not hand-tune our own grade. We commit to keeping this policy in plain or moderately complex language. If a future version ever reads as "Dense," we have failed our own premise, and rewriting it is a correction we owe you;
• a certification tier, computed by the same rules as everyone else's. Right now our software computes this policy as Declared — it is newly published, so that is all it can be. The higher tiers are earned, not asserted: Committed after 90 days live with at least one revision, Verified once our domain is verified and the policy is linked to it, and Exemplary only once a public correction log and a documented yearly review both exist. We show whatever tier the platform computes — you can watch it climb, or stall, in public.

By our own three tests, described on /about: this document is self-defined (we chose every line; no regulator made us), publicly stated (not an internal memo, not our Terms), and independently citable (licensed, timestamped, versioned). We built the test. This is us taking it.

We review this policy at least once a year (our earlier example said quarterly; yearly plus on-material-change is a cadence we will actually hit), and we revisit it whenever the platform changes in a way that affects what is written here.

Preview note: the platform is in Research Preview. When it reaches general availability, every published policy — including this one — is automatically unpublished to a private draft and must be consciously re-published. We will re-publish this one deliberately, so our own example never quietly disappears.

Version 1.0 · drafted 30 June 2026. The published version is dated and timestamped automatically when it goes live, and every change after that is recorded in this policy's public version history.