The Articled No-AI Policy
This rule, written down so it can be used against us: a generative tool should not touch any Articled file in any stage. This page states the parameters, the people bound by this rule, and the consequence of breaking this rule.
What the No-AI Policy Prohibits
There is a prohibition on drafting and not just on generative AI. A writer cannot ask a model for an outline or first line or a transition, a variant of a headline, a meta description, a summary of a paper they have not read, or a translated quote. If a language model wrote the words, they cannot go into your file, even if it is just a fragment of a draft or a single sentence that was copied.
People overlook the reason research is banned. Because chatbot summaries of literature can’t tell you where they came from, a writer who starts there builds on something no one can check. Our writers read the primary source, cite it in the file, and include the link. A claim that cannot be sourced comes from the draft.
This proposal covers rewriting and paraphrasing tools. Machine text goes through a rewriter and ends up with machine text with a new vocabulary, which is a rearrangement of words meant to address the purpose of this rule. If it is sold as a study aid, a grammar assistant, or a productivity tool tucked inside the paid editing software, then the ban is applicable.
- Model-written drafts, sections, intros, headlines or meta descriptions.
- AI outlines and content plans handed to a person to fill in.
- Chatbot research summaries standing in for reading the source.
- Using machine translations for quotes, interviews, research.
- Any text edited with the so-called humanizing tools, paraphrasers or spinners.
- Any editing tools and browser extensions or plugins that are equipped with rewrite features.
Where the line sits
Generative tools require an AI as the writer does not use a dictionary, spell checker, search engine, or citation manager when composing a sentence. Our question is straightforward: did software write this sentence, or was it a human? Note: The assigned tools include a dictionary, spell-checker, search engine, and citation manager. A style guide detects plagiarism.
How the No-AI Policy Applies to Every Writer
Writers sign this policy before they get their first brief and it is a term of engagement rather than a suggestion in a handbook. It runs across the entire file: notes, outlines, drafts, revisions, and the research summary that comes with it. It also travels with a brief so assigning an Articled job to someone else is not a possibility, no matter what.
Editors do the same things. Anyone who can’t fix a clumsy paragraph by hand is not fit to edit, so there is no quiet exception to polish a paragraph. There is no exception for length. The same term applies to a 100-word product blurb as a 5,000-word report. The greatest temptation for a shortcut is on the work that appears trivial and insignificant.
This is the final restriction expressed simply. We would never put monitoring software on a writer’s machine and would never engage with a writer that said they would let us do that. Instead, we have signed terms, an editor that double checks every submission with its source, twelve checking tools and a right to request any writer’s records when we want.
- Signed before the first brief of work, and reaffirmed with each new assignment.
- Covers notes, outlines, drafts, revisions, and research summaries.
- No subcontracting, no ghost-editing, no carve-out for short pieces.
- The same applies to editors and reviewers.
How We Enforce the No-AI Policy
BEFORE any software runs, an editor checks the draft against the brief and each source it cites. Editorial issues tend to be the most important. Examples include sources cited that do not support the attached sentence or a confident statement with no accompanying source, and arguments that are repeated several times over three (or more) paragraphs. These issues are missed by detection software. However, a reader who understands the sources will not miss these.
We then check the draft with Originality.ai, GPTZero, Turnitin, Copyleaks, Winston AI, ZeroGPT, Sapling, Content at Scale, the Writer.com detector, Crossplag, and Grammarly Authorship. For duplication, we use Copyscape. We send you those reports with the delivery email. If one of the reports shows a high reading, that by itself would not end your career, as these tools will give false positives on genuine human writing. However that does start the file.
An open file implies that the writer is being asked for their working record, which includes their notes, earlier drafts, document or browser history, and an explanation of their reasoning for each section. Writers who can justify their text can remain, and we inform the client in writing. Writers who cannot justify their text are let go from Articled immediately, with no notice period and no second brief.
- A flag is raised by an editor, a detector reading or a client.
- Delivery is paused when the writer is asked to provide notes and drafts.
- A confirmed breach permanently ends that engagement.
- You can either choose a new draft from a different writer or opt for a full refund.
- Copyright in anything already delivered still transfers to you.
A policy, not legal advice
This page explains how we operate. If you are using our clauses as a template for your supplier contracts, get someone with legal knowledge to review your version to see if it conforms to your jurisdiction before using them.
No-AI Policy FAQs
No. Both drafting and rewriting fall under the umbrella of generation, and both are completely banned. This includes a full first draft, a single rewritten paragraph, a rewritten headline, and any suggestion from a generative editor that is accepted. A writer who needs a paragraph rewritten must do so by themselves or have an editor do it.
The delivery gets held back as we request notes, drafts, and sources from the writer. If confirmed, the breach takes effect immediately and will result in no further briefs. You will then have to choose between an entirely new draft, on us, or a full refund.
And the difference matters. Sending us AI-drafted briefs, existing pages, or quotes from a chatbot would mean we will have to write around what you send us. Without any assurance of authorship, we cannot guarantee that we have not plagiarized. Simply notify your editor when you place your order, and they will highlight sections augured for you.
This page is the policy. We do not have a policy document behind a sign up wall, nor do we create accounts. Nothing changes after you pay. If procurement requests the same terms within a supplier agreement, send a request to hello@articled.org. Someone will email you the agreement for signing.
With each order. All twelve reports will be sent in the delivery email, including the Copyscape result. If you receive a report that seems unusual, reply to that email and an editor will provide an explanation of what the measurement was and the reason why we sent the file.
As with everything else. AI shortcuts feel the most harmless when solving small and mechanical tasks, which is why no exceptions were made for them. A proofread is a person reading and the two revision rounds included with your order follow the same principle.
First of all, publishing a policy is inexpensive, but OPEN PAID costs time and money. This one costs us speed of delivery, since a person needs up to three days, and a model only needs seconds. It has cost us writers we enjoyed working with. However, we believe the trade-off is justified and the reports you see in your inbox are the trade you are checking.
Content a person actually wrote
$10 per 100 words, and the writer keeps all of it. Our 1% sits on top, 0.5% goes to trees, and no generated text appears anywhere in the process.