The Articled Zero AI Guarantee
Most statements claiming AI absence actually refer to a detector score. This statement concerns authorship. A score of zero implies no generated text, no generated outline, and no generated paraphrasing, and the definition is meant to provoke argument.
What the Zero AI Guarantee Means
Since zero is a count, it should have a unit. We use the sentence. If a language model generated a sentence, that sentence does not appear in your document, paraphrased, shortened, or used as a scaffold that a human writer edits until it resembles something else. This count also considers headings, captions, bullet fragments, alt text, and the image meta description.
The unit preceding the sentence also counts. An outline determines what is said and in what order, so a generated outline that is subsequently filled in by a human is a generated document with human decoration. The same applies to a generated research summary, a generated set of subheadings, and generated questions used to organize an interview.
Tools that do not generate text are not included in the ban, and pretending they are would be theatrical. Writers are free to use spell-checkers, dictionaries, search engines, library databases, and citation management systems. For the same reason, machine transcription of an interview the writer recorded is fine, provided the writer compares it to the audio before quoting.
- A generated sentence, however heavily edited afterward.
- A generated outline that was completed by a human.
- A generated summary of a source that the writer has not read.
- A generated headline, caption, subheading, or meta description.
- A paraphrase of some generated text, made possible with any tool.
The unit is the sentence
Zero is quantifiable because we have given zero a unit. One artificially intelligent (AI) generated sentence anywhere in this file voids that agreement. Invoking the remedy occurs with breaking the agreement rather than a dialog regarding degree.
How the Zero AI Guarantee Is Enforced
The majority of the work on a brief happens before an assignment is made. Writing samples for which employees are interviewed and for which employees are paid to write a sample and talk through how they researched it take roughly an hour to complete and are difficult to fake. Many applicants end it here. The ones that continue submit the policy and sign it along with a name on every document they are asked to bring.
After that it is editorial. Every draft is read against its sources before a tool touches it, and the editor is hunting for the specific failures of the generated prose: claims that the cited source does not support, a paragraph that restates the previous one in greater detail, and precision that was never in the reporting to begin with. Then the twelve-tool panel runs and the reports go out with the file.
We also perform a random audit. Each month, some files go through a secondary audit. This involves the editor reviewing the writer’s notes and drafts along with raw and finished copies, regardless of whether anything seemed wrong at the time. Audits that target a specific file based on a complaint tend to catch the careless writers and miss the careful ones, which is the opposite of what is intended.
- Interview, paid sample, and a discussion on methods.
- Policy signed before the first brief, name on every file.
- An editor reads the draft and references each source it cites.
- Twelve detection and originality checks, reports sent to you.
- Randomized audits of files submitted against their notes and drafts.
Why the Zero AI Guarantee Goes Beyond AI Detection
Detectors rate the text, not the process, and therefore, tend to reward unoriginality. Formulating more complex responses only improves the likelihood of a response being detected as being human (but can still end up being detected as a machine response). Conversely, simple, monotonic responses are more likely to be detected as being a response generated by a machine. Detector fail safes are necessary because both types of errors happen frequently. This is why we cannot rely on the number alone.
The gap runs the other way as well, and that is the reason why humanizer tools exist. Anyone can push generated text through a rewriter until the score drops, so a supplier who only offers a low detector reading is making a promise you could buy for a few dollars a month. That is a promise about a score. It is not a promise about authorship.
The guarantee is based on the chain. It includes the name of the writer, the editor who checked the sources, a record of employment, a request for which we can contact any time, and the remedy for the chain failure. Detector reports are supporting evidence and we submit all twelve, even if some come back looking a little strange. They affirm the guarantee. They are not the guarantee.
- Whether the writer researched the topic or invented it.
- Whether generated text was pushed through a humanizer first.
- Whether a genuinely human draft was flagged by mistake.
- Whether the sources cited actually support the claims made.
- Who wrote the file, or whether anybody can still be asked.
False positives are real
We send the record of the work we did instead of rewriting the file if your human writing gets flagged for a simple, direct writing style that is technical in format. We will wait for a tool to give us more flexibility.
Zero AI Guarantee FAQs
No. When you click the “rewrite” button in their editing tool, they begin editing sentences by rewriting, expanding, condensing, and restyling them. For our purposes, it behaves like a generation system that would break the 50 count. spelling and grammar checks should remain allowed because they contribute nothing to the submission.
Yes. The first proof is included without request. You receive twelve reports attached to your delivery email. Beyond the first proof, you are able to request the writer’s notes, previous drafts, a typing provenance record, or a brief call with the writer at any time within the 14-day revision period.
The guarantee that we have ensured is the promise and the remedy that is connected to it. Zero has no right to become mostly silent through this definition written down on this page. This is the page that you need to copy if you are preparing to draft a supplier clause.
Not for anything that reaches your file. A writer may use the translation tool as an evaluation method to determine whether a source is worth pursuing, but a translated quote must be evaluated by someone who reads that language. If no one does, the quote is dropped rather than guessed.
A keyword tool that reports search volumes is data, and it remains permissible. A tool that provides finished headlines is composing, and it is therefore out. The rule has nothing to do with which vendor builds the tool and has everything to do with whether a sentence is provided back.
Yes. The act of reading is involved in writing, which is one reason delivery takes up to three days rather than three minutes. We read, name, and store sources, and our separate no-AI research policy outlines how writers deal with materials they cannot verify.
The strict version of this promise is inconvenient for us. It rules out shortcuts that would be invisible to you, and it means turning down work that comes with a generated draft and a request to clean it up. We prefer to publish definitions that you can hold us to rather than statements that you are required to believe.
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.