Verified With GPTZero, Originality.ai and Multiple AI Detectors
Twelve tools, run by an editor rather than by the writer, on the final draft. Reports by the detection system and why one detector is never enough. The promise we refuse to make regarding the scores.
How Verified by AI Detectors Fits the Articled Process
Detection runs last. This positioning is intentional. The checks are not a filter that generated text must get around because no generated text is introduced at any point in the process. They are a check to determine if our claim is valid in the tools used by our clients.
An editor uses the tools; a writer does not. The person whose interest is at stake is not the one conducting the test. Results are not ignored and are referenced against the order. This helps us evaluate a writer’s pattern across twenty pieces rather than one.
The scope of the question is much smaller than it appears. We know who wrote your draft because we commissioned the writer. What we’re examining is whether a reader who copies the draft to GPTZero in the coming week will find something that will raise a concern for them. If there is a concern, we want to know whether it is due to the draft, the tool, or the way the tool interprets this particular type of writing.
- Brief received, quoted, and assigned to a writer who knows the format.
- Writers conduct their own research before drafting and performing post-draft read-through.
- Editors read drafts in the context of both the brief and the source lists.
- Editors then run the detection and similarity tools once the text is drafted.
- Both reports and drafts are sent out in the same email.
Why Verified by AI Detectors Uses More Than One Tool
Send the same three paragraphs to five detectors and five different answers come back. They were trained on different material, set their thresholds differently, and are trained on a schedule that the vendors do not disclose. A measurement in March with one particular tool is not the same measurement that would be made in October with that same tool.
Results from multiple tools created by independent groups are more informative than a strong result from a single tool. For example, if eleven checks consider a draft to be at a human level and one check flags it, we will treat the check as an outlier. However, if the majority of them flag it, we will consider it worthy of a real conversation with the writer, and that conversation will happen before we send you the file.
Here is what stacking tools cannot solve. Their errors are intercorrelated. Most of them misread the same sorts of writing. So running twelve does not eliminate an embedded bias against bland procedural writing or against writers whose first language is not English. More tools diminish random noise. They do not turn a heap of estimates into a proof.
- AI detection: GPTZero, Winston AI, ZeroGPT, Sapling, Content at Scale, Writer.com AI Detector, Crossplag.
- AI detection with plagiarism: Originality.ai and Copyleaks.
- Similarity against an academic index: Turnitin.
- Authorship and writing-session provenance: Grammarly Authorship.
- Duplicate content across the open internet: Copyscape.
Twelve opinions, not twelve proofs
Every tool listed is capable of being incorrect about the same paragraph in the same manner. We state what each tool said independently, where they disagreed, and we do not average them into a verdict.
What Verified by AI Detectors Means for Customers
The supplied draft report files come labeled by tool and dated in a single file in your email. As for the content, nothing has been consolidated to a single figure, and nothing has been withheld because it read poorly. You have the option to send them to your client, attach them to a compliance file, or just discard them since the copyright is yours upon receipt.
What we don’t get with the label is a promised number. We cannot guarantee a draft scoring below some number on your preferred tool, because your settings, that vendor’s next upgrade, and your own copy edits all are beyond our control. What we can guarantee is the name of the author and the fact that we will show our work.
Consider check discrepancies standard events. In the event of a check discrepancy, provide the name of the tool, score, and the relevant passage. An editor will reopen the file, and will send notes, sources, and draft history along with reruns of the checks. You have two rounds of revisions for fourteen business days following the delivery. These revisions are backed by our money-back guarantee policy.
- That a report identifies the author of a document.
- That a clean detector result indicates originality of idea.
- That your tool will give the same result as our tool.
- That a score will be permanent once the company updates the model.
- That AI detection is a reliable science.
What lands in your inbox
Your final draft, detection and similarity reports, source list and, the name of the editor. No account. No dashboard. No login necessary to retrieve files.
Verified by AI Detectors FAQs
They arrive with the draft, one attachment per tool, in the delivery email. No request is necessary. If you lose the email or need the files in a different format for compliance purposes, email support@articled.org with your order number.
No, and we would be suspicious of any provider who offered one. Detectors provide results that vary based on settings, sample length, and model updates. What we guarantee is that a specific person wrote the piece, and we will show you the notes, sources, and drafts.
For a hundred-word order, the detection tools are at their most unreliable, since there is too little text for the models to make a judgment. We run them nonetheless, and the results are attached with a caveat stated in the report and not hidden in a footnote.
Yes. Because a revised draft contains new text, it is sent for another round of detection and similarity. Therefore, you will receive the second set of reports in addition to the first, rather than instead of the first.
Let us know when you place the order and we will tell you honestly if we have it. If we do not, we will tell you this without quietly replacing it with a similar check.
It will most likely. An automated process assigns a reviewer after you flag the document. Reviewers generally schedule time to discuss improvements with the author. This process takes a few days for us to complete. If the process takes longer, we will send an email with the revised draft once it is complete. We will not send a draft that we have not finished checking.
Think of this page as a description of a process, not a badge. It accurately depicts the orders it describes because an editor performs these steps and keeps the output on file. If any aspect of your order was handled in a different way, the reports will confirm this, and so will we.
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.