Academic managers online nowA person writes it. Tools are allowed to do the dull parts.Submit assignment →
AI usage policy

A person writes it. Tools are allowed to do the dull parts.

Pretending nobody in this industry touches a language model would be a lie, and promising you a machine wrote nothing at all is a claim we could not enforce. So the line is drawn task by task instead, with what happens if a specialist crosses it.

Five commitments
01Argument and prose are humanwritten by a named specialist who can be asked why a paragraph says what it says.
02Tool use is disclosed by categorythe task-by-task list below is the whole of it, and specialists sign it.
03Your work never trains a modelno brief, file or delivered piece is used as training data, by us or a supplier.
04A detection report comes with every deliveryso you see what a marker running one would see, before they run it.
05Generated work is a full refundand the specialist leaves the panel. Stated as a term, not a courtesy.

The line, task by task.

Eight jobs that go into a piece of work. Select one to see the rule, why it is drawn there, and how it is enforced.

The task
Never

Forming the argument

Never generated. The thesis, the line of reasoning and the decision about which authorities matter are the specialist’s own, developed against an outline a manager approved.

Why the line is here

This is the thing you are paying for. A generated argument is confident about the wrong things, and neither the specialist nor we could defend it if asked.

How it is enforced
Outline approved before drafting on anything above 1,500 words
The second reader can ask the writer to justify any claim, and does

About AI detectors, honestly.

Detectors are unreliable in both directions, and anyone selling you a guaranteed score is selling you something they cannot deliver.

We run one on every delivery and attach the report, including the sections it flagged rather than only the headline number.
We tell you which tool produced it and on what date, because tools change their behaviour between versions.
We do not treat a low score as proof of anything, and we will not quote you a target score.
We do not optimise writing to beat detectors. That is evasion, and it degrades the prose.
We will not tell you a flag on human writing is impossible — formal academic prose is exactly what these tools most often misread.

If your work gets flagged.

Step 1Send us the report and the flagged passages. No form, no ticket queue — your manager handles it.
Step 2We re-run our own checks on the delivered file and give you the second-read note and the audit trail.
Step 3We tell you what we actually think, including if we believe the flag is justified. You are better served by that than by reassurance.
Step 4Where the fault is ours, the refund follows the published policy automatically — and we say so in writing you can show.

Your brief, your files, and what never goes into a model.

Six undertakings about your data specifically, separate from the confidentiality policy.

Undertaking 01

No training on your work

Your brief, uploads and delivered files are never used to train or fine-tune any model, by us or by a supplier acting for us.

Undertaking 02

No pasting your brief into a public tool

Specialists may not paste your brief, files or identifying details into a consumer AI service. This is a panel condition, not a suggestion.

Undertaking 03

Contractual opt-outs with suppliers

Where a tool is used operationally, training-data use is contractually excluded and the supplier is named in the privacy policy.

Undertaking 04

Identity separated from the brief

The specialist never sees your name, email or institution — so even a mistake could not associate the work with you.

Undertaking 05

Detection is a one-way check

Files sent to a detector are submitted with retention disabled where the tool supports it, and never added to a repository.

Undertaking 06

Deletion on request

Ask and your files are deleted within 30 days, no reason required, under the privacy policy.

Using AI in your own studying.

Not our decision to make for you, but here is what we would tell a friend.

Find your institution’s actual rulemost now publish one, and some set it per module or per assessment rather than centrally.
Declaration requirements are the trapseveral institutions permit tool use but require you to declare it — and treat the missing declaration as the offence.
Keep your drafts and promptsa version history is the most useful evidence you can have if a question is ever raised.
Never let a tool supply citationsfabricated references are the single most common way students are caught, and they are trivially checkable.
Ask before, not aftera question to your module leader is not an admission. A hearing is much worse than an awkward email.

What we will not build.

Each of these has been suggested to us, and each would sell.

A "humanise this text" toolthe entire purpose is evasion, and we will not sell it.
Instant generated draftswe could ship it cheaply tomorrow; it would make the specialists pointless and the product worthless.
Detector-score optimisationwriting to defeat a tool is writing worse, and it puts you at more risk rather than less.
A model trained on client workno matter how anonymised, the answer here is simply no.

Questions people actually ask.

So is any AI involved in my order at all?

Possibly, in the categories marked "allowed" on the list above — spellcheck, reference formatting, literature search, statistical software. Never for argument, analysis or prose. If you would rather no tool touched your order in any capacity, say so on the brief and we will staff it accordingly and note it on the file.

Why not just claim to be 100% AI-free?

Because it would not be true unless we also banned spellcheck and reference managers, which no serious writer works without. A promise that specific is either false or means less than it sounds, and we would rather publish the actual line.

Can I ask for a fully tool-free order?

Yes. Ask your manager and it is recorded as a condition of the brief. The specialist works without any AI-assisted tool, and the piece takes slightly longer — usually a few hours on a standard essay.

How do you actually know a specialist did not generate a draft?

Three ways, none perfect: the detection report on the delivered file, a second reader who can question any claim in the piece and does, and the outline agreed before drafting. A generated draft rarely matches an outline it was not built from. Where we find one, the arrangement ends.

Is my brief sent to any AI service?

No. Specialists are contractually prohibited from pasting your brief or files into consumer AI tools, and the operational tools we do use exclude training-data use by contract. The processor list is in the privacy policy.

My institution says AI use is misconduct. Does that affect ordering?

It affects what you can do with the work, which is covered in the plagiarism policy — read that one too. As to us: our specialists write the pieces, and if you need to state on a declaration that no AI generated the content of a piece we supplied, that statement is accurate.

If a generated draft were good enough, we would not employ specialists.

The whole cost structure of this business is people reading and writing. That is the simplest reason to believe the policy above.