L
📕 Companion article · 5 of 5

Saying how the work was made

why detection fails · what you can actually claim · the four-line account

Two questions get collapsed into one, and they have different answers. Whether a tool was used cannot be established by anybody, including you six months later. Whether the work holds can be established, in four lines, by the person who did it.

Detection · fails both directionsProvenance · cannot be provedWhat you checked · can
What you'll learn

Two questions, and only one of them is answerable

How it was made, and whether it holds. Almost every argument about AI and honesty is people answering the first question at each other while the second one goes begging.

  • Find the rule that applies to you, written down, rather than guessing at the norm.
  • Stop relying on detection — it clears machine text and accuses real people.
  • Claim what you verified, specifically, because that is the part that can be true.
  • Name what you could not close, which is the line that makes the rest believable.
Why it matters · where it's tested

Why the honest position is also the practical one

An account of what you checked survives being questioned, and a claim about how the words arrived does not survive the first person who asks how you would know.

Companion to the 25-day courseBuilds on days 9, 15, 20, 25Stands alone

People look for: AI disclosure, AI detection limits, evidence records, audit trails, citation checks, and accountable approval.

See defensibility built into the work

Everything in the top half is a claim you cannot support if pressed. Everything in the bottom half you can, in a sentence, with a document behind it.

What you can and cannot claim
The claimCan it be established?By what
No AI was used in this documentNo Nothing. Nobody can show it, including you, once time has passed
This was written by a personNo Detection clears real machine text and accuses real people. It settles nothing
AI was used only for editingNo A boundary nobody can locate afterwards, including the author
Every figure was checked against the source named beside itYes Open the source. Two minutes
Every name was confirmed with the personYes The messages exist
This claim is not verified, and here is whyYes It is a statement about your own work, and it costs you to make it
These three decisions were mine, and here is the reasoningYes The record from day 20, written at the time rather than afterwards
The four-line account

What the piece claims. What you checked, and against what. What you could not close, and what would close it. Who decided the things that were judgement rather than fact. Four lines, written while the work is fresh, and they answer every reasonable question anybody will ask — while the question everybody argues about goes politely unanswered, because nobody can answer it.

How evidence makes work defensible

Defensibility comes from evidence and named verification, not from guessing how prose was produced or attaching a tool name without context.

Find the rule before you decide the norm

Somebody has written a rule, or nobody has. Go and find out which, and read the actual words rather than the summary somebody gave you in a meeting. Calibrating to what colleagues seem to be doing is how people end up on the wrong side of a policy that was sitting in a shared folder.

Detection is not evidence

It fails in both directions, and the second failure is the serious one: careful, plain, well-structured prose is exactly what gets flagged, so the writers most likely to be accused are the ones who took the most care. Never rest a claim on a detector, and never accept one resting on you without asking what its error rate is.

Write it at the time, or not at all

Reconstructed six months later, an account is a reconstruction and everybody can tell. Written the afternoon you finished, it costs four lines and it is the difference between somebody who checked and somebody who says they checked.

Where disclosure genuinely is required
  • Follow applicable contracts, institutional rules, professional duties, and laws.
  • Disclose assistance when it materially changes authorship, evidence, risk, or the reader's decision.
  • Protect confidential inputs even when disclosure of the workflow is required.
Two things that make an account credible
  • It names the checks actually completed and the evidence used.
  • It records unresolved limits and the person who accepted them.

Defensibility in real decisions

The client who asked directly

The team answered with four concrete lines: assistance used, sources supplied, checks performed, and final approver. The account was useful because each statement could be examined.

The student cleared by nobody

A detector score was treated as proof until reviewers asked about false positives and direct evidence. The decision was restarted using process records and the student's actual work.

The figure that came back a year later

A published number was challenged after its source changed. The saved source version, calculation, review note, and approver allowed the organisation to explain and update it without guessing.

Glossary — the 4 words that unlock it

Provenance

What it means
A claim about how a document was produced, rather than about whether it holds.
Why it matters
Nobody can establish it, including you, once time has passed. It is the question everybody argues about.
Example
No AI was used in this document.
Key question
How would you show that, if somebody pressed you?

Verification

What it means
A claim about what you personally checked, and against what.
Why it matters
It can be true, it can be tested, and it is the one that carries weight.
Example
Every figure checked against the source named beside it.
Key question
Which of your claims could somebody else confirm in two minutes?

The named gap

What it means
A written statement of what you could not verify, and what would close it.
Why it matters
It costs you something to write, which is exactly why it makes the other lines credible.
Example
Not verified: the 70% figure. Every trail reaches one unsourced 1996 book.
Key question
What in your last piece did you quietly hope nobody would ask about?

Detection

What it means
A tool that guesses whether text was machine-written.
Why it matters
It clears real machine text and accuses real people, so it settles nothing in either direction.
Example
Careful, plain, well-structured prose is exactly what gets flagged.
Key question
What is the error rate of the thing being used to judge you?

Questions about evidence and disclosure

What makes AI-assisted work defensible?
ConceptualWhatcomplexity 2

Defensible work has traceable sources, named checks, recorded limitations, clear decision ownership, and an account of where assistance was used when that fact matters. The aim is not to prove a document's hidden production history; it is to support the claims and decisions people rely on.

Can an AI detector prove that a person used an AI tool?
ConceptualWhatcomplexity 2

No detector can establish authorship reliably from prose alone. False positives can accuse careful human writers, while edited machine output can evade detection. Treat a detector score as an uncertain signal, never as proof, and use process records or direct evidence for consequential decisions.

What should an AI-use disclosure say?
ConceptualWhatcomplexity 2

State the material activity rather than naming a tool without context: what assistance was used for, which sources or inputs governed it, what a person checked, what remained uncertain, and who approved the result. The detail should match the audience, risk, contract, and applicable policy.

Do citations make AI-assisted work trustworthy?
ConceptualWhatcomplexity 2

Citations make individual claims inspectable when they point to relevant, accessible sources and the author has checked that those sources support the claim. A long reference list does not prove the analysis, and an assistant's citation must be opened and verified before it carries weight.

What records should a team keep for important AI-assisted work?
ConceptualWhatcomplexity 2

Keep the source materials or versions, key instructions, material outputs, model or tool version when relevant, verification performed, unresolved limitations, changes made by people, and final approver. Retention should follow the work's risk and privacy rules rather than storing every conversation indefinitely.

Key takeaways

  • Detection is not proof of authorship.
  • Disclose material assistance in terms the audience can use.
  • Open and verify every citation that carries a claim.
  • Record checks and unresolved limits at the time.
  • A named person or team owns the final decision.

Where this lesson leads

Move from a claim about production to evidence about verification, limitations, and accountable decisions.
✓Source record
✓Verification record
✓Disclosure
→Risk-based retention
→Independent review
→Defensible governance

Keep making work accountable

You cannot prove how it was made. You can prove what you checked, and only one of those was ever the question.

▶ Try the playground
© Pawan Nayar · LLOS.ai · 2026 — Saying how the work was made.Original pedagogy, voice, and design — all rights reserved.