24 September 2026

From Level 0 to Level 99: A Simple Way to Explain How You Use AI

A numbered AI level from 0 to 99 gives readers a quick, honest answer about how much of a piece AI wrote and how closely a person checked it.

BOUTIQACADEMY

Article

AI in Practice

From Level 0 to Level 99: A Simple Way to Explain How You Use AI

BoutiqAcademy · September 2026 · Source: own IP + named sources. AI level 3.

You will already know that AI can play a part at any stage of producing professional content, from research through to final formatting. The harder part is explaining where. A client reading your article, a learner working through your course, or a peer scrolling past your post deserves a quick, honest answer to one question: how much of this came from a person, and how much came from a machine?

Why a clear AI statement matters

A single label such as "AI-assisted" covers too much ground. It could describe a book you wrote yourself and had AI format, or a post AI drafted in full while you gave it a final read. Those two pieces carry very different levels of human judgement, and a reader can only weigh them fairly if they can tell them apart.

The cost of a vague label runs both ways. Overstate the human effort and you risk your credibility when the truth comes out. Understate it and you undersell the thinking and review that went into the work.

An AI level is a short, numbered statement of how much of a piece AI produced and how closely a person checked it. It gives the reader a precise answer in a few words. A number also makes AI use easy to compare across pieces. A reader who sees Level 1 on one piece and Level 5 on another can tell at once which one carries more of the author's direct work.

The question every piece of content has to answer

How much of this did a person write, and how closely did a person check it?

From the author's side, the matching question is about effort and risk: what level of review does this piece deserve, given what it asks the reader to trust? A course module that teaches professional practice carries more weight than a short social post, and the review should reflect that.

How can you describe your AI use in one line?

A useful starting point is to separate three things and be clear about each one.

Source, Research, Level

The source: Where the ideas and substance come from.
The research: How the supporting facts and examples were gathered.
The level: How much of the writing AI did, and how closely a person reviewed it.

The source

The source is the foundation of the piece. At BoutiqAcademy, every article, course and post starts from our own intellectual property: the frameworks, methods and practice material BoutiqAcademy has developed. Named external sources are added where they strengthen the argument.

Own IP covers the frameworks, models and course material the author or organisation created.
Named sources cover published research, standards and frameworks, cited by name so a reader can check them.

Stating the source first matters because it answers the question readers care about most: whose ideas are these? When the source stays constant, a higher AI level changes the wording, and the substance still traces back to the author.

The research

AI supports the research behind all BoutiqAcademy content from Level 1 upward, from gathering examples to checking facts against published sources. Treating research as a shared baseline keeps the scale simple.

At Level 0, the author does the research without AI.
From Level 1 upward, AI supports the research and the author decides what to use.

The trade-off is transparency against length. Listing every research tool on every piece would add length to every piece, so one baseline statement covers it.

The level

The level is the number that changes from piece to piece. It runs from 0, where a person does everything, to 99, where AI produces and releases the work with no human review. The gap between 5 and 99 is deliberate. It shows how far a fully automated output sits from anything a person has checked.

The guiding principle is simple: the more a piece carries professional judgement, the more closely a person checks it.

Choosing a level before starting a piece is also a quick way to decide how much review time it needs.

Lower levels suit work where the author's own words and judgement are the product.
Middle levels suit structured work built from established IP, where review focuses on accuracy.
Higher levels suit short pieces with lower stakes, where a final read is proportionate.

The trade-off is speed against assurance. Each step up the scale saves the author time and shifts more of the checking onto tools and final reads, so the level should rise only as far as the stakes allow.

The scale, level by level

Each level describes a different split of writing and review. One piece of content sits at one level.

Level 0: All human
No AI at any stage, research included.
Level 1: Human-authored, AI-formatted
The author writes every word. AI handles layout, conversion and presentation.
Level 2: Human-led, AI-assisted
The author sets the structure and argument. AI expands sections, and the author edits line by line.
Level 3: AI-drafted, expert-reviewed
AI drafts from the author's IP. The author reviews the substance, corrects it and approves it.
Level 4: AI-produced, quality-checked
An automated pipeline builds the output. The author checks a sample, and checking tools flag issues for attention.
Level 5: AI-generated, light-touch review
AI writes to a brief in the author's voice. The author gives it a final read before release.
Level 99: All AI
AI produces and releases the output with no human review.

Where the idea comes from

The scale borrows its shape from the SAE levels of driving automation, which rate vehicles from Level 0, where the driver does everything, to Level 5, where the vehicle drives itself. Borrowing a familiar structure means the content version needs little explanation.

Other frameworks cover related ground. Sheridan and Verplank described ten levels of automation in 1978, focused on decisions between people and computers. The AI Assessment Scale, developed by Perkins, Furze, Roe and MacVaugh, gives educators five levels for how students may use AI in assessment. Kari Weaver's Artificial Intelligence Disclosure (AID) Framework offers a structured statement that lists AI use task by task.

Each of these serves its own purpose. The content scale takes the simple numbered structure and applies it to professional publishing, where the reader wants to know both how much AI wrote and how closely a person reviewed it.

Where to start

A one-line tag is enough to put the scale to work: "Source: own IP + named sources. AI level 3." Before you add one to your own material, three questions help.

Which pieces of your content carry the most professional judgement, and do they get the closest review?
If a client asked how a specific piece was produced, could you answer in one sentence?
Is any client-facing work currently going out at Level 99 without you realising it?

When the source is clear, the level is honest, and the review matches the weight of the work, readers are more likely to trust both the content and the person behind it.

This article is Level 3: AI-drafted from BoutiqAcademy IP and named sources, then reviewed and approved by the author.