Simplified Technical English, checked against the rule that says so.
ASD-STE100 defines 53 writing rules and a controlled dictionary for technical documentation, so that a maintenance instruction means one thing to a reader anywhere. This section publishes all 53 rules with worked examples, and the full set of system prompts for checking text against them, so you can run every check against your own model without giving us anything. A checker that runs them for you, with the dictionary lookup built in, is the next phase.
Controlled language is an old answer to a new problem.
Aerospace and defence have spent forty years making technical writing unambiguous for readers who are not native English speakers, and ASD-STE100 is what that produced: a fixed vocabulary, a fixed grammar, and a rule you can cite for every judgement.
That turns out to be the problem language models create at scale. Output that reads well and cannot be verified against anything is the failure mode, and a controlled language standard is a specification a machine can be held to. This section is an AI governance practice taking that seriously enough to build the tool.
The prompts are published. Take them.
Every system prompt this section is built on is published here, one per phase, rendered from the rule pack. Six of them work with any model and need nothing but your text. Copy them and run the whole thing yourself.
What a copied prompt cannot do alone is the dictionary. An approved word is approved as one part of speech with one meaning, so knowing which words in your text are on the list, and what to use instead, needs a lookup against the extraction. Three of the ten prompts take that lookup as an input block; the dictionary page describes what it holds, and the live lookup is the next phase of this section.
See the promptsAll 53 rules, with worked examples.
Each rule page carries the rule in plain terms, what a checker looks for, how to rewrite what it finds, before-and-after examples, and the page of the standard it comes from.
Browse the rulesGuides
Two explainers, for two different starting points.
How to Make an LLM Write in Simplified Technical English
Pasting all 53 rules into a prompt and asking for a rewrite does not work well. What does.
ASD-STE100 Explained: Simplified Technical English
53 writing rules plus a controlled dictionary of about 2,000 words. What it requires, and what it is not.
About this rule pack
How AI is used in this section
The rule pages and the published prompts are written and rendered from a structured rule pack; no model wrote them. The worked example on the prompts page is stored model output, reviewed once and kept with its errors showing. Nothing here runs a model on your text: you run the prompts against a model you choose, and what comes back is a set of candidate findings, each naming the rule it cites, for a person who knows the standard to accept or reject.
The ASD Simplified Technical English Maintenance Group does not endorse AI tools, this one included, and its June 2026 white paper on STE and AI(opens in a new tab) sets out its position: the standard takes priority, AI supports and does not replace a human author, and responsibility for the text stays with the person who signs it off. This section is built on that basis. Do not present any output derived from it as evidence that a document conforms to ASD-STE100.
ASD-STE100 is an international standard owned and published separately by the ASD Simplified Technical English Maintenance Group(opens in a new tab). Axiom Verity is not affiliated with ASD or the STEMG, and this site does not reproduce the standard. What is published here is a rule-by-rule commentary with our own examples, and a structured extraction of the controlled dictionary. If you need the standard itself, obtain it directly from ASD.
The dictionary extraction holds 2,037 entries against the 2,149 the standard declares, which is 94.8%. It was produced programmatically from the Issue 9 document and spot-checked rather than verified word by word. A word the checker cannot resolve is reported as needing review and is never passed silently.
The same discipline, applied to AI governance.
Every finding traceable to a citable rule, nothing asserted bare, and the limitations stated in the open. That is the standard the rest of this site is held to.