Plain language is a set of techniques with a published standard behind it, not simplification, and the plain version is sometimes longer than the original.
We have not measured this one. It is published because its techniques are named and individually checkable, not because it has been compared against a run without it.
What it knows, concretely. First, the nominalisation pass, which is the highest-yield single technique and the one most often described without being made operational. The skill carries the list: make a decision becomes decide, conduct an investigation becomes investigate, is in receipt of becomes receives, and about fifteen more, together with the reason it works, which is that a nominalisation hides the actor and forces a weak verb like make, give, effect or undertake to carry the sentence. Second, the rule for terminology, which is where most plain-language advice quietly does damage. A term of art whose boundary matters to the reader's rights is kept and defined, not replaced, and the file states the cost asymmetry that decides the ambiguous case: an unnecessary definition costs one sentence, and an approximation that moves the boundary costs the reader a wrong action they believed was right. Third, an honest treatment of readability formulas, including what they actually count, which is sentence length and word length and essentially nothing else, and why that makes both gaming them and passing them uninformative.
It also names the real standards precisely rather than gesturing at them: the international plain language standard, ISO 24495-1, was published in 2023 and sets four governing principles, and separately several jurisdictions impose plain-language obligations by law, which is a different kind of thing and the file says so.
Who it is not for. If your document is already short, task-focused and read by specialists, most of this will make it worse by expanding terms your readers already know. And if the document fails because of its layout rather than its sentences, this pass will polish text nobody reaches.
Measuring one skill honestly costs about twenty model sessions: five runs with it, five without, on real material, each output graded alone by a session that is not told the other arm exists, against a rubric written by somebody who never saw the skill. We have not spent that on this one yet, so it ships labelled rather than ships silently.
How it would be measured. Tier A spine, with a Tier B question underneath it. Both runs rewrite the same three dense sources, a consumer contract clause, a letter from a public body, and a set of dosing instructions, each built to contain a nested conditional and a term of art whose boundary matters. The rubric is fixed before either run and is answerable from the output: is every nominalisation resolved to its verb, does the condition precede the instruction, is the nested conditional converted to a table, is the term of art retained and defined rather than approximated, and does an actor appear before the action in each sentence. Whether readers actually understand the result is a separate measurement that needs people performing a task.
There is a good objective spine here and a hard question sitting underneath it. The spine is that each technique leaves a mechanical trace: an unresolved nominalisation is findable, a condition placed after its instruction is findable, a term of art swapped for an approximation is findable by comparison with the source. That half could be scored without judgement.
The hard question is that the point of plain language is comprehension, and comprehension is measured by giving real people a task and watching whether they complete it. No output-only rubric measures that, so a full test would need a reader panel and a task, which is a research exercise rather than a scoring script.
The material also has to be genuinely dense rather than merely wordy, and it has to contain a term whose boundary matters, since the interesting failure is not verbosity but a rewrite that reads beautifully and has quietly changed what the reader is entitled to.
The rule that decides pass or fail was written down before any run was executed and it does not move afterwards. It is in the method note on the hub, along with the full results table including every skill that was tested and cut.
Stated plainly, because a skill that claims everything is useful for nothing.
name and description in the file's frontmatter, so you can also invoke it by name.A plain-language specialist is better than this and the gap is large, particularly on anything legal or clinical, where the risk of a rewrite is that it is clearer and wrong.
The free government guidance is more thorough than this file and should be treated as the authority. Its shape is a reference work, organised by topic, which is exactly right for looking something up and less useful when the question is what to do first on a paragraph in front of you. That ordering is what this adds.
A readability tool is cheap, instant and answers a narrower question than people think it answers. Use one as a smoke alarm.
The model with no skill at all will rewrite a dense paragraph into something markedly easier on request, and for internal documents that is usually enough. What this file carries that a general rewrite does not is a stated test for which terms of art survive untouched even when you were asked to simplify them, and the position that a plain version is frequently longer than the original rather than shorter.
---
name: plain-language-rewrite
description: Rewrites dense, formal or bureaucratic text using named plain language techniques, in an order that puts the highest-yield pass first. Covers the actor-before-action rule, the specific cases where the passive voice is correct, subject-verb distance, one idea per sentence, condition before instruction, the nominalisation pass with a list of the common ones, reader vocabulary versus institutional vocabulary, converting nested conditionals into tables, when a technical term must be kept and defined rather than replaced, and the honest use of readability formulas. This skill should be used when rewriting a contract, policy, notice, form, letter, instruction or help page for a reader who has no choice about reading it.
---
# Plain language rewrite
## The claim this skill is built on
Plain language is a set of named techniques with a published standard behind it. It is not simplification, and treating it as simplification is how rewrites go wrong.
The obvious approach is to shorten the sentences and swap the long words for short ones. It produces a document that scores better on every readability formula and is frequently no more usable, because neither sentence length nor word length is what was stopping the reader. What stopped them was usually one of four things: they could not tell whether the document applied to them, they could not find the part that concerned them, they could not tell who was supposed to do what, or a condition arrived after the instruction it governed so they had to read the sentence twice to find out that they did not have to read it at all.
Every technique below targets one of those. None of them is about being brief. Several of them make the text longer.
## The standards, described honestly
**ISO 24495-1, Plain language, Part 1: Governing principles and guidelines, was published in 2023.** It is a voluntary international standard, not a law, and it sets four governing principles: readers get the information they need, can find what they need, can understand what they find, and can use it. The value of the standard is not that it is prescriptive, because it largely is not, but that it makes plain language a thing that can be specified in a contract or a brief rather than argued about. Further parts covering specific domains have been in development, and you should check the current status on the ISO site rather than trusting a list in a file.
**Separately, and this is a different kind of obligation, several jurisdictions require plain language by law for certain material.** In the United States, the Plain Writing Act of 2010 requires federal executive agencies to use plain writing in documents they issue to the public. In the European Union, the General Data Protection Regulation requires information to data subjects in clear and plain language, and consumer contract terms have long been required to be in plain, intelligible language under the unfair contract terms rules. A number of United States states have their own plain-language requirements for consumer contracts and insurance policies, some dating to the late 1970s.
**The distinction matters.** A standard is something you choose to conform to. A statute is something that applies to you whether or not you have heard of it, and its scope is narrow and specific. Do not tell a client that plain language is required by law without checking what actually applies to their document, their sector and their jurisdiction, and the honest answer for most commercial writing is that nothing legally requires it and it is still worth doing.
## The techniques
**1. Put the actor before the action.** Mechanism: readers build a sentence around a doer. When the doer arrives late or never, the reader holds the whole clause in memory until it does. Before: "Consideration of the application will be undertaken." After: "We will consider your application." Ask of every sentence: who is doing this? If the answer is not in the sentence, that is the defect, and it is usually deliberate.
**2. Prefer the active voice, and know the four cases where the passive is correct.** The passive is correct when the actor is genuinely unknown ("the form was submitted on 3 March"), when the actor is irrelevant and naming them adds noise, when the receiver is the topic and keeping it in the subject position preserves the thread across sentences, and when naming the actor would be an accusation you cannot support or a disclosure you should not make. Outside those cases the passive is usually hiding somebody, which is why institutions like it.
**3. Keep the subject and its verb close together.** Mechanism: a reader who has a subject but no verb is holding an incomplete clause, and everything inserted between them is held with it. Before: "The tenant, in the event that the property is damaged by fire, flood or storm during the period of the tenancy, must notify the landlord." Thirty words between subject and verb. After: "If the property is damaged by fire, flood or storm during the tenancy, the tenant must tell the landlord."
**4. One idea per sentence.** Not one clause, which produces a staccato text that is harder to read, and not a word limit. One idea. The test is whether you can split the sentence at a conjunction without either half needing the other to be understood; if you can, it was two.
**5. Put the condition before the instruction.** Mechanism: a reader who meets the condition first can skip everything that follows if it does not apply, and a reader who meets the instruction first has to hold it while deciding whether it is theirs. Before: "You must complete form B4 if you were self-employed for any part of the tax year." After: "If you were self-employed for any part of the tax year, complete form B4." This one is nearly free and it changes how a whole document feels, because most policy documents are written the wrong way round throughout.
**6. Replace a nominalisation with the verb it came from.** The highest-yield technique. Its own section below.
**7. Use the reader's word, not the institution's word.** Mechanism: institutional vocabulary is precise inside the institution and empty outside it. Readers say pay, not remit; ask, not enquire; end, not terminate; house, not dwelling; before, not prior to; about, not in respect of. The exception is a term of art, covered in the terminology section, and the difference between jargon and a term of art is the entire skill of this technique.
**8. Turn a complex conditional into a table.** Mechanism: prose is linear and a conditional with three variables is not, so the reader has to do the transposition themselves and they do it wrong. Once a rule has more than two conditions or more than three outcomes, a table with the conditions as rows and the outcome as the final column will beat any sentence. This is the technique people skip because it does not look like writing, and it is often the single largest improvement available.
## The nominalisation pass
A nominalisation is a verb turned into a noun. The damage is not the noun itself, it is that the sentence now needs a filler verb, usually make, give, take, effect, provide, conduct, undertake or perform, and that filler verb has no actor attached to it. Resolving nominalisations therefore fixes three things at once: it restores the real verb, it shortens the sentence, and it forces the actor into the open.
Run this pass first, on the whole document, before anything else. It is fast, mechanical, and it exposes the sentences that were hiding who does what, which is where the real problems will turn out to be.
The common ones and their verb forms:
- make a decision, become decide
- give consideration to, become consider
- conduct an investigation, become investigate
- provide assistance, become help
- undertake a review, become review
- carry out an assessment, become assess
- perform a calculation, become calculate
- reach an agreement, become agree
- make an application, become apply
- make a submission, become submit
- provide notification, become notify
- give authorisation, become authorise
- make a payment, become pay
- effect a reduction in, become reduce
- is in violation of, become breaks or breaches
- is in receipt of, become receives
- issue a determination, become decide
- have a requirement for, become need
- carry out an inspection, become inspect
- take into consideration, become consider
The pass has one exception worth stating: keep the nominalisation where it has become the name of a thing rather than a description of an act. An application in the sense of the document you submitted is a thing. An assessment that has a reference number is a thing. Turning those into verbs produces nonsense.
## Readability formulas, handled honestly
**What they measure.** Flesch Reading Ease, published in 1948, and the Flesch-Kincaid Grade Level derived from it in 1975 for the United States Navy, both combine average sentence length with average syllables per word. Gunning Fog, from 1952, counts sentence length and the proportion of words of three or more syllables. SMOG, from 1969, counts polysyllables. Coleman-Liau and the Automated Readability Index substitute characters per word for syllables. That is essentially the whole family: **sentence length and word length, and nothing else.**
**Why that makes them easy to game.** Split a sentence at a comma and the score improves without a word of meaning changing. Swap a familiar three-syllable word for an unfamiliar two-syllable one and the score improves while comprehension falls. A rewrite driven by the number will do both of these, because they are the cheapest ways to move it.
**Why a text can score well and be incomprehensible.** The formulas cannot see whether a word is known to your readers, whether the sentences are in a sensible order, whether the conditions are stated before the instructions, whether anything is missing, or whether the document answers the question the reader arrived with. A page of short sentences full of undefined jargon in random order scores excellently.
**The correct use.** Treat a score as a smoke alarm, not a target. A document scoring far above the reading level of its audience tells you something is wrong somewhere and is worth investigating. A document scoring well tells you nothing at all. Never set a score as an acceptance criterion, because a criterion turns the formula into a target and the gaming follows automatically. And note that the formulas are built for continuous English prose, so they are meaningless for tables, forms, lists and headings, which is often most of the document.
## Terminology, and the decision rule
This is where careless plain-language work does real harm, because replacing a term of art with an approximation produces text that is easier to read and no longer says the same thing. The reader then acts confidently on a wrong understanding, which is worse than being confused, because confusion produces a phone call and false confidence does not.
- **The term has a legal, clinical or technical definition whose boundary affects the reader's rights, money or safety:** keep it. Define it at first use, in one sentence, and use it consistently thereafter, never varying it with a synonym. The style guide conformance skill sets out why that rule is absolute in documentation and what it costs to break.
- **The term is institutional jargon with an everyday equivalent that means exactly the same thing:** replace it. Remit becomes pay, commence becomes start, in the event that becomes if.
- **The term is jargon whose everyday equivalent means nearly the same thing:** keep the term, define it, and give the everyday word as a gloss. Nearly is precisely where the harm lives.
- **You cannot tell whether the boundary matters:** keep the term and define it. The asymmetry decides it. An unnecessary definition costs the reader one sentence. An approximation that shifts a boundary costs them a wrong action they believed was right, and costs you the argument about what the document meant.
## Structure as a plain language technique
Sentence-level work is the visible part and is not usually the largest win.
**Headings that answer questions.** A heading reading "Cancellation" is a filing label. A heading reading "How to cancel, and what it costs" is a promise about content and lets a reader skip. Where the document exists to answer a question, use the reader's question as the heading, in their words.
**Most important information first.** Within the document and within each section. The reader's first question is always whether this applies to them and what it means for them, and they should not have to earn that answer.
**Invert the usual chronological order.** Institutional documents are written in process order, because that is the order the institution experiences it: we received your form, we assessed it against these criteria, this took place at this stage, and therefore your payment will change on 1 October. The reader wants the last clause first. Putting the outcome first is not a stylistic preference, it is what lets a reader decide whether to read the rest, and readers who cannot make that decision quickly stop reading altogether rather than reading everything.
## Testing with real readers
Everything above is a prediction. Testing is the only thing that is not.
The cheapest version that still works: five people who are not on your team and match the real audience, one realistic task each, and you watch without helping. The task is the important part. Not "does this seem clear?", which returns politeness, but "using this document, work out how much you would pay if you cancelled in the second month", which returns an answer that is right or wrong. Record where they stop, what they re-read, and what they answer.
Three people find the fatal problems. Five finds most of the rest. Cheaper still, and worth doing when nothing else is possible: one person reading aloud, marking every sentence they had to read twice. The re-read is the signal, and people can report it reliably even when they cannot say what was wrong.
## What plain language is not
**Not shorter.** Often longer, and the clearest cases are the ones where a cross-reference is replaced by the thing it points to.
**Not simpler in content.** The content is whatever it is. Plain language removes the difficulty added by the writing, not the difficulty inherent in the subject, and a document about a complicated entitlement stays complicated.
**Not condescending.** Explaining a term is not talking down, and the reader who already knows it loses two seconds. Talking down is a matter of tone, not vocabulary, and the reliable way to sound condescending is to tell the reader something is simple.
**The example where longer is better.** Original: "Refunds are subject to clause 7.2." Nine words, and the reader now has to find clause 7.2, read it, and work out which part applies. Plain version: "You can get a full refund if you cancel within 14 days of your order. If you cancel after 14 days, we refund what you paid minus the delivery cost. Clause 7.2 has the full rules." Three times the length, and the reader is finished.
## Worked example, one dense paragraph
The contract wording below is invented for this example.
**Original.** "Notification of the cessation of service must be provided by the customer to the company not less than thirty (30) days prior to the intended date of termination, and in the event that such notification is not received within the aforementioned period, the customer shall be liable for the payment of charges in respect of the subsequent billing period, save where the cessation arises in consequence of relocation to a premises at which the company does not provide service, in which case the requirement for notification shall be reduced to fourteen (14) days."
**Nominalisation pass.** Notification becomes tell. Cessation becomes stop or end. Payment of charges becomes pay. Relocation becomes move. Four filler verbs go with them: provided, received, arises, shall be liable for the payment of.
**Actor before action.** The original hides both actors behind passives: provided by the customer, is not received. The customer tells us. We charge.
**Reader's word.** Prior to becomes before. In the event that becomes if. Save where becomes unless. In respect of becomes for. Premises becomes address. Aforementioned goes entirely, since it points backwards at something the reader has to hold.
**Condition before instruction, and one idea per sentence.** The original is one sentence carrying a rule, an exception, and a penalty. That is three ideas and it needs three sentences, or better, a table.
**Complex conditional to a table.** Two conditions, moving or not moving, and two outcomes, notice period and charge, which is exactly the shape prose handles badly.
**The rewrite.**
"To end your service, tell us at least 30 days before the date you want it to stop. If you give us less than 30 days' notice, you pay for one more billing period. If you are ending the service because you are moving to an address where we do not operate, you only need to give 14 days' notice."
| Your situation | Notice you must give | If you give less notice |
| --- | --- | --- |
| Ending the service | 30 days | You pay for one more billing period |
| Moving to an address we do not serve | 14 days | You pay for one more billing period |
**Verdict.** The original is 92 words in one sentence with two passives, five nominalisations and the exception buried at the end where a reader who is moving house will never find it. The rewrite is 71 words plus a table, so slightly shorter in prose and longer overall, and the moving-house case is now visible in two seconds instead of being the last clause of a sentence most readers abandon. The one thing that has not been checked, and cannot be checked here, is whether "billing period" carries a defined meaning elsewhere in the contract, which is a question for the lawyer and not for the rewriter.
## Failure modes
**Rewriting to a score.** Sentences split at commas and words swapped for shorter ones, the number improves, and nothing is clearer. The tell is that the changes are all mechanical and none is structural.
**The approximated term of art.** A defined term replaced by an everyday near-equivalent. The document reads better and now means something else, and nobody notices until it is relied on.
**Sentence-level work on a document that fails structurally.** Every paragraph polished, and the reader still cannot find whether it applies to them, because the answer is on page three.
**Keeping the institution's order.** The document runs in process order because that is the order it happened in, and the outcome the reader came for sits at the end.
**The condition after the instruction, throughout.** Individually invisible, collectively the reason a policy document feels heavy. Readers hold instructions they did not need.
**Explaining that it is simple.** Telling the reader something is straightforward, which adds nothing when they agree and is insulting when they do not.
**Testing by asking whether it is clear.** People say yes. Comprehension is measured by a task with a right answer, never by an opinion.
## What this skill does not do
- It does not confirm legal correctness. A rewritten clause may carry different obligations, and only a lawyer can say whether it still does what it must.
- It does not measure comprehension. Everything here predicts what readers would do, and only readers doing a task can tell you.
- It does not design the document. Layout, typography, page order, form field sequence and what appears on which screen are frequently the real defect, and a designer beats a rewrite there.
- It cannot tell you which laws or standards apply to your material. It names them accurately as of August 2026 and the scoping question is jurisdiction-specific.
- It does not handle languages other than English well. The techniques are English-centred, and other languages have their own plain-language traditions that this does not carry.
These skills all ask your assistant to check things against your actual codebase, your actual schema, your actual design system. Locul keeps that context current on its own, from the files you already have, on your machine. Mac and Windows, free to start.