Plain language does not fail at the draft. It fails after launch.

In March 2021, Ethos CRS and VisibleThread scored 136 documents from 35 Australian Government agencies against a plain language benchmark. Every one fell below it. The average document sat at a reading level of 15.2, which is third-year university.

The same report scored 2 agency websites. Services Australia's came in at 119.3 against a benchmark of 100, at a year 7 reading level, with 0.3% long sentences and 98% active voice.

Same agencies. Same writers. Same standards.

The websites passed and the documents failed, and the difference was never writing skill.

What a website has that a report does not

Two things. A team whose job is the whole of it. And a standard applied on the way out.

Policy areas write reports. Executives clear them. They publish once. Nobody reads them across. Nobody holds them to a reading level. No publishing step asks whether a person could use this, with somebody standing there who can say no.

That is a governance finding, not a literacy one.

The number you will be handed, and what it is worth

Somebody will quote you 79%. It comes from a VisibleThread study of 28 federal and state agency websites, sampled in May 2016. Only 6 scored acceptable readability. It is a decade old, it mixes federal and state bodies, and it measures against VisibleThread's own benchmark rather than an official standard.

Use it to date the problem. Do not use it to describe where government content sits now.

Adobe's Digital Government Index is more current. It scored 15 Australian agencies at 69.4 out of 100 in 2025, up from 58.8 in 2022 and flat since 2023. Readability is one input to that composite, and Adobe sells the tooling. Treat it as a direction, not a score.

Australian government content got better, then stopped. That is the problem your agency is in.

Why more training will not move it

Plain language often gets treated the way work health and safety training does. One course, signed off, ticked.

The guidance is not the gap. The Australian Government Style Manual recommends a year 7 reading level, which around 83% of the population is likely to understand. It sets sentences at an average of 15 words, with 25 as the ceiling. Standards Australia adopted the international plain language standard, AS ISO 24495.1:2024, in April 2024. Criterion 4 of the Digital Inclusion Standard names the Style Manual by title. That makes it a compliance instrument, not a style preference.

A trained writer produces a good draft. Then the page goes live beside legacy content nobody has touched in years. Six months later a policy area adds a section. A year later somebody updates the form instructions. Two years on, the page reads like a committee that never met.

Training improves the draft. Nothing in an agency's normal operating rhythm improves the page after that.

Four roles most content teams do not have

None of these need to be full-time. Two are already inside somebody's job without a name on them.

Name an owner for every page

Not a branch, not a team inbox. A person.

Their job is one question, asked on a cycle: is this still true, and is it still the fastest way to the answer.

Take your 10 highest-traffic pages and write a name beside each one.

The pages where you cannot fill in a name are the pages that rot.

Add a clarity gate, separate from the accuracy gate

This is the missing one. Every agency has a step where a subject matter expert confirms the content is correct. Almost none have a step where somebody confirms a person can use it.

Those are different questions and they need different approvers. The person who guarantees legal accuracy has every reason to add qualifiers, not remove them.

A clarity approver needs the standing to hold a page until it reads.

Measure the whole site on a cycle

Somebody who reads across, not down. Quarterly, against the Style Manual and AS ISO 24495.1. One score, one list of what slipped, nothing else.

In most teams this is the content lead or an external reviewer. It does not need a new hire.

It needs the measurement to exist as an artefact somebody receives.

Give somebody the power to delete

Readability reports sink because the only action they can trigger is a rewrite. Rewriting 200 pages is not a plan.

When I led the ACT Government's digital consolidation we decommissioned 60 websites. That was half of all ACT Government sites outside schools. On what remained we often cut 90% of the content before it moved to act.gov.au. Almost none of that work was rewriting.

Retirement is faster than rewriting, and on a large site it is the single intervention with the largest readability effect.

A report with no retirement power is a to-do list, and a to-do list is where site-wide readability goes to die.

Do them in this order

Owners first, because nothing else works without a name. Then the clarity gate, because it stops new drift while you deal with the old drift. Then measurement. Then the retirement decision, which is the one that moves the number.

Doing measurement first is the common mistake.

An audit delivered into a system with no owners and no retirement power produces a very good document and no change.

What good looks like

• Every page in your top 50 has a person's name against it, not a team.

• Somebody other than the accuracy approver can hold a page at publish for clarity alone.

• A named person receives a readability score for the whole site every quarter, in writing.

• You retired a page in the last 12 months because it was not worth fixing.

• Your last readability report produced a deletion list, not a rewrite list.

A page on your site drops below standard 6 months after going live. Whose job is it to notice? And whose job is it to decide the page should not exist any more?

If you cannot name both people, the failure rate is not surprising.

Photo by Nathan Bailly on Unsplash.