AI content review checklist for Australian government content teams
This is a white-label AI content review checklist for Australian government content teams. It is a pre-publication check for any content drafted or edited with generative AI tools, built to catch the failure modes a standard editorial pass misses: plausible inaccuracies, invented sources, tone drift, accessibility gaps, and undocumented AI use. Copy it into your own style guide, intranet, or content management system, and adapt the bracketed sections to your agency. Keep it short enough that a reviewer will actually use it on every piece.
How to use this AI content review checklist
Run this checklist on the near-final draft, after editing and before approval. It does not replace your normal editorial review or your agency's clearance process. It sits alongside them and adds the AI-specific checks those processes were not designed for. Treat any unchecked box as a blocker, not a preference: if a line cannot be ticked, the content is not ready to publish.
Assign one named reviewer per piece. Where your agency already has plain language, accessibility, or governance standards, link to them from the relevant section so reviewers check against the source, not their memory.
The AI content review checklist
Before you start
Confirm you have:
The final draft, with all AI-assisted sections identified
Access to the primary sources behind any facts, figures, or policy references
Your agency's style guide and accessibility standard
Your team's AI use policy or, if none exists yet, the Australian Government's policy for the responsible use of AI in government
Section 1: Accuracy
AI tools produce fluent text that can be confidently wrong. Verify, do not assume.
[ ] Every fact, figure, date, and policy reference has been checked against a primary source, not against AI output
[ ] No AI-generated statistic, quote, or citation has been used without independent verification
[ ] Legal, regulatory, or policy wording matches the authoritative source document word for word
[ ] Any named program, payment, form, or service exists and is current
[ ] Links resolve to the correct live page and are not invented or outdated
Section 2: Plain language
Plain language is a requirement for Australian government content, not a polish step. AI drafts often default to long, formal, generic prose.
[ ] Content meets the plain language guidance in the Australian Government Style Manual
[ ] The reading level suits the audience and the task
[ ] Acronyms are spelled out on first use and jargon is removed or defined
[ ] Sentences are short and carry one main idea
[ ] The content leads with what the reader needs to do or know
Section 3: Accessibility
Accessibility is a legal obligation under the Disability Discrimination Act 1992. AI output can introduce structural and accessibility problems that look fine on screen.
[ ] Heading hierarchy is logical and follows a correct order (no skipped levels)
[ ] Every meaningful image has accurate alt text written or checked by a person
[ ] Link text describes the destination and is not "click here" or "read more"
[ ] Colour is not the only way meaning is conveyed
[ ] Content meets WCAG 2.2 Level AA
Section 4: Tone and voice
Generative AI has a recognisable register. Left unchecked, it flattens an agency's voice and can mishandle sensitive material.
[ ] Tone matches your agency's voice and the seriousness of the topic
[ ] The content does not read as generic, over-formal, or machine-generated
[ ] Sensitive or high-risk topics are handled with appropriate care and reviewed by a subject matter expert
[ ] No em dashes, filler phrases, or hedging language that AI tends to add
[ ] The content says something specific, not a survey of everything that could be said
Section 5: Governance
If you cannot say who reviewed the content and how AI was used, you cannot defend it later.
[ ] AI use is documented in line with your team's AI use policy
[ ] The named reviewer and approver are recorded
[ ] A review date and content expiry date are set
[ ] The content is consistent with your agency's published AI transparency commitments
[ ] A human is accountable for the final published version
Sign-off
Complete before publishing.
Content title: [title]
Reviewed by: [name, role]
Approved by: [name, role]
AI tools used: [tool names and what they were used for]
Date reviewed: [date]
Next review due: [date]
Adapt this for your team
Add or remove checks to match your context. A regulatory team might add a check for statutory wording. A service team might add a check that the content matches the actual steps in the service. Keep the checklist to one page so it is used every time, not just for high-profile content.
Download the checklist
Did this land?
I write these to be used, not just read. Tell me if this wouldn't work in your organisation, and why.