AI Disclosure Checklist for UK Journalism — When to Declare AI Use
There is no UK statute on AI disclosure in journalism and no settled regulatory rule, which leaves each newsroom improvising a line between a transcription tool nobody needs to know about and a drafting tool readers plainly should. The distinctions that matter are what the tool touched and whether a human checked the result.
Last reviewed: Next review due:
AI Disclosure Checklist
Determine whether you need to disclose AI use in your published work, and generate sample disclosure text. Guidance: NUJ AI guidance and IPSO position.
Information only — not legal advice. AI disclosure standards are evolving. Check the latest NUJ and IPSO guidance for your publication type. See full disclaimer.
What was AI used for? (check all that apply)
Were facts independently verified by a human?
Is the published version primarily AI-generated content?
How it works
- You tick which of nine uses applied — drafting, research and summarising, transcription, translation, image generation, image editing, headline suggestions, fact-checking, code generation — then set the human edit level, whether facts were independently verified by a human, whether the published version is primarily AI-generated, and your name.
- The rules run in a fixed order and the first match wins. “No” to human verification overrides everything and returns “Do not publish” in red, regardless of what the tool was used for.
- Image generation, or answering yes to “primarily AI-generated”, returns “Disclose explicitly”. So does drafting that has only been lightly or moderately edited.
- Drafting that was heavily rewritten and fully verified, or image editing, or partial verification, all return “Brief mention recommended”. Support-only uses — research, transcription, translation, headlines, fact-checking, code — with full human verification return “No disclosure required, document internally”.
- The “disclose explicitly” and “brief mention” outcomes come with sample disclosure wording naming you as the journalist who edited and verified, with a copy-to-clipboard button. The “do not publish” and “no disclosure required” outcomes show an instruction instead — there is nothing to copy for either.
When to use it
- Before filing anything where a generative tool touched the words, images or headline.
- When drafting or updating a newsroom AI policy, as a starting map of which uses your policy needs to name.
- On a freelance commission, to check whether your disclosure meets what the commissioning outlet expects before it becomes an argument.
- When an AI-assisted image or illustration is being considered for a news story rather than a feature.
- When you are asked to sign off someone else’s AI-assisted copy and need a consistent line to apply.
What it does not do
- Not legal advice, and not a regulatory ruling. AI disclosure standards are unsettled and moving; the tool encodes one reasonable reading of current NUJ and IPSO positions, not a rule you can point to.
- Your own publication’s policy overrides this entirely, and many outlets and platforms now require disclosure in cases where this tool would say none is needed.
- It cannot see the output. “Verified” and “heavily rewritten” are your assessments, and the tool has no way to detect a fabricated citation, an invented quote or a hallucinated statistic that survived your edit.
- It says nothing about copyright in AI-generated material, about whether training a model on your work was lawful, or about the data protection implications of putting source material into a third-party tool — which for confidential source material is the more serious question.
- The sample wording is a template. Publishing a disclosure line that is not actually true of how the piece was made is itself an accuracy problem.
More tools
This is one of the free tools on UK JournoHub. See the full tools index for the rest.