AI Disclosure Statement Examples for Articles and Posts

A blank grey paper page with a dark strip pinned across the top and a smaller grey strip tucked under a square picture panel like a caption

The more the tool shaped what the reader receives, the higher up the page the line goes.

What a good disclosure line says

A disclosure line tells a reader how the piece was made, in words they can act on. Three facts do that. What the tool did. What a person checked. Who is accountable for the result. A line with all three is specific enough to trust. A badge that says AI-assisted tells the reader only that something happened.

What the tool did: name the task, such as summarising sources, writing a first draft or generating an illustration.
What a person checked: facts, quotes, figures, sources, or the whole text. Say which, because readers weigh a checked quote differently from a checked comma.
Who is accountable: the named author, the editor or the publication. Responsibility stays with a person; the tool cannot hold it.

Publishers that write formal policies land in the same place. Springer Nature's policy on AI use in manuscript preparation grades uses by risk, asks authors to clearly describe their use and confirm their accountability, and ends on a plain rule: authors remain fully accountable for everything submitted in their name. The logic carries over to a blog: grade by what the tool did, and keep a person's name on the result.

AI disclosure statement examples, level by level

The lines below are written for this guide and free to adapt. Each fits in one or two sentences, names the task, names the check and keeps a person responsible. Swap the bracketed parts and keep the wording fixed across posts, so readers learn it.

Level	Example line	Where to put it
No AI use	Researched, written and edited by [name]. No AI writing tools were used.	End note, only where your policy promises a statement on every post
Research help	[Name] used an AI tool to find and summarise background sources. Every source cited here was read and checked by the author.	End note, or a short line under the byline
Editing	An AI tool suggested grammar and clarity edits. The wording, argument and conclusions are [name]'s own.	End note or the policy page
Drafting with human rewrite	A first draft of this article was produced with an AI tool from [name]'s notes, then rewritten and fact-checked by [name], who is responsible for its content.	Directly under the byline
Mostly generated	This summary was generated by an AI tool from [source]. An editor checked names, figures and dates against the source before publication.	At the top, before the first paragraph
AI illustration	Illustration generated with an AI tool for this article. It does not show a real event or person.	Image caption, plus alt text that describes it
AI-edited photograph	Photo by [name]. Background extended with an AI editing tool; the subject is unaltered.	Image caption

The no-AI line is optional: most publications say nothing when no tool was used, and a statement on every post soon becomes noise. The editing line is the one publishers disagree about most. Springer Nature asks authors to describe even grammar and language editing; many newsrooms and blogs leave spelling and grammar tools out of disclosure entirely. Pick one rule, write it on your policy page and apply it every time.

Where on the page a disclosure belongs

The line should reach the reader before the part of the piece it describes. A reader who learns at the end that a tool drafted the article has already read it on other assumptions.

Near the byline or at the top: for drafting and generated text. This is where readers look to learn who wrote the piece, and the guide to byline schema in WordPress covers marking up the human author beside it.
In image captions: for generated or AI-edited pictures. A caption sits next to the thing it explains. The guide to labelling AI-generated images in link previews covers alt text and the share card.
As an end note: for research help and editing. The detail is there for readers who want it, without weighing on the opening.
On a policy page: for the rules behind all of it, linked from every line. The guide to AI-generated content in social feeds covers how to decide when you disclose at all.

Most readers meet a post as a share card first, and an end note never reaches the card. Wording and placement for titles, descriptions and first sentences are covered in AI disclosure lines that survive a share, so this guide stays with the page itself.

Lines to avoid

Line or habit	Why it fails	Better
This content may contain AI.	Says nothing about what the tool did or who checked it	Name the task and the check
AI-assisted, as a badge	Covers everything from a spelling check to a full draft	Use the line for the actual level
Disclosure only in the site footer	Readers never connect a footer to a specific post	Line on the post, policy in the footer
Written by AI, when a person rewrote it	Overstates the tool and hides the human work	Drafting with human rewrite line
Content was generated and reviewed.	Passive voice hides who did the review	Name the editor or author

The vague line is the most common failure. Readers cannot tell a lightly edited post from a generated one, so they discount both, and a writer whose work is later questioned has nothing specific to point to; the guide for anyone accused of posting AI content shows why a precise, dated line is worth keeping.

How disclosure relates to the rules

In the EU, the relevant law is the AI Act, Regulation (EU) 2024/1689. Its transparency duties sit in Article 50, and they are narrower than most summaries suggest. The European Commission's text of Article 50 requires those using an AI system to disclose deep fakes: image, audio or video content that has been artificially generated or manipulated. It also requires disclosure of AI-generated or manipulated text published to inform the public on matters of public interest. That text duty does not apply where the content has undergone human review or editorial control and a person or organisation holds editorial responsibility for publishing it. Where disclosure is required, Article 50 says it must be clear and distinguishable at the latest at the first exposure.

Read plainly, an edited publication that stands behind its articles is in a different position from a feed of unreviewed text. Separate duties in Article 50 fall on the providers of AI tools, who must mark synthetic output in a machine-readable way; that is the tool maker's job, and it does not replace a line a reader can see. This summary is not legal advice. If your publication is in scope, check the current text and take advice on your own case.

Search engines are a different matter, with no duty to label. Google Search Central's guidance on generative AI content asks publishers to fact-check and review AI output before publishing, including titles, descriptions and alt text, and suggests adding information on how content was created where that makes sense for the audience. It also warns that generating many pages without adding value may breach its spam policy on scaled content abuse.

A worked example: one month at a local news site

The numbers in this example are invented for illustration. A local news site publishes 40 items in a month. Under its policy, 22 involve no AI tool and carry no line. In 9, a reporter used a tool to summarise council papers before reading the originals; each gets the research help line as an end note. In 4, the subeditor ran an AI grammar pass; the policy page covers these and the posts carry no line. Three feature articles began as tool drafts built from the reporter's notes; each carries the drafting line under the byline, naming the reporter.

Two weekly sports results roundups are generated from league data and checked by an editor against the league tables; both open with the mostly generated line. Separately, 6 posts use AI illustrations, each captioned with the illustration line, and none shows a real person. The editor logs each line with the post, level and date. When a reader asks whether a feature was machine-written, the answer is already under the byline, with a name attached.

Pitfalls

Changing the wording from post to post. A fixed phrase per level is easier for readers to learn and for editors to apply.
Disclosing at the end what the reader needed at the start. A drafting line in a footnote arrives too late to matter.
Labelling an illustration in the text but not in its caption. Readers who skim see the picture and the caption, rarely both paragraphs around it.
Treating a disclosure line as a substitute for checking. The line describes the review; it does not perform it.
Promising in a policy what the posts do not deliver.

Disclosure is one signal among several that tell a reader who made a piece and how. The provenance section covers the others: bylines, author tags, image credentials and the share card.

Sources

Questions

What should an AI disclosure statement include?

It should include what the tool did, what a person checked and who is accountable for the result. One or two plain sentences cover all three.

Where should I put an AI disclosure on an article?

Put it near the byline when AI drafted or generated the text, and in the caption when it made an image. Research help and light editing can sit in an end note linked to a policy page.

Do I need to disclose using AI for grammar checks?

Usually not for a blog or news site, though some publishers ask for it. Whatever rule you choose, write it on your policy page.

Does the EU AI Act require every AI-assisted article to be labelled?

No. Article 50 requires disclosure of deep fakes and of AI-generated text published to inform the public on matters of public interest, and the text duty does not apply where the content had human review or editorial control and someone holds editorial responsibility. This is not legal advice.

Is a vague line like may contain AI enough?

No, because it tells readers nothing they can use. Use the line for the actual level, naming the task, the check and the person responsible.

Hannah Voss, Editor. Checks every guide against a working WordPress install and the networks' current documentation. Last reviewed September 2026.

About the editor and our methods