No. If an AI model drafted a post you publish on your own domain, no rule in Google's documentation and no US law requires you to label it. Google goes further than "not required" — it says that giving AI an author byline "is probably not the best way to follow our recommendation to make clear to readers when AI is part of the content creation process." The move that matters is the opposite of a disclaimer: put a real, named human in the byline — the person who read the draft, checked the facts and approved it — and describe your process where a curious reader can find it.
If that feels too easy, look at practice. Researchers analysing 186,000 articles from 1,500 American newspapers published in summer 2025 found roughly 9% were partially or fully AI-generated; a hand audit of 100 flagged articles turned up five disclosures. The norm founders lose sleep over barely exists. The one that does — the one every public AI-content scandal has actually turned on — is accountability: whose name is on it, and can that person answer for it?
Start with the hardest case, because if you clear it you clear everything below it. The EU AI Act's Article 50 transparency obligations became applicable on 2 August 2026, and Article 50(4) is unusually direct: "Deployers of an AI system that generates or manipulates text which is published with the purpose of informing the public on matters of public interest shall disclose that the text has been artificially generated or manipulated."
Two limits sit inside that sentence. The first is subject matter: the European Commission's FAQ defines "matters of public interest" as a specific list — politics and democratic processes, public administration, justice and law enforcement, fundamental rights, public security, public health, environmental protection, consumer safety, and economic, financial, political, scientific or cultural developments relevant to public debate. A post about frontmatter schemas isn't on that list.
The second limit is worth pinning to your wall. The obligation "shall not apply … where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content." The Commission defines those terms tightly: human review is "deliberate examination of the substance of the content by one or more natural persons possessing relevant knowledge and professional judgement," editorial control is the authority to "approve, alter or reject the substance of the text," and editorial responsibility means someone holds ultimate legal responsibility. Spell-checking doesn't count. In effect, the strictest AI transparency regime on earth says a label is what you owe when nobody is accountable.
And yes, it reaches you: Article 2(1)(c) covers deployers "located in a third country, where the output produced by the AI system is used in the Union." Your blog is read in the EU, so you're in territorial scope — almost certainly out of substantive scope, and covered by the carve-out anyway if a named person approved the post. The Commission's guidelines on transparency of AI-generated content were updated on 6 August 2026.
Google's position is softer than the AI Act's but points the same way. Its guidance on using generative AI content, last updated 10 December 2025, contains no disclosure mandate at all — only an invitation: "Sharing information about how a piece of content was created can help give your readers more context." Its older framing named the trigger precisely: disclosures "are useful for content where someone might think 'How was this created?'" That's a reader-expectation test, not a compliance test.
The more actionable document is Creating helpful, reliable, people-first content, also refreshed on 10 December 2025. It asks you to self-assess under Who: "Is it self-evident to your visitors who authored your content? Do pages carry a byline, where one might be expected? Do bylines lead to further information about the author or authors involved?" And under How: is the use of automation self-evident, are you providing background about how it was used, and are you explaining why it was useful?
Notice what's being asked. The byline question is conditional — where one might be expected. The automation questions ask for background and rationale, which is a described process, not a badge. A sentence like "our drafts are researched and written with AI assistance, then fact-checked and approved by a named editor before publishing" answers all three How questions at once. "AI-generated" answers none of them.
Keep this separate from the penalty question, which founders reliably conflate with it. Quality raters are told the Lowest rating applies when content is "copied, paraphrased, embedded, auto or AI generated, or reposted from other sources with little to no effort, little to no originality, and little to no added value" — effort and value are the test, and the guidelines say nothing about disclosure either way. If ranking risk is the real worry, our close reading of what Google's scaled content abuse policy really bans covers it properly.
If your blog emits Article structured data, the question returns in miniature. Google's Article markup guidance offers two author types — Person and Organization, no non-human option. Can you just credit the company? It satisfies "self-evident who authored this," but Google also asks whether bylines "lead to further information about the author," and a bare company name doesn't. So where a real person approved the post, name them and give the Person entry a url or sameAs that resolves.
Your own domain is the permissive case. Three other surfaces are not.
Someone else's platform. Medium requires that "AI-assisted text needs to be disclosed at the beginning of the story, within the first two paragraphs" and enforces it through distribution: disclosed AI writing is eligible for General Distribution, undisclosed writing reaches only your existing followers, and AI-generated writing has been barred from behind the paywall since 1 May 2024 either way. Syndicate and you inherit the destination's rules.
Ads. On 9 July 2026 Google began surfacing a "How this ad was made" panel in My Ad Center, showing whether ad creative was made or edited with AI. It covers paid advertisements, not organic content — but turn a post into ad copy and the answer changes.
Copyright registration. Register a work containing more than a de minimis amount of AI-generated material and you must disclose it and describe the human contribution. The Copyright Office's Part 2 report of 29 January 2025 held that "prompts may reflect a user's mental conception or idea, but they do not control the way that idea is expressed" — prompting alone isn't authorship, though human-authored elements stay protectable.
One law gets cited at founders wrongly: California's AI Transparency Act (SB 942, as amended by AB 853) binds a "covered provider" — an entity producing a generative AI system with over 1,000,000 monthly visitors or users. Publishing blog posts doesn't make you one.
Suppose you decide to disclose anyway. What does it buy you? Less than you'd hope — and it depends on what else you show.
The Reuters Institute's Generative AI and News Report 2025 — six countries, roughly 2,000 respondents each, fieldwork 5 June to 15 July 2025 — measured a clear "comfort gap". Only 12% were comfortable with news produced entirely by AI and 21% with a human in the loop, but 43% were comfortable when a human leads with some AI help, against 62% for entirely human-made news. Read that as a spec, not a verdict: audiences aren't rejecting AI in the pipeline, they're rejecting AI at the wheel. The distance from 21% to 43% is what a human who leads rather than rubber-stamps is worth.
Labels themselves carry a cost. A survey experiment with a nationally representative sample of 3,861 people found that explicitly labelling a news article as AI-generated reduced its perceived accuracy — and that increasing the salience of how AI was used reduced that penalty. Vagueness stings; specifics soften it.
Which leads to the most useful finding here. Toff and Simon, in the International Journal of Press/Politics (30 December 2024), found audiences perceive AI-labelled news as less trustworthy — but that "negative effects associated with perceived trustworthiness are largely counteracted when articles disclose the list of sources used to generate the content." The receipt beats the label. If you publish anything about how a post was made, make it the sources — which doubles as the structure that gets you cited by ChatGPT and AI Overviews.
Two caveats, since over-claiming here would be its own dishonesty. Every study above measures news, not B2B or developer-tools blogs; no published evidence says an "AI-assisted" note on a SaaS post costs you readers. And a small 2026 study of 40 participants found about two-thirds preferred detailed disclosures even though detail was the version that dented trust. Forty people is directional at best — but "readers don't want to know" isn't a claim the evidence supports.
Every AI-content scandal founders remember gets misfiled as a disclosure failure. Look at what actually happened in each.
Sports Illustrated, November 2023. Futurism found articles under authors like "Drew Ortiz" and "Sora Tanaka" whose profile photos traced to a marketplace selling AI-generated headshots, listed with descriptions such as "neutral white young-adult male with short brown hair and blue eyes." After Futurism approached the publisher, the bios vanished; the Arena Group attributed the content to a contractor and ended the partnership. The offence was an invented human, not an admitted machine.
CNET, January 2023. Roughly 78 articles ran over about two months under "CNET Money Staff", the AI involvement discoverable only on the author page. One compound-interest explainer claimed $10,000 at 3% annual interest earns $10,300 in the first year — it earns $300. CNET didn't catch it; Futurism did. The outlet later corrected 41 of 77 AI-written stories, better than half the run. A byline pointing at a department rather than a person is how nobody noticed for two months.
"Margaux Blanchard," August 2025. A fabricated freelancer placed AI-generated work in Wired and Business Insider, among others; editors eventually spotted invented sources, and both outlets said the pieces did not meet their editorial standards.
The pattern is consistent. Nobody was punished for saying "AI helped." They were punished for a byline that pointed at someone who could not be held responsible — and for errors a substantive human review would have caught.
It also answers the fear behind most defensive disclosure: that a detector will out you. The FTC alleged Workado promoted its AI Content Detector as "98 percent" accurate when independent testing showed 53 percent on general-purpose content — no better than a coin toss, in the agency's words — and researchers have found widely used GPT detectors "consistently misclassify non-native English writing samples as AI-generated". Evading detectors isn't a strategy, and neither is placating them. Being accurate and accountable is.
Here's the practical version, in the order it costs you effort:
Person entry in your Article schema with a url or sameAs that resolves — an author page, a company team page, a real profile.DigitalSourceType: TrainedAlgorithmicMedia requirement for AI-generated images sits under its ecommerce/Merchant Center policies, not general web guidance — an obligation for a store, good hygiene for a blog hero image.That list is, transparently, why Magic Share is built the way it is: the agent researches and drafts, checks that 100% of cited links resolve, and attaches a verification report — then the draft waits. A person approves it, and on a static site merging the pull request is the publish step, leaving a timestamped record of who signed off. Same shape as the AI Act's carve-out and Google's "Who." For the manual version of that review, our 20-minute fact-checking checklist for AI drafts walks through it link by link. And plainly: most posts here, including this one, are agent-drafted and human-approved.
Do I have to tell Google that a post was written with AI? No. Google's generative-AI guidance, last updated 10 December 2025, sets no disclosure requirement for organic web content — only a suggestion to share context where it helps your audience. Its systems assess quality and effort, not authorship method.
Can I list an AI tool as the author?
Not advisable. Google has said giving AI an author byline "is probably not the best way" to make AI involvement clear, and its Article structured data spec supports only Person and Organization. Name the person who reviewed and approved the post instead.
Does the EU AI Act apply to my US company blog? Territorially, probably yes — Article 2(1)(c) covers third-country deployers whose AI output is used in the Union. Substantively, probably not: the text duty covers content published to inform the public on matters of public interest, and doesn't apply where the content had human review with a named party holding editorial responsibility. This is general information, not legal advice.
Should I add an AI disclosure to posts I already published? There's no obligation to retrofit labels. The higher-value retrofit is a real byline with an author page that resolves, plus fixing stale facts and dead links — which is what readers and quality assessment actually respond to.
Strip away the jurisdictions and the studies and one idea survives: a byline means "I read this and I stand behind it." Regulators wrote that instinct into an exemption, Google wrote it into a self-assessment question, and three newsrooms learned it the expensive way. If the name on your post can honestly make that promise, you don't need a disclaimer. If it can't, no label fixes it — reading the draft does.
If you'd like drafts that arrive with every cited link verified and a report attached, waiting for exactly that human approval, see how Magic Share researches, drafts and verifies each post — then put your name on the ones you approve.
Magic Share researches, writes and fact-checks posts like this for any site — point it at your URL and review your first draft today.
Plant your first post