AI Content Ethics: What Google Actually Requires You to Disclose

Google doesn’t require disclosure on every AI-assisted page, but it does require it when a reader would reasonably wonder how the content was created — and it flatly requires disclosure for AI-generated content on sensitive topics like health, finance, and news, regardless of how the content otherwise reads.

The confusion most businesses run into isn’t the rule itself, which is fairly clear once stated — it’s that the rule gets applied against a backdrop of near-constant, often contradictory public commentary about AI content policy, much of it from sources with no actual authority over Google’s guidelines or platform-specific ad policies. Sorting the genuine requirements from the speculation and the outdated advice still circulating from earlier, stricter periods of AI-content anxiety is where most of the actual confusion lives.

Creation Method Isn’t the Ranking Factor. Quality Is.

Google has been explicit that AI-assisted or AI-generated content can rank when it’s useful, original, accurate, and satisfies search intent — and equally explicit that using AI provides no special ranking advantage. That cuts both ways: content isn’t penalized for being AI-assisted, and it isn’t rewarded for it either. The quality bar is the same one that’s always applied. What’s changed is the disclosure expectation layered on top of it.

This is a genuinely different position than Google held a few years earlier, when the safest public assumption in the SEO industry was that any detectable AI involvement carried real ranking risk. That earlier caution pushed a wave of businesses toward either avoiding AI tools entirely or using them covertly and hoping detection tools wouldn’t flag the output. Both reactions now look like overcorrections in hindsight, chasing a penalty that Google’s own guidance says was never the actual mechanism. The mechanism was, and remains, content quality — AI just changed how cheaply low-quality content could be produced at scale, which is what actually triggered Google’s tightened spam and quality enforcement around AI-generated content in 2024 and 2025. Businesses that focused on the real target (quality, verified by genuine expertise) rather than the perceived target (avoiding AI detection) ended up in a stronger position regardless of how the specific policy language evolved.

Try It: Do You Need to Disclose?

Where the Rules Get Strict

The clearest hard line sits in advertising, not organic content: all ads using AI-generated images, voices, or text must carry an “AI Generated” label, and deepfake-style content depicting real people is prohibited outright. YouTube specifically requires AI disclosure inside the video itself, not just in the description. Organic blog content has more latitude, but the direction of travel across jurisdictions is toward mandatory disclosure becoming the default, not the exception — several regions are actively legislating this now.

The advertising rules in particular have real enforcement teeth behind them in a way that organic content guidelines currently don’t. Google Ads and Meta’s ad platforms both run automated and manual review processes specifically flagging undisclosed AI-generated creative, and violations can result in account-level penalties that go well beyond a single ad being rejected — repeated violations put the whole advertiser account’s standing at risk. Organic content policy violations, by contrast, are enforced indirectly through ranking and visibility rather than through a formal takedown or account penalty in most cases, which is part of why the two categories feel like they carry different weight even though both are described as “requirements” in Google’s guidance.

Disclosure Requirements at a Glance

The regulatory direction is consistently toward more disclosure, not less, and toward disclosure becoming a legal requirement rather than a platform-specific best practice. The EU’s AI Act includes transparency obligations for AI-generated content that are being phased in, and several individual jurisdictions are drafting their own labeling requirements independent of the EU framework. A business publishing content today with no disclosure practice at all is building on ground that’s shifting under it — what’s merely “recommended” in one category can become “required” with limited notice as legislation catches up to the technology.

Our Actual Position on This

We use AI tools in parts of our own content workflow — research, drafting speed, structure — and we don’t think that’s something to hide or apologize for. What we won’t do is publish AI-drafted content without a person who actually understands the topic reviewing, correcting, and taking ownership of it before it goes live. The ethical line isn’t “did AI touch this content.” It’s “does a real, accountable expert stand behind what’s published.” That’s the standard Google’s E-E-A-T signals are increasingly built to detect anyway, so it’s not just an ethics position — it’s also, in our experience, the one that ranks.

Clients occasionally ask us to disclose exactly what percentage of a given piece was AI-assisted, expecting a precise number the way one might expect a nutrition label. We push back gently on that framing, because it implies a level of precision that doesn’t reflect how the actual workflow operates — a piece might start from an AI-generated outline, get substantially rewritten by a human writer, get fact-checked and corrected by a subject-matter reviewer, and go through two further editing passes, at which point asking “what percent is AI” is a bit like asking what percent of a renovated house is the original structure after several decades of updates. The more useful question, and the one we actually answer for clients, is “who is accountable for the accuracy of what’s published” — and the answer to that is always a named person on our team or the client’s, never “the AI.”

What “Genuine Review” Actually Means in Practice

“A human reviewed it” gets thrown around as a compliance checkbox, and it’s worth being specific about what a genuine review actually requires versus what merely looks like one. A superficial review — skimming for typos and obvious factual errors, then publishing — doesn’t meet the bar Google’s E-E-A-T signals are designed to detect, and it doesn’t meet the ethical bar either. A genuine review means someone with real, verifiable expertise in the topic checks the substantive claims against their own knowledge, catches the subtle inaccuracies that read as plausible but are wrong (a persistent failure mode of AI-drafted content, which tends to state incorrect specifics with the same fluent confidence as correct ones), and is willing to put their name or their organization’s professional reputation behind the published result. That last part — accountability — is the piece a lot of AI-content workflows skip entirely, publishing under a generic “Team” byline that diffuses responsibility for accuracy across nobody in particular.

The Reputational Risk Beyond Search Rankings

Framing this purely as an SEO compliance question undersells the actual risk. A business that publishes AI-drafted content with a confidently wrong factual claim on a health, financial, or legal topic faces liability and reputational exposure that has nothing to do with whether the page ranks well — and everything to do with whether a reader acted on bad information and can trace it back to the publisher. We’ve seen this specifically in the financial and legal services clients we work with, where a single confidently-wrong claim about a regulatory requirement or tax threshold, however plausible-sounding, can do more lasting brand damage than years of good content can offset. The SEO disclosure question and the liability question point toward the same practical answer — genuine expert review before publication — but it’s worth understanding they’re two separate risks compounding on each other, not one risk with two names.

A Workable Internal Policy

Rather than a blanket “no AI” or “AI is fine” rule, the more defensible internal policy separates the question by stage and by topic sensitivity. Using AI for early-stage research, outlining, and first-draft structure is low-risk regardless of topic, since a competent human writer is still shaping and verifying every claim before publication. Using AI to generate the final published sentences on a YMYL topic without a named, credentialed reviewer attached is high-risk regardless of how good the output reads, both from a Google policy standpoint and a liability standpoint. Everything in between — general blog content, opinion pieces, lower-stakes commercial content — sits on a sliding scale where the right level of human involvement depends on how much a wrong claim could actually cost the business or the reader if it slipped through. Building this as an explicit, written internal policy, rather than leaving it to individual writers’ judgment case by case, is the difference between a defensible position if a regulator or a reporter ever asks about it, and an ad hoc practice nobody can clearly explain.

How Swiss and EU Rules Interact

Switzerland isn’t an EU member, but Swiss businesses selling into EU markets, or operating EU-facing digital properties, need to track EU AI Act transparency obligations regardless of domestic Swiss law, since the obligations attach to the audience being reached, not just the company’s home jurisdiction. Switzerland’s own regulatory approach to AI content has moved more cautiously and with less prescriptive detail than the EU’s so far, which means a Swiss business operating purely within the domestic market currently faces a lighter formal requirement than one with EU customers — but “currently” is doing a lot of work in that sentence, given how quickly this area is moving. Building a disclosure practice now that would satisfy the stricter EU standard, rather than the current lighter Swiss-domestic standard, avoids having to retrofit a compliance practice under time pressure once Swiss rules inevitably tighten to track the broader European direction.

A Worked Example: The Plausible-Sounding Wrong Number

A financial services client’s marketing team drafted a blog post explaining a specific Swiss pension contribution threshold using an AI writing tool, intending it as a quick informational piece rather than formal advice. The draft read fluently and confidently, citing a specific franc figure for the threshold — a figure that was subtly wrong, off by an amount that would have meaningfully changed a reader’s actual contribution planning if they’d acted on it directly. Nobody on the marketing team caught the error before it was scheduled for publication, because the number looked exactly as plausible as the correct one would have, and nobody involved happened to have the specific figure memorized well enough to flag it from memory alone.

It was caught only because the client’s internal review policy required a credentialed compliance reviewer to sign off on any content mentioning specific regulatory figures before publication — a policy that existed for exactly this scenario. The reviewer cross-checked the figure against the current official threshold, caught the discrepancy in about ten minutes, and the correction went out before publication rather than after a reader had potentially acted on bad information. This is precisely the scenario a “workable internal policy” is meant to catch: not sloppy or obviously wrong AI output, which most reviewers would catch instinctively, but confidently-stated, plausible-sounding errors that require someone with actual subject knowledge, not just a general edit pass, to verify against a real source before anything goes live.

Related Guides

• how YMYL rules change SEO for fintech and finance sites — why finance and fintech content faces a stricter quality bar.
• why most guest posting “opportunities” are worthless — how to tell a legitimate guest posting opportunity from a worthless one.
• why most top-ranking pages now show author credentials — the brand-authority signals increasingly built into ranking pages.
• our full guide to semantic SEO and topical authority — the entity-based approach behind most of the strategies discussed here.

Frequently Asked Questions

Want content that’s genuinely expert-reviewed and accountable, not just AI-polished and hoping nobody checks? See our content strategy service.

References

• RankAI — Google Policy on AI Content: 2026 SEO Compliance Guide
• Skycrumbs — AI Content Labeling Laws in 2026