Document icon with a quotation mark badge, representing Google's actual quoted AI content policy text

What Google’s AI Content Policy Actually Says

by

Ask five people what Google’s policy on AI content actually says and you’ll get five different confident answers. One says AI content is banned outright. Another says it’s fine as long as you disclose it. A third says there’s a magic percentage — under 30% AI-written, you’re safe. None of them are quoting an actual Google document, because most of what circulates as “Google’s AI policy” is a paraphrase of a paraphrase, several summarization steps removed from the four pages where Google actually wrote this down.

In my own pass through those four documents — not the SEO-blog recaps of them, the primary sources themselves — the real position is more specific than any of the paraphrases, and less dramatic than the panic that shows up every time Google runs a core update. Google has never published a single “AI content policy” page. What exists is a 2023 blog post that set the position, a spam-policies page that defines the actual violation, a content-quality page with a specific self-assessment framework, and a technical guidance page for anyone automating content at scale. Read separately, each sounds vague. Read together, they say something precise.

This post walks through what each of those four documents actually says, quoting the wording directly rather than summarizing it, so you’re working from the source instead of the fifth-hand version. It’s the reference companion to two other pieces on this site: how to tell if AI content is hurting your WordPress SEO covers the diagnostic checks to run against your own site, and this one covers what the policy those checks are based on actually says.

The Four Documents, Not One Policy Page

Google’s position on AI-generated content is spread across four separate pages on developers.google.com, published and updated at different times:

None of these four is titled “AI content policy.” That’s part of why the paraphrased version drifts so far from the source — there’s no single page to link to, so most explanations are built from memory of all four rather than a direct read of any of them.

The Core Position, In Google’s Own Words

The 2023 blog post states the position plainly: “Google’s ranking systems aim to reward original, high-quality content that demonstrates qualities of what we call E-E-A-T: expertise, experience, authoritativeness, and trustworthiness… Our focus on the quality of content, rather than how content is produced, is a useful guide that has helped us deliver reliable, high quality results to users for years.”

The post draws a direct historical parallel to make the point: “about 10 years ago, there were understandable concerns about a rise in mass-produced yet human-generated content. No one would have thought it reasonable for us to declare a ban on all human-generated content in response.” The production method was never the target. The quality bar was.

Then the actual rule, stated without hedging: “Using automation — including AI — to generate content with the primary purpose of manipulating ranking in search results is a violation of our spam policies.” And immediately after, the counterweight most paraphrases drop: “it’s important to recognize that not all use of automation, including AI generation, is spam. Automation has long been used to generate helpful content, such as sports scores, weather forecasts, and transcripts.”

That’s the whole position in two sentences: automation used to manipulate rankings is a spam violation, automation used to genuinely help readers is not, and Google has treated it this way for longer than generative AI has existed.

What Actually Counts as a Violation

The spam policies page defines the violation precisely, under the heading “Scaled content abuse”: content generated “for the primary purpose of manipulating search rankings and not helping users,” typically involving “large amounts of unoriginal content that provides little to no value to users, no matter how it’s created.” That last clause matters — the policy isn’t AI-specific. It targets a pattern (volume without value), and AI is simply the newest way to produce that pattern at scale.

Google lists specific examples of what counts, and only the first one is AI-specific:

  • Using generative AI tools “to generate many pages without adding value for users”
  • Scraping feeds or search results to generate pages, including through synonymizing, translating, or other obfuscation
  • Stitching or combining content from different pages without adding value
  • Creating multiple sites specifically to hide the scaled nature of the content
  • Producing pages where the content “makes little or no sense to a reader but contains search keywords”

The technical guidance page adds a detail almost no summary includes: this behavior maps to specific sections of Google’s internal Search Quality Raters guidelines — section 4.6.5 for scaled content abuse specifically, and section 4.6.6 for “main content created with little to no effort, little to no originality, and little to no added value.” Google is explicit that these guidelines “are not a guide to ranking first in Google” and that rater scores “don’t directly influence ranking” — the raters evaluate how well Google’s automated ranking systems are already performing, they don’t manually rank individual pages.

The “Who, How, Why” Self-Check

The helpful-content page gives a concrete framework for self-assessment, and it’s worth using literally rather than as a vague idea. Three questions, in Google’s own wording:

Who created the content — “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?” This is the same ground building E-E-A-T for a website covers in more depth — it’s not a separate concern from AI content, it’s the same author-identity signal applied to any content regardless of production method.

How it was created — specific to automation and AI: “Is the use of automation, including AI-generation, self-evident to visitors through disclosures or in other ways? Are you providing background about how automation or AI-generation was used to create content? Are you explaining why automation or AI was seen as useful to produce content?”

Why it was created — Google calls this “perhaps the most important question to answer about your content.” The distinction it draws: content made primarily to help people who’d land on the page directly is aligned with what the ranking systems reward; content made primarily to attract search visits is not, and if automation is used specifically to manipulate rankings, “that’s a violation of our spam policies” — the same line from the blog post, repeated here as the actual enforcement trigger.

The Disclosure Question

Google’s stance on disclosure is more permissive than most guides suggest, and it’s conditional rather than mandatory. From the FAQ: “AI or automation disclosures are useful for content where someone might think ‘How was this created?’ Consider adding these when it would be reasonably expected.” There’s no blanket rule requiring an “AI-assisted” label on every post — the test is whether a reasonable reader would specifically wonder how the piece was made.

One place the guidance does get specific: bylines. “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” — a direct answer to a question that comes up constantly, phrased as a caution rather than a hard prohibition, consistent with how the rest of the guidance reads.

The technical page adds one more layer for anyone publishing at real scale: e-commerce and structured-data content has firmer expectations than a blog post does. AI-generated images should carry IPTC DigitalSourceType metadata identifying them as algorithmically produced, and AI-generated product titles or descriptions should be identifiable as such in the underlying data — a requirement with teeth in a way “consider disclosing” isn’t.

Where This Meets 2026’s Enforcement

The policy itself hasn’t changed since 2023. What changed is how actively Google enforces it. The March 2026 core update named scaled content abuse as a direct enforcement target, and sites that had been publishing AI-generated pages at volume without real editorial oversight saw traffic drops in the 50-80% range — figures covered in more detail in how to use AI to write blog posts. Nothing about that enforcement contradicts the 2023 position; it’s the same “volume without value” rule the spam policy always described, applied with more force once scaled AI publishing made it a bigger real-world problem than it was when the position was first written.

What the Policy Does Not Say

Reading all four documents directly rules out most of the folklore that circulates about this topic. Google does not publish a maximum percentage of AI-written content a page is allowed to contain — there’s no threshold anywhere in any of the four pages, and any guide citing one is inventing a number. Google does not require a visible “written with AI” disclaimer on every AI-assisted post — disclosure is conditional on reasonable reader expectation, not universal. Google does not treat AI-generated and human-generated low-quality content differently — the scaled content abuse policy explicitly says the abuse pattern applies “no matter how it’s created,” so a manually written content farm is judged by the same standard as an AI-generated one. And Google does not run AI-generated pages through a separate detection classifier before ranking them — nothing in any of the four documents describes an AI-detection step in the ranking pipeline; the systems evaluate helpfulness signals, not authorship method.

Practical Tips

Apply the Who/How/Why framework literally, not loosely — write down the actual answers for a page before publishing it, the same way you’d fill in a checklist. Keep a real, verifiable byline on anything AI-assisted, since the “Who” question is the one Google is most explicit about wanting answered. Add a disclosure line only where a reader would plausibly ask “how was this made” — an AI-assisted round-up of statistics or a bulk-generated product description clears that bar; a blog post you drafted with AI and then rewrote in your own voice usually doesn’t. Treat the spam policy’s volume-without-value description as the actual line, not a vibe — if a batch of pages could be described honestly as “generated to occupy search real estate,” that’s the violation, independent of what tool produced them.

Common Mistakes

Citing a percentage-AI threshold as Google policy — it doesn’t exist in any of the four source documents, and a piece of content that’s 90% AI-drafted but genuinely useful and heavily edited is treated no differently than one that’s 10% AI-drafted and thin. Treating the Quality Raters guidelines as a ranking factor — Google states directly that rater scores “don’t directly influence ranking,” they’re a feedback loop for evaluating the automated systems, not a manual review layer. Assuming disclosure is always required — over-disclosing (an “AI-generated” banner on content that’s been substantially edited and fact-checked) isn’t wrong, but it isn’t what the guidance asks for either; it asks you to judge reasonable reader expectation. Conflating “AI content” with “spam” as if they’re the same category — the spam policy’s own wording treats scaled, low-value content as the violation regardless of authorship, which means a site of thin human-written pages is exposed to the identical rule.

When to Read This vs the Other AI Posts Here

This page is the primary-source reference — what Google’s own documents actually say. For a step-by-step audit of whether your own AI content is already hurting your rankings, how to tell if AI content is hurting your WordPress SEO is the diagnostic companion. If you’re deciding whether a third-party AI-detection score should factor into an editorial or hiring decision, AI content detection tools: what they actually measure covers why those tools aren’t part of Google’s own evaluation at all. For the editing discipline that turns AI drafts into content that clears the bar this post describes, see how to edit and fact-check AI-generated content. And if you’re choosing which AI writing tool to use in the first place, best AI writing tools for blog content compares what each one is actually built for today.

Conclusion

Strip away the paraphrasing and Google’s position on AI content is narrower and more consistent than the surrounding noise suggests: quality and usefulness are what get rewarded, volume produced primarily to manipulate rankings is what gets penalized, and that’s been true since well before generative AI existed — automation is just the latest way to produce the pattern Google has been policing for years. Read the Who/How/Why framework as an actual checklist, keep a real byline on what you publish, disclose where a reader would reasonably ask how something was made, and the policy stops being something to worry about and starts being something you’re already meeting.