What Google Says About AI-Written Content and Scaled Content Abuse
What Google's spam policies, people-first guidance and generative AI guide say about AI-written pages, what changed in 2026, and a test to run before publishing a batch.
On this page
- What do Google's three documents say about AI-written pages?
- How does AI-generated content quality affect search rankings?
- Which of these rules changed in 2026?
- What did Google update about AI content on October 1, 2026?
- Was there a helpful content update about AI writing in October 2026?
- What does the people-first page ask about first-hand expertise?
- Documented policy and detection folklore, side by side
- Is AI at scale a worse bet than human writing for SEO?
- A pre-publish test for a batch of AI-assisted pages
- Sources
Google Search Central's guidance on scaled content abuse and AI content is spread across three documents, and all three make the same call: a page is judged on why it exists and what it adds, whatever tool produced it. Our view is that most arguments about AI detection are aimed at a rule Google never wrote. Every quote and date below comes from pages we read on October 7, 2026.
What do Google's three documents say about AI-written pages?
The spam policy is the document that sets out a violation and its consequence. Google defines the offence in two sentences: "Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users. This abusive practice is typically focused on creating large amounts of unoriginal content that provides little to no value to users, no matter how it's created." The first example it lists is "Using generative AI tools or other similar tools to generate many pages without adding value for users", and the page states the stakes near its top: "Sites that violate our policies may rank lower in results or not appear in results at all."
| Document | A sentence that settles the question | Dates on the record |
|---|---|---|
| Spam policies, scaled content abuse | "If you're hosting such content on your site, exclude it from Search." | Policy added March 5, 2024; page stamped 2026-08-28 |
| Guidance on using generative AI content | "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." | Page added May 21, 2025; updated October 1, 2026 |
| Creating helpful, reliable, people-first content | "If you use automation, including AI-generation, to produce content for the primary purpose of manipulating search rankings, that's a violation of our spam policies." | Page stamped 2026-10-05 |
The added and updated dates come from the documentation changelog kept by Google Search Central.
How does AI-generated content quality affect search rankings?
AI-generated content quality reaches search rankings by the same two routes as any other content, which is the design of Google's scaled content abuse policy and its people-first guidance. One route is ordinary ranking, where a 2023 FAQ from Google says "Using AI doesn't give content any special gains." The other is spam enforcement, which the policy page describes this way: "We detect policy-violating practices both through automated systems and, as needed, human review that can result in a manual action."
The quality rater guidelines get cited as if they were ranking rules. The generative AI guide from Google says rater ratings "don't directly influence ranking."
Which of these rules changed in 2026?
Google Search Central's AI-generated content guide, its helpful content guidelines on people-first content and the reach of the spam policies all changed in 2026, and we found no changelog entry that rewrites the definition of scaled content abuse.
| Date in 2026 | What changed | How we know |
|---|---|---|
| April 14 | "Google may now use spam report submissions to take manual action." | Changelog |
| May 15 | Google "Clarified that our spam policies also apply to generative AI responses in Google Search." It also published a guide to optimizing for generative AI features | Changelog |
| October 1 | The generative AI guide gained a passage on hallucinations and manual fact checking | Changelog, page comparison |
| By October 5 | The people-first page gained a "Main content" section and a paragraph on deceptive authorship | Page stamp, page comparison |
The May entry matters to anyone tracking AI share of voice, because the policy page now counts "attempting to manipulate generative AI responses in Google Search" as spam. The guide published that day by Google, as it reads now, extends the scaled content rule to query variants. It says that creating separate content for every variation of how people might search, fan-out queries included, primarily to manipulate rankings or generative AI responses "violates Google's scaled content abuse spam policy."
What did Google update about AI content on October 1, 2026?
The Google Search Central update on AI content dated October 1, 2026 is an edit to the generative AI guide, logged as "Updated the using generative AI content guide with information from the Search Quality Raters guidelines." An Internet Archive capture from September 20, 2026 already carried the pointer to rater guideline sections 4.6.5 and 4.6.6. New since that capture is a passage on hallucinations, ending with the manual fact-check sentence in the first table.
The people-first page from Google changed as well. Its stamp reads 2026-10-05, and as of October 7 we found no changelog entry for the edit. Set against an Internet Archive capture from September 2, 2026, today's page adds a "Main content" section and an authorship paragraph. The section defines effort as an attribute Search Quality raters are trained to evaluate and says, "Conversely, automatically generating pages from feeds or using generative AI to produce large amounts of text without manual oversight or curation represents little to no effort." The paragraph calls fabricated creator profiles, such as AI-generated headshots, made-up names or false credentials, a form of deception.
Was there a helpful content update about AI writing in October 2026?
We found no Google helpful content update on AI writing in October 2026; the helpful content update's AI writing guidelines sit on the people-first page, and the update itself no longer runs as a separate event. A changelog entry dated August 15, 2024 reads: "Also, the helpful content system became part of core ranking systems (as we previously shared in March 2024)." On the Search Status Dashboard from Google, the last entry named a helpful content update is dated 14 Sep 2023. Its ranking history for 2026 lists a Discover update, two core updates (March 27 and May 21) and four spam updates (March 24, June 24, August 18 and September 24). None is named for AI content, and the September spam update still had no duration listed on October 7.
On disclosure, the people-first page gives advice: "Consider adding these when it would be reasonably expected." The one labeling requirement we found is for ecommerce, where the generative AI guide says Google Merchant Center's policies require AI-generated images to carry IPTC metadata and AI-generated product data, such as titles and descriptions, to be labeled as AI-generated.
What does the people-first page ask about first-hand expertise?
Google Search Central's helpful content page ties people-first content to first-hand expertise in one question: "Does your content clearly demonstrate first-hand expertise and a depth of knowledge (for example, expertise that comes from having actually used a product or service, or visiting a place)?" A second question accepts review as well as authorship: "Is this content written or reviewed by an expert or enthusiast who demonstrably knows the topic well?" By our reading, a model-drafted page that a practitioner corrects and extends fits that wording, and a page nobody with experience has touched fails it.
The generative AI features guide makes the same demand from the other direction: "Don't just recycle what others on the internet have already said, or could easily be produced by a generative AI model." That guide speaks only for Google Search, but we hold GEO work and any effort to get cited by ChatGPT, Perplexity and AI Overviews to the same sentence.
Documented policy and detection folklore, side by side
| Claim in circulation | What the documents say |
|---|---|
| Google demotes text because a model wrote it | "Appropriate use of AI or automation is not against our guidelines." |
| A low AI-detector score makes a batch safe | We found no detector score, threshold or safe share of AI text on any page cited here |
| The method has to be established before a page is rated as scaled abuse | "Even if you are unsure of the method of creation, e.g. whether or not the page is created using generative AI tools, you should still use the Lowest rating when you strongly suspect scaled content abuse after looking at several pages on the website." |
The last row quotes section 4.6.5 of the rater guidelines, in the September 11, 2025 edition published by Google. On how spam gets found, the 2023 FAQ offers one sentence: "We have a variety of systems, including SpamBrain, that analyze patterns and signals to help us identify spam content, however it is produced." A detector score measures something the written policy never asks about.
Is AI at scale a worse bet than human writing for SEO?
The trade-offs of AI-generated content at scale versus human-written content, as far as SEO quality goes, come down to review capacity, because Google's pages grade the effort behind a page whoever typed it. The people-first page flags human mass production as well as automation: "Is the content mass-produced by or outsourced to a large number of creators, or spread across a large network of sites, so that individual pages or sites don't get as much attention or care?" A freelancer farm fails the same test as an unreviewed model run.
Scale buys coverage and speed. Its price is a manual fact check on every page, metadata included, plus an original contribution that someone has to supply. A plan with more pages than review hours is the exposed one, and that arithmetic belongs in any decision on how to split a content budget between SEO and AI answer visibility.
A pre-publish test for a batch of AI-assisted pages
Run the test on the batch as a whole, since section 4.6.5 has raters judge scaled abuse across several pages of a site.
| Check | Drawn from | A page passes when |
|---|---|---|
| Purpose | People-first page, "Why" | It would be worth publishing with no search traffic at all |
| Added value | Spam policy example | A reviewer can name what it has that competing pages lack |
| Variants | Generative AI features guide | It answers a distinct question, with phrasing variants merged |
| Accuracy | Generative AI guide | A person checked every claim, title, meta description, structured data field and alt text |
| Oversight | People-first page, "Effort" | A record shows who curated it and what they changed |
| Authorship | People-first page, "Who" | The byline is a real person with real credentials |
A hypothetical batch shows the mechanics. Take 40 AI-assisted comparison pages. The variants check finds 12 that rephrase 4 questions, so they merge into 4 pages and 8 URLs are never created. Of the other 28, a reviewer cannot name the added value on 6, and those wait until someone with product experience supplies it. The remaining 22 and the 4 merged pages, 26 in all, pass to the accuracy check and publish. Every figure here is an assumption for illustration; we found no pass rate or sample size on the Google pages cited here. For failing pages that are already live, the policy's instruction is the sentence in the first table.
Across all three documents the offence is a pile of unreviewed, interchangeable pages, and a team that can show its review trail is making the argument Google's own wording invites. Further resources: our AI search optimization service and the AI visibility checker.
Sources
- Google, Spam policies for Google web search
- Google, Guidance on using generative AI content
- Google, Creating helpful, reliable, people-first content
- Google Search Central, Latest documentation updates
- Google, Optimizing your website for generative AI features
- Google, Guidance about AI-generated content (February 8, 2023)
- Google, Search Quality Evaluator Guidelines (September 11, 2025)
- Google, Search Status Dashboard, ranking history
- Internet Archive, generative AI guide as captured September 20, 2026
- Internet Archive, people-first page as captured September 2, 2026