You have probably read both claims by now. One says Google is fine with AI-assisted content, the other says AI pages get deindexed. You are publishing anyway, and the quiet worry is that the pages you shipped last quarter are costing you rankings right now.
This article does not settle that with vendor claims. It reads Google’s own published pages, the ones that define what gets a page demoted or removed, and turns them into a standard a small team can actually run. Every term you may have seen thrown around without explanation, including manual action, scaled content abuse and helpful content, gets defined where it first appears.
If you are scanning for one question, jump to it:
Whether AI authorship itself is penalized
What Google actually penalizes
Whether a penalty is a manual action, an algorithm change, or just not ranking
What Google means by helpful, people-first content
Where E-E-A-T fits
What publishing standard a one- or two-person team should run
How to audit pages you already published
Key Takeaways
Google’s guidance does not prohibit AI authorship; the trigger is content produced at scale to manipulate rankings, however it was produced.
A manual action is a human reviewer’s decision recorded in Search Console, and Google’s help page never lists “AI content” as a category.
Google publishes no page-count threshold for scaled content abuse, so “how many AI pages am I allowed” has no numeric answer.
E-E-A-T is not a score and rater ratings do not directly affect rankings, so no tool can tell you your E-E-A-T number.
Does Google penalize AI-generated content just because AI wrote it?
No. Google does not penalize AI-generated content for being AI-generated, and its own guidance says so in plain language. The page that answers this question most directly is Google’s guidance on AI-generated content, last updated 1 October 2026, which treats generative AI as a legitimate research and structuring tool and names only one failure mode: generating many pages without adding value for users, which it says may violate the scaled content abuse policy.
The myth that authorship is the violation has a long tail. It survives because “penalty” gets used loosely for four different things, and because a traffic drop feels like a verdict. Google’s February 2023 announcement, Google’s February 2023 announcement, set the frame that still holds: Google’s focus is on the quality of content rather than how content is produced, and not all use of automation is spam. The same post says using AI gives content no special gains either way.
“Appropriate use of AI or automation is not against our guidelines.” Google Search Central, 8 February 2023
The September 2025 rater guidelines carry the same logic into how human evaluators judge pages: using generative AI tools alone does not determine the level of effort or the Page Quality rating, because those tools can produce both high and low quality work.
What does Google actually penalize?
Google penalizes intent and scale, not the tool. The enforceable rule is the scaled content abuse policy, which defines the practice as “many pages are generated for the primary purpose of manipulating search rankings and not helping users,” typically large amounts of unoriginal content that provides little or no value, “no matter how it’s created.”
The examples on that page are worth reading as scenarios rather than abstractions. Generating many pages with generative AI without adding value is one. Scraping feeds or search results and running automated transformations is another. So is stitching content from different pages, running multiple sites to hide the scale, and publishing keyword-stuffed pages that make little sense to a reader. A 40-page programmatic location set with swapped city names and no local substance sits squarely in the first category.
When Google rewrote the policy in March 2024, it made the method-agnostic framing explicit: the policy covers content “produced through automation, human efforts, or some combination.” That announcement also states the consequence, that sites violating spam policies “may rank lower in results or not appear in results at all.” Danny Sullivan’s April 2025 warning repeated that how the scaling is done does not matter to enforcement.
Policy example | The same practice in plain language |
|---|---|
Generative AI used to generate many pages without adding value | A template filled with a new keyword per page, no original research |
Scraping feeds or search results with automated transformations | Rewriting competitors’ posts with synonyms and republishing |
Stitching content from different pages | A “guide” assembled from paragraphs of other sites’ articles |
Multiple sites hiding the scaled nature | The same content network spread across several domains |
Keyword-stuffed pages that make little sense | Pages written for the query string, not for a reader |
How many AI pages you are allowed to publish has no published answer. Google gives no numeric threshold for “many pages” or “large amounts,” so any vendor quoting a page count is inventing it. The test is purpose and scale, judged together.
Is a penalty the same as a manual action, an algorithm change, or just not ranking?
They are four different things, and only one of them is a human decision. A manual action is issued when a human reviewer at Google has determined that pages on a site are not compliant with the spam policies, as Google Search Console’s Manual actions report explains. That report is where the decision is recorded, alongside the Search Console message center.
An algorithmic action is the automated equivalent: Google’s systems detect and demote spam without a person reviewing the site. A core or spam update is a change to ranking systems that can move many sites at once with no action against any of them. And “just not ranking” is the ordinary case, where a page is indexed, eligible, and simply less useful than the alternatives.
Google’s help page on manual actions is blunt about how invisible this is: most manual actions result in pages or sites being ranked lower or omitted from search results without any visual indication to the user. That is why a traffic drop alone cannot tell you which of the four you are looking at. The report can. It also never lists “AI content” as a category; the closest entry is “Major spam problems,” which names scaled content abuse among aggressive techniques.
Google’s EU transparency disclosure puts the ratio in perspective: 318,012 manual actions against spam policies versus 38,837,715 algorithmic actions in the reporting period, with 39,125,118 unique domains affected by the combined enforcement. Read that as one regulatory reporting period, not a yearly total, and note that Google publishes no figure for scaled-content-abuse manual actions specifically. If someone tells you a specific number of sites were hit for AI content, ask which report it came from.
What does Google mean by helpful, people-first content?
Helpful content is content created primarily for people rather than to manipulate rankings, and Google gives you three questions to test it. Google’s people-first content guidance frames the self-assessment as Who, How and Why: who authored the piece, how it was created, and why it exists. AI use belongs in the “How” answer, which is why the page lists automated, AI-generated and AI-assisted content as things a reader may reasonably want explained.
The “Why” carries the enforcement weight. The same people-first guidance states that using automation, including AI generation, to produce content for the primary purpose of manipulating search rankings is a violation of the spam policies, and that AI or automation disclosures are useful where someone might wonder how a piece was created. Disclosure is framed as useful context, not a requirement, for ordinary editorial content.
Google’s self-assessment questions turn “helpful content and AI” from a slogan into checks you can run on a draft: does it provide original information, reporting, research or analysis; a substantial and complete description of the topic; insight beyond the obvious; and substantial value rather than a rewrite of other sources. A draft that reads fine can still fail all four. Google describes signals and questions here, not a formula, so no checklist converts a weak draft into a strong one.
Where does E-E-A-T fit, and is it a score Google ranks on?
E-E-A-T is not a score, and there is no number to optimize. Google’s description of E-E-A-T says its systems identify a mix of factors that help determine which content demonstrates experience, expertise, authoritativeness and trustworthiness, that trust is the most important of these, and that E-E-A-T itself is not a specific ranking factor. The factors are inputs to automated systems, not a grade you can look up.
The rater programme is where the “score” myth comes from, and Google is explicit about its role. Google’s own explanation of its rater program describes more than 10,000 raters who use the public guidelines to assess result quality, and states that their ratings do not directly impact how a page or site appears in Search; they measure how well the systems are working.
“The ratings they provide don’t directly impact how a page or site appears in Search.” Google, rater guidelines overview
The guidelines still matter as a reading list. The October 2026 update to Google’s AI content guidance points publishers to sections 4.6.5 and 4.6.6 of the rater guidelines, covering scaled content abuse and main content created with little effort, originality or added value, while repeating that the guidelines are not a guide to ranking first and that ratings do not directly influence ranking. The edition to read is the 11 September 2025 version; no 2026 edition has been published.
What publishing standard should a one- or two-person B2B team actually run?
Run four checks before publish, and accept that none of them makes a page immune. The standard below comes from Google’s own pages, not from a tool, and it fits a team where one person writes and the other reviews.
Check 1: every number and quote traces to a source. Google’s warning about accuracy explains why: generative models don’t retrieve facts, but predict a likely sequence of words from training data, so outputs can contain inaccuracies. The same page calls manual fact-checking critical before publishing and extends the review to titles, meta descriptions, structured data and image alt text, all of which can appear in Search results.
Check 2: the page answers a question a named reader has. Run the Who, How and Why questions from the people-first guidance against the draft, and write the “How” answer down even if you never publish it.
Check 3: the batch is not a template. If a set of pages differs only by a swapped term, that is the scaled content pattern, whatever produced it.
Check 4: consider a short disclosure. Google’s disclosure guidance suggests explaining how content was created where readers would reasonably wonder. There is a supporting reason beyond search: Article 50 of the EU AI Act requires disclosure for AI-generated text published to inform the public on matters of public interest, and exempts content that has undergone human review or editorial control with a named person holding editorial responsibility. Documented editorial control is the same mechanism Google’s guidance rewards, so the two point the same way for a small team.
What this standard does not buy: a ranking, a guarantee against a manual action, or immunity from an algorithm update. It buys a defensible position, which is what a two-person team can actually hold. AI-assisted content for B2B marketers works best where the model does research, drafting and repurposing while a person keeps judgment, the same division of labour described in using AI to increase brand awareness.
How do I audit the AI-assisted pages I already published?
Start in Search Console and work outward, because the presence or absence of a manual action changes everything that follows. Open the Manual Actions report first. If an action is listed, the remediation path is prescribed: fix the issue on all affected pages, since fixing only some earns no partial return, then request review and describe the exact quality issue, the steps taken and the outcome, per Google’s reconsideration guidance. The same page notes that most reconsideration reviews take several days or weeks.
If no action is listed, you are not looking at a manual penalty, and the work is editorial. Pull the pages that lost the most impressions, group them by template, and ask of each group whether it would survive the scaled content test: does each page carry something the others do not? Where the answer is no, consolidate rather than rewrite, because a rewritten template is still a template.
The honest gap is recovery timing. Google publishes no timeline for how long an algorithmic demotion lasts, and the “several days or weeks” figure applies only to reconsideration reviews of manual actions. Anyone promising a recovery window is guessing.
Frequently Asked Questions
Does Google penalize AI-generated content?
Not for being AI-generated. Google’s guidance treats generative AI as a legitimate tool and targets content generated at scale to manipulate rankings, regardless of whether automation, humans or both produced it. A page can be entirely AI-drafted and still comply, and a human-written page can still violate the scaled content abuse policy.
Will an AI detector score get my page penalized?
No. Google does not use third-party AI detectors as a ranking input, and its guidance says AI use alone does not determine Page Quality. Detector scores also vary by tool: Graphite’s 31,493-keyword analysis reported a 4.2% false-positive rate for the detector it used, which is enough error to make a single score a poor basis for deleting a page.
Do I have to label AI-assisted posts?
For ordinary editorial content, Google says disclosure is useful where readers would wonder how a piece was created, not mandatory. The hard labelling rules apply to ecommerce: AI-generated images in Merchant Center must carry IPTC DigitalSourceType metadata and AI-generated product data must be labelled as such. In the EU, Article 50 adds a disclosure duty for AI text on matters of public interest, with an exemption for documented editorial control.
My traffic dropped. Was I penalized?
Check the Manual Actions report before assuming anything. Most manual actions demote pages without any visual indication, so the report is the only reliable signal, and its absence means the drop came from ranking systems or from competitors rather than a human decision. A core update can move a site with no action taken against it at all.
If AI content is allowed, why did my page not rank?
Being allowed and being chosen are different tests. Google’s AI features documentation notes that meeting every requirement does not mean Google will crawl, index or serve a page, and the people-first questions ask whether a page offers original information and insight beyond the obvious. A compliant page that restates what already ranks has cleared the policy and still lost the comparison.
Conclusion
The answer to whether Google penalizes AI-generated content is that it penalizes purpose and scale, not authorship, and it says so across five separate documents: the AI content guidance updated 1 October 2026, the spam policies, the 2023 and 2024 announcements, and the rater guidelines. The empirical studies that seem to contradict this are detector-dependent and point in different directions, which is a reason to read them as context rather than rules. Google’s own enforcement data does not corroborate an authorship penalty at all.
What that leaves a small team is a standard rather than a reassurance. Trace every number, answer a real question, avoid the template batch, and document your editorial control. None of it makes a page immune, and no vendor can sell you immunity.
The next step is small and free. Open Search Console, go to the Manual Actions report, and confirm whether anything is listed. Then pick one batch of AI-assisted pages you published this year and run the four checks against them. That audit tells you more about your real risk than any detector score, and it takes an afternoon.
If you want to go further into how visibility works once the quality bar is cleared, the comparison of SEO versus GEO is the logical next read.