What Google now says about AI-written pages, and how to build a workflow around it

The question

Does Google penalize pages written with AI?

The short answer

Not for using AI. Google warns that generating many pages without adding value may violate its scaled content abuse policy, its helpful content page calls large amounts of AI text produced without manual oversight or curation "little to no effort," and it treats fabricated author profiles as deception.

Source: Search Central documentation updates, Google, Google Search's guidance on using generative AI content, Google, Creating helpful, reliable, people-first content, Google, Spam policies, Google

TL;DR: On 1 October 2026 Google updated its guidance on AI content. It still doesn't ban AI writing. It points to rater guidelines on low-effort content, and its helpful content page says using AI "to produce large amounts of text without manual oversight or curation" is little to no effort. Fabricated author profiles are called deception. Human review is necessary, but not enough on its own. The page also has to be accurate and add something useful.

Here's how a review can miss the real problem. A model drafts twenty service pages. Someone fixes the spelling and schedules them. Nobody checks whether the pages are accurate or whether they say anything a reader couldn't find elsewhere.

That review checked spelling, not accuracy or added value. The pages are now very well spelled.

On 1 October 2026, the Search Central changelog recorded an update: "Updated guidance on using generative AI content," with information from Google's Search Quality Raters guidelines. Google's helpful content page now carries sections on main content and authorship that cover the same ground. First, the headline some people missed: Google did not, in fact, ban AI writing. Please return your torches. Here's what the two pages say, my read on them, and the review workflow I use for AI-assisted pages.

What the guidance says

AI is allowed. The generative AI content guide says AI "can be particularly useful when researching a topic, and to add structure to original content."

Volume without value is the risk. The same guide says using AI "to generate many pages without adding value for users may violate" the scaled content abuse spam policy. That policy describes many pages generated mainly to manipulate rankings rather than help users, "no matter how it's created."

Raters have named categories. The guide points to rater guidelines section 4.6.5 (scaled content abuse) and 4.6.6 (main content with "little to no effort, little to no originality, and little to no added value"). Google says raters help evaluate its ranking systems, and their ratings "don't directly influence ranking."

Effort has a definition. The helpful content page lists four things raters judge: effort, originality, talent or skill, and accuracy. Under effort it says "using generative AI to produce large amounts of text without manual oversight or curation represents little to no effort." It adds that giving credit to other sources "doesn't replace the need for original effort."

Fake authors are deception. The same page says "fabricating creator profiles (such as by using AI-generated headshots, made-up names, or false credentials to make content appear as if it was written by human experts) is a form of deception."

Fact-check the metadata too. The AI guide says to "manually factcheck and review all AI-generated content," including title elements, meta descriptions, structured data, and image alt text.

My read

This is my interpretation, not Google's wording.

Human review is the floor, not the finish line. Google names effort, originality, skill, and accuracy. My rule: a quick skim isn't enough to check any of them. An editor who knows the subject, checks every fact, and adds what only they know gets much closer.

The authorship paragraph is worth acting on first. If your site presents content as written by an expert who doesn't exist, with a made-up name, a generated headshot, or a false credential, Google's docs describe that as deception. I'd fix false authorship before anything else. Dr. Jane Placeholder, 20 years of experience, generated last Tuesday, has to go.

Linking sources isn't enough. Summarizing five news articles, with links, is still a summary. The value has to come from something you add: a test you ran, a number from your work, a decision rule, a worked example.

A five-step review workflow for AI-assisted pages

This is the workflow I use and recommend. For each page, save four things: the note on what it adds, the sources checked, the editor's name, and the review date.

1. Start with what you add. Before anyone prompts a model, write down what this page will add: customer questions you've answered, results from your own work, a comparison you've actually done, photos. If you can't name anything, the page probably isn't ready.
2. Let AI draft, not decide. Give the model your notes, the primary sources, and the format. Use it to outline and write a first pass. Don't ask it to "write a blog post about X" from nothing.
3. Edit as the expert. Someone who knows the topic rewrites the draft. Cut generic paragraphs. Add the example, the caveat, the number. Cut any paragraph that adds nothing to the reader's understanding or decision.
4. Fact-check every claim, metadata included. Open each source and confirm the claim matches what it says. Check the title, meta description, structured data, and alt text the same way, since Google names all four.
5. Make authorship accurate. Add a byline where readers would expect one, and make sure the person or brand behind it is real. Where it helps, the AI guide suggests telling readers how the content was made.

Pre-publish checklist

Run this on every AI-assisted page:

• What useful information, analysis, or help does this page add beyond existing sources?
• Did someone with relevant knowledge edit it, not just skim it?
• Does every factual claim link to a source I opened before publishing?
• Are the title, meta description, structured data, and alt text checked?
• Is the author real, and is everything said about them true?
• For topics that could significantly affect people's lives or well-being, like health or money: is it consistent with expert consensus? Google's helpful content page says these "must be highly accurate."
• Would I publish this page if it got no search traffic at all?

If a whole batch of pages struggles with the first question, stop publishing that batch and review the existing pages against the scaled content abuse policy. If they fit its definition, Google's instruction is to exclude them from Search.

Where to start this week

Audit your last ten AI-assisted pages with the checklist and record a decision for each: keep, improve, merge, or remove. Fix false authorship and factual errors first. For pages with no clear added value, decide whether to improve, merge, or remove them. Improve means adding your own example, data, or decision rule, not more words. Adding words alone doesn't add value.

For the broader approach behind this checklist, see Practical AI without the theater.