On October 1, Google quietly updated its guidance on using AI to write web content. No press tour, no keynote, just a documentation page that got a little more specific about what it considers lazy. Meanwhile its September spam update has been rolling through search results since September 24, and as I write this it still isn’t done.
So if your website was written by a chatbot in one afternoon, this is a good week to reread it. Slowly. Like a stranger would.
The short version
Google refreshed its page on using generative AI content and pulled in material from its Search Quality Rater Guidelines. In plain words: the rules didn’t flip, but the explanation of what “bad” looks like got sharper.
The guide says using AI to crank out many pages without adding value for users may violate Google’s spam policy. It also flags that AI tools make stuff up (it calls them hallucinations, which is a polite word for confidently wrong) and says manual fact checking before publishing is critical. And it now references the rater guidelines on scaled content abuse and on pages made with little effort, little originality and little added value.
That last phrase is the one I’d tape to the monitor. Not because it’s new, but because it describes about half the service pages I see.
No, not just for being AI written. Google cares whether the page helps somebody, not which keyboard it came from.
Here’s the catch, though. “Helps somebody” is exactly where most AI drafts fall on their ass. Ask a chatbot for a plumbing page and you’ll get 600 words about how water is important and your family deserves peace of mind. Nothing about your service area, your actual prices, what you won’t touch, or why the guy before you left a P trap held together with tape. It reads fine. It says nothing. Google’s raters have a name for that, and it isn’t a compliment.
I see this all the time: a site with a dozen city pages that are the same page with the town name swapped. Palm Coast, Ormond, Bunnell, same three paragraphs, same stock photo of a handshake. That’s not a strategy. That’s a mail merge with ambition.
Timing, mostly. Google started its September 2026 spam update on September 24, and its status dashboard still lists it as active. Google hasn’t said what specific tactic it’s going after, so I’m not going to pretend I know.
What I do know: when Google publishes clearer language about low value AI pages in the middle of a spam update, it’s not a coincidence I’d bet the business on ignoring. If your traffic dipped in the last week or so, don’t panic and don’t start deleting pages at 2 a.m. Wait for the rollout to finish, then look at which pages lost ground. If they’re the thin, copy paste ones, you have your answer.
Use it like an intern, not like a ghostwriter you never check on.
Let the robot do the first draft, the outline, the “help me say this shorter” stuff. Then a human who knows the business has to go through it and do three unglamorous things:
The guide also mentions telling readers how content was made where that makes sense, and if you sell products, it has specific rules about labeling AI generated images and product data. Most local service businesses won’t need a disclaimer on every page. But if you’re running an online store and your product photos came out of an image generator, read that section. Seriously.
I wrote about how AI search is changing the way people find local businesses in my post on Google AI Mode, and the theme is the same: the machines are getting better at spotting pages with nothing to say. If you want help making your site the opposite of that, my AI services page covers how I build content that holds up.
Google encourages sharing how content was made when readers would reasonably want to know, but it doesn’t require a badge on every service page. Ecommerce is stricter: AI generated product images and product data have their own labeling rules in the guide.
Not automatically, and not during a rollout. Let the spam update finish, see which pages actually dropped, then improve or merge the thin ones. Nuking half your site in a panic is how you lose the pages that were working.
Yes, if a person who knows the subject edits it, checks the facts and adds real experience. The tool isn’t the problem. Publishing whatever it spits out without reading it is the problem.
Nothing about this update is scary if your website sounds like you. It’s only a problem if your website sounds like everybody, which, to be fair, is what a chatbot does best. Write less, say more, and check the damn hours before you hit publish.
If your site is mostly robot filler and you’d rather have pages that actually sell, tell me about your project and I’ll tell you honestly what’s worth keeping.
