Sometime in the last year or two, somebody had AI write part of your website. Maybe it was the service pages. Maybe it was a few hundred product descriptions, or the little summaries that show up under your page titles in Google. It was fast, it read well enough, and nobody went through every line.
Somewhere in there is a sentence that isn't true. A "same-day service" you have never offered. A price from two specials ago. "Licensed in every state," on a business licensed in one. It has probably been sitting there for months without doing much harm, because almost nobody reads a service page all the way to the bottom.
Two things changed this fall that make that sentence matter. On October 1, Google wrote down, plainly, that a person is expected to check every AI-written word before it goes live. And over the same stretch, Google moved more advertisers onto settings that let it write ad copy straight from your website. The sentence nobody read on your page can now be the first thing a stranger reads in your ad.
Google Did Not Ban AI Content on October 1. It Put the Review Rule in Writing.
Start with what actually happened, because the headlines around it have been louder than the change.
Google keeps a short page in its Search Central documentation about using generative AI to create website content. On October 1 it added a few sentences. The first explains why AI gets things wrong: generative models "don't retrieve facts, but predict a likely sequence of words based on their training data," so the output can contain inaccuracies. The second is the one that matters: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing."
That is not a ban. The same page still says AI can be genuinely useful for research and structuring content, and Google's own FAQ on the subject says outright that it does not ban AI-generated content. Google also did not announce a new penalty. It described the edit as bringing its documentation in line with what it already tells developers at its events.
So why care? Because "nobody told us" used to be a reasonable position, and it isn't anymore. The expectation is in writing. If a page full of unchecked AI claims stops performing, or a customer holds you to a promise your website made, the defense that nobody said you had to read it is gone.
The older rule still sits right next to it, too. Churning out lots of AI pages that add nothing for the reader can fall under Google's spam policy on scaled content abuse. That has been true for a while. What October 1 added is that even a single page is expected to have a human behind it.
Your Title Tags and Image Descriptions Show Up in Search. Nobody Proofreads Them.
Here is the part of the update most owners will skim past, and it is the part that applies to almost everyone.
Google's new wording says the review also covers metadata: the page title, the meta description, structured data, and the alternate text on images. In plain terms:
- The page title is the blue headline people click in search results.
- The meta description is the short summary that often appears under it.
- Structured data is the behind-the-scenes code that tells search engines your hours, prices, ratings and address.
- Alt text is the written description attached to each image, which screen readers read aloud and search engines read too.
These are exactly the fields that website tools and AI plugins now fill in automatically, in bulk, across every page at once. They are also the fields an owner almost never looks at, because they don't appear on the page itself. Your web person sets up a plugin, it writes three hundred meta descriptions in an afternoon, and the first human to read most of them is a stranger seeing your business in a search result.
If you sell products, there is one more layer. Google's Merchant Center rules require product titles and descriptions generated by AI to be labeled as AI-generated, and AI-generated product images to carry a label in their file data. If an app wrote your product feed, that is worth a direct question to whoever set it up.
Text Customization Turns Your Web Pages Into Ad Headlines You Never Approved
This is where a search housekeeping note turns into a business decision.
Google Ads has a feature called text customization. It writes extra headlines and descriptions for your ads, and Google says it builds them from your domain, your landing pages and your existing ads. Its help page is specific: it takes snippets from landing page titles, descriptions and meta tags, and uses generative AI grounded in your landing page content. Read that list again next to Google's new review rule. They name the same fields.
Text customization now lives inside AI Max, and three things put it in more accounts than ever this fall:
- New Search campaigns start with AI Max switched on. You have to turn it off if you don't want it.
- Campaigns that used the older "automatically created assets" setting were moved into AI Max during September, with text customization carried over. We walked through that change in Google Switched Your Search Campaign to AI Max on September 1. You Never Said Yes.
- Performance Max turns text customization on by default, using content from your landing page and domain.
Google says it checks these generated assets for accuracy and refreshes them at least every 48 hours. But its own help page puts the job of a truthful source back on you: "Ensure your website content is accurate, not misleading, and not in violation of Google Ads policies or applicable laws." Google's checks can confirm that an ad matches your website. They cannot confirm that your website is right.
None of this means AI Max is bad. It means your website has quietly become part of your ad account. If you have not decided whether you want these settings on at all, Google Wants You to Turn On AI Max. Most Small Businesses Shouldn't Yet. is the place to start.
A Wrong Price in a Google Ad Costs More Than One Buried on Your Website
A false sentence on a page and the same sentence in an ad are not the same problem.
On the page, the wrong line is usually buried. It sits in paragraph six of a service page that most visitors leave after a few seconds. It is wrong, and it rarely does much.
In an ad, it is the headline. It is the first promise a stranger sees, it is the reason they click, and you pay for that click. Then they call. A plumbing company's AI-written page that still says "$49 drain cleaning" from a spring promotion becomes a phone call from someone expecting a $49 drain cleaning. Your office now has two bad choices: honor a price you stopped offering, or start the relationship by telling a new customer the ad was wrong.
Some claims carry more than awkwardness. A credential the business does not hold, a "guaranteed" result, a license, a "certified" anything, a health or safety claim: those are the words that get ads disapproved and, in some industries, get a business a letter from a regulator. Advertising law in the United States has always expected claims to be truthful, whoever or whatever typed them. "The software wrote it" has never been an argument that works.
And there is a quieter cost. An ad that promises one thing and lands on a page that says another trains people not to trust you. They may not complain. They just don't call back.
AI Can Write Your Website. It Cannot Vouch for a Single Line of It.
I use AI every day in my own business, and I would not tell an owner to stop. It is very good at a first draft, at structure, at getting you past a blank page. Banning it from your website would be a strange overreaction to a problem that has nothing to do with the tool.
The problem is volume without reading. AI made it possible to publish three hundred pages in the time it used to take to write three, and almost every business that did it skipped the step where someone reads them. That is the failure Google is pointing at, and it is the same failure that shows up everywhere AI gets used carelessly: the work gets faster, and the judgment quietly drops out of the process.
We have made a version of this argument before about where automation belongs, in Everyone Is Automating the Part Your Customers Want Human. And if you want a better job for AI in your marketing than writing unread copy, Stop Asking AI to Write Your Ads. Ask It to Read Your Account. makes the case for pointing it at your data instead.
The rule I would use is simple. AI can write anything. A person reads everything that makes a promise.
Every Price and Guarantee on Your Website Is a Promise Somebody Has to Keep
You don't need to reread your whole website line by line tomorrow. You need to find the sentences that cost money when they are wrong and put a human name next to each one.
Find what AI touched. Ask whoever built or maintains the site which pages, product descriptions, titles and meta descriptions were generated. If nobody knows, assume all of them. Most website platforms can export page titles and descriptions to a spreadsheet, which turns an invisible problem into a list you can read in an hour.
Read for the claims that carry risk. Skim past the filler and stop on anything that is a promise or a fact:
- Prices, specials and anything with a dollar sign
- Guarantees, warranties and "risk-free" language
- Licenses, certifications, credentials and awards
- Service area, hours and response times ("same-day," "24/7," "within the hour")
- Numbers of any kind: years in business, customers served, ratings
- Anything medical, legal, financial or safety related
Give each one an owner. The person who signs off on "same-day service" should be the person who knows whether the crew can actually do it. That is usually not your web designer and never the software.
Then fence the ads. If you run AI Max or Performance Max, use the text guidelines in each campaign: up to 25 words or phrases Google may never use, and up to 40 restrictions on the kinds of claims to avoid. Check the asset report for headlines Google wrote, which are marked as added by Google AI. And when an offer changes on your site, remember that the ads can follow it within a couple of days, for better or worse.
Google Reads Your Website to Write Your Ads. Somebody at Your Business Has to Read It First.
For years, the website and the ads were separate jobs. One person built the site, another ran the campaigns, and the owner looked at whichever one was making noise that month.
That separation is gone. Google now reads your pages to decide what to show in search, what to write in your ads, and where to send the click. The October 1 update did not create that situation. It just said out loud what it means: if you publish it, somebody at your business is expected to have read it.
That is not a reason to be afraid of AI. It is a reason to spend one afternoon reading the parts of your own website that are quietly making promises on your behalf, because Google is already reading them, and so are your customers.
