Someone recently blamed Google’s “AI” for telling searchers that their site had been offline since early 2026. The headline of their blog post leaned hard into tech jargon — something about “cross-page AI aggregation” and “liability vectors.” It sounded serious. It also didn’t really mean anything.
Instead of arguing on Reddit, they linked to their post. That gave Google’s John Mueller a chance to look at the actual site. And within minutes, the mystery wasn’t mysterious anymore.
It wasn’t AI. It was JavaScript.
Where the confusion started
The site was showing placeholder text in its base HTML — something like “not available.” Then JavaScript would load and replace that text with the real content.
For users with modern browsers and scripts enabled, everything looked fine. For anything that didn’t run the JavaScript immediately (including search crawlers in certain situations), the placeholder text was what got seen and indexed.
So Google read “not available” and took it at face value.
The site owner wasn’t sure how Google could even detect downtime. They speculated about AI systems pulling in fresh information, maybe through some kind of cross-page process. They wondered if Google was blending unrelated content into answers.
I get the instinct. AI search feels opaque. It’s easy to imagine a giant model hallucinating something weird about your site.
But in this case, nothing exotic was happening.
What’s actually going on with Google’s AI answers
Google’s AI search doesn’t magically know things. It still relies on traditional search systems to fetch pages. The AI layer then summarizes what it finds.
Think of it like this: someone Googles your site, reads what’s on the page, and explains it in plain English. If the page says “not available,” that’s what they’ll repeat.
There’s no need to invent new phrases about “liability vectors.” The crawler saw the HTML before the script rewrote it. End of story.
Mueller’s fix (and it’s boring, which is good)
John Mueller responded calmly and pointed out the issue: don’t use JavaScript to swap critical text from “not available” to “available.” If a client doesn’t execute the script, it sees the wrong message.
He even compared it to another long-standing warning: don’t use JavaScript to flip a robots meta tag from noindex to something else after load. Crawlers may never see the updated version.
The safer approach is simple. Put the correct content directly in the base HTML. If you need JavaScript, use it to enhance the page — not to fix or replace essential information.
It’s not flashy advice. But it works.
more: Best ai website builder
The real lesson here
What stands out to me isn’t the JavaScript mistake. Plenty of developers have made similar ones. It’s the leap to blaming “AI” before confirming the basics.
The site owner admitted they didn’t fully understand how Google’s AI summaries are assembled. That’s fair — most people don’t. But then they started guessing. They removed a pop-up as a “shot in the dark” fix. They speculated about scraping and cross-page aggregation.
All while the root problem was sitting in the source code.
I’ve seen this pattern a lot in SEO. Google feels mysterious, so we assume the issue must be mysterious too. Sometimes it is. Often it’s just rendering, indexing, or a mismatch between what users see and what bots get.
Before blaming AI, check the HTML.
If the raw page says your site is unavailable, Google will believe it. And honestly, so would I.