AI Visibility Pitfall: Cache Showing An Old Version Of Your Site

I found this out the hard way, on my own site.

While testing one of my tools, I asked an AI tool to tell me what my company did. It struggled to find the right reference, and then said it didn’t like my headline. As it read it back to me, I laughed (and then freaked out) because it was an old headline from before a major site overhaul. I frantically checked my live site to see if it somehow was pointing to an old version, but nope – the new one was live. I was looking at the current site, live, on my phone, with the actual content on it.

So which one of us was wrong?

Neither. And that’s the part that should worry you.

As a web optimization strategist who helps businesses keep their sites structured in a way that can be read, I was horrified.

But here’s the lesson:

The web isn’t one web. It’s a thousand cached copies of the web.

Kinetic Pencil

What was actually happening?

My phone was pulling the site from a Cloudflare edge node near me. The AI tool was pulling it from a server somewhere else entirely — and that server was still serving a snapshot from before I relaunched. Same URL. Two completely different sites. The one being synthesized to strangers was not only wrong, it was over multiple months old.

How did this happen? I had changed hosting providers prior to relaunch, which was a factor in my case. I also had a content delivery network (CDN) created by a group of connected servers spread throughout the world. Their role is to store and deliver content geographically closer to the user requesting it. This actually means your site isn’t fetched freshly every time someone visits. Instead, it’s actually serving pre-baked copies from wherever’s closest to the request — and if you don’t explicitly tell it to throw the old copy of the website away, it won’t. It’ll happily keep serving 2019 to some crawler in another timezone while you’re staring at 2026 on your own screen, completely convinced everything’s fine.

And here’s the uncomfortable part: Googlebot, GPTBot, ClaudeBot, or PerplexityBot don’t necessarily hit your live site. They hit whatever edge node their infrastructure happens to land on. If that node is holding a stale cache, your new positioning, your new services, your new content — none of it exists yet, as far as the thing that’s supposed to be finding you is concerned.

You can publish the best page of your life and still be functionally invisible.

Kinetic Pencil

01Checklist

What to actually check

  • Purge your CDN cache globally, not just for your region.

    If you’re on Cloudflare: Caching → Configuration → Purge Everything. A regional purge fixes what you see and leaves everyone else looking at the old version.

  • Purge any caching plugin separately.

    WP Rocket, LiteSpeed Cache, WP Super Cache — these sit on top of your CDN, not inside it. Clearing one doesn’t clear the other.

  • Use Google Search Console’s “Test Live URL” tool.

    This is the only way to see, with certainty, what Googlebot itself is currently pulling — not what you assume it’s pulling.

  • Check incognito, from a different network than your usual one.

    If your regular browser and your incognito window show two different sites, you’ve confirmed it’s a caching issue, not a “my eyes are lying to me” issue.

02Checklist

Once your site is actually visible — get it seen faster

Fixing the cache gets you back to zero. These get you ahead:

  • IndexNow.

    A free plugin that pings Bing and its partner engines (Yandex included) the moment you publish or update a page, instead of waiting for them to wander back on their own schedule. It does not cover Google — Google runs its own separate system.

  • Google Search Console’s URL Inspection tool.

    After you publish, manually request indexing for that specific URL. It’s not instant, but it’s faster than waiting for a crawl you didn’t ask for.

  • An updated XML sitemap, submitted in Search Console, that actually reflects your current pages

    — not the sitemap from whatever theme setup you started with.

  • robots.txt, checked by eye, not assumed.

    Visit yourdomain.com/robots.txt yourself. Confirm nothing is quietly blocking the crawlers you’re trying to attract.

Kinetic PencilCache & crawl checklist

None of this is difficult. It’s just invisible — which, it turns out, is exactly the problem it’s solving.