The Phantom Subdomain: When AI Advice Costs You 10 Years of SEO
How one missing step in a ChatGPT-guided migration tanked a site’s traffic — and the 30-second fix that resolved it
The Problem
A client came to me after six weeks of failed attempts to get his website indexed by Google.
He ran a niche history blog — serious primary source research with over 130 posts built up over nearly a decade. The kind of content Google should love: original, authoritative, deeply researched.
But his Search Console told a different story: 48 pages “not indexed.” Redirect errors. 404s. Crawl failures across the board.
Traffic had collapsed from thousands of monthly visitors to 61.
He’d tried everything he could think of. Resubmitted his sitemap. Clicked every button in Search Console. Asked the same AI that got him into this mess how to get out. Nothing worked.
Six weeks of this. For a site that had been performing fine for ten years.
The Backstory
Here’s what happened.
The client had originally built his blog on a subdomain — let’s call it old.example.com. It worked. It ranked. It brought in steady traffic for a decade.
Then he asked ChatGPT for advice on improving his SEO. The AI told him two things:
-
You need HTTPS. Correct — Google has used HTTPS as a ranking signal since 2014.
-
You should consolidate to your main domain. Also correct — subdomains are treated as separate sites by Google, so moving to
www.example.comwould consolidate his authority.
Good advice. Technically accurate. The kind of thing any SEO consultant would recommend.
The problem: ChatGPT told him what to do without telling him how to do it completely.
What a Proper Migration Looks Like
Moving a site from one domain (or subdomain) to another isn’t complicated, but it has steps that can’t be skipped:
1. Set up the new domain with SSL
2. Move all content to the new location
3. Configure redirects from every old URL to its new equivalent
4. Update canonical tags
5. Submit a change of address in Google Search Console
6. Delete or redirect the old DNS records so Google stops trying to crawl a dead site
7. Monitor Search Console for errors during the transition
The client did steps 1-3. The AI never mentioned steps 4-7.
So he ended up with his content in the right place — but the old subdomain still existed in DNS, pointing to… nothing. No valid SSL certificate. No content. Just a broken endpoint that Google kept trying to crawl because it remembered URLs from the past decade.
The Diagnosis
When I started investigating, Search Console showed 48 pages not being indexed:
- 9 redirect errors
- 20 not found (404)
- 19 crawled but not indexed
The client assumed these were problems with his new site. He’d spent six weeks trying to fix www.example.com.
But when I actually looked at the URLs Google was complaining about, none of them were on his main site.
100% of the 48 errors came from the orphaned subdomain.
His main site had zero indexing issues. It was working fine. Google just couldn’t tell him that because the Search Console data was polluted with errors from a subdomain he thought he’d left behind.
The Fix
I logged into his DNS provider and deleted one record: the CNAME pointing old.example.com to Blogger’s servers.
That’s it. One record.
After updating, the subdomain stopped resolving immediately. Google will hit “domain not found” instead of SSL errors and 404s. The 48 errors will clear from Search Console over the next 1-2 weeks as Google discovers there’s nothing there anymore.
His actual site — the one with 130+ posts of original historical research — will index normally on Google’s standard timeline for migrated content. Probably 2-4 weeks to see meaningful progress, 4-8 weeks for full indexing.
The Lesson
The AI wasn’t wrong. HTTPS matters. Domain consolidation builds authority. These are real SEO principles.
But “technically correct advice” isn’t the same as “complete implementation guidance.”
ChatGPT gave him the strategy without the execution plan. It didn’t mention DNS cleanup. It didn’t warn him about Search Console pollution. It didn’t tell him what to monitor or how to verify the migration was complete. And when things broke, it had no way to diagnose what went wrong — because it couldn’t see his actual configuration.
The client isn’t stupid. He followed reasonable-sounding advice from a source that seemed authoritative. He had no way of knowing the advice was incomplete until the consequences showed up six weeks later.
The Cost
-
10 years of accumulated domain authority — disrupted (recoverable, but it takes time)
-
Traffic — from thousands to 61 monthly visitors
-
Time — 6 weeks of frustration before finding the real problem
-
Trust — in tools that give confident answers without complete context
The actual fix wasn’t anything cryptic or complicated. Finding it just took an experienced (human) eye.
The Takeaway
AI can give you a checklist. It can’t verify you completed it. It can’t see your DNS records, your Search Console errors, or the gap between what you intended and what actually happened.
If you’re making infrastructure changes to your website — migrations, SSL, DNS, redirects — and something breaks, the problem usually isn’t that you did the wrong thing. It’s that you did the right thing incompletely, and nobody told you what “complete” looks like.
That’s where human expertise still matters.
Related Reading
-
What Is GEO? The Next Evolution of SEO in the Age of AI
AI-generated answers are reshaping how customers find businesses online. Is your site GEO optimized? -
Google Visibility That Lasts: Security Is SEO
You could lose search rankings overnight — not from bad content, but from hidden security issues. -
The Value of Page One: Why Your Search Position Matters
One shift in your Google ranking could cut your traffic in half. The math might shock you.