AI tools can now build a good-looking website in an afternoon. That's a real win for a small business that couldn't afford a developer.
The catch is that a site can look finished to you and still be a dead end for Google.
TL;DR: AI site builders don't know your business or your SEO rules, so a vibe-coded site can look done and still be hard for search engines to find. Search Engine Land lays out seven checks to run before launch and after every big change.
🧱 A Site That Looks Finished Isn't Always Findable
Vibe coding means describing a website to an AI tool and letting it write the code. It lowers the barrier a lot. A small business can launch without hiring a full-stack developer.
The problem is what the AI doesn't know. It doesn't know your business, your marketing goals, or why a specific SEO element exists. It builds what you ask for, and sometimes what it assumes.
A human visitor sees a working page. A search crawler may see something very different, and that gap is where the trouble starts.
Static sites have a close cousin of this problem: the SEO work moves onto the site owner instead of going away.
🧠 The AI Doesn't Remember Why Your SEO Rules Exist
Here's the example from the article. Say a promo page at /promo/teeth-whitening has a canonical tag pointing to /services/teeth-whitening, so ranking signals pile up on one page. You ask the AI to update the promo page.
The AI may change the canonical to point to itself, because that's what canonicals usually do. That undoes the whole point of the tag.
It gets worse over time. Switch AI models or lose the original chat, and the AI has no record of why a rule was set.
Schema has the same gap. The AI writes valid markup, but it may guess your hours, prices, or address, or make details up. Whether a solo practitioner is a Person or an Organization, and which business type fits, are your calls.
Platforms change too. The article notes that Lovable moved new projects to server rendering while older projects stayed on dynamic rendering. Moving an old site over is effectively a full site migration.
Hit reply
When you ask an AI tool to edit one page, do you crawl the whole site afterward to see what else changed? Hit reply and tell me what you check, or what the AI quietly changed on you.
🚧 Crawlers Can Get Locked Out Without Anyone Noticing
Two problems hide from you here. First, content that only loads after JavaScript runs may never show up for crawlers that don't run it, which the article says includes many AI search tools. Second, the services behind your site can fail or block crawlers.
A free database plan might pause when the project sits idle. Bot protection can mistake Googlebot for a bad bot.
The consequences are specific. Google slows its crawling when it hits frequent 5xx server errors, and it can eventually drop those URLs. Pages that return a 403 get removed from the index.
Security is the last gap. AI-written code can work fine and still have holes, and forms, input fields, and upload handlers are common entry points. A hacked site can end up with spam pages in Google's index and a warning in search results.
🛠️ Seven Checks Fit Into Three Habits
Test What Crawlers Actually See
Right-click your page and choose View Page Source. Don't use Inspect, because that shows the page after JavaScript has run. If your key content isn't in the raw source, fix that first.
Then list every service your site depends on, including hosting, database, and bot protection. Learn their plan limits and what happens when the site goes idle. Confirm your bot protection allows Googlebot, and check the Page Indexing report in Search Console for 5xx and 403 errors.
Write Your SEO Rules Down for the AI
Put your canonicals, redirects, and noindex tags in the AI tool's project instructions, along with the reason for each one. Tell it not to change them unless you say so. After any AI edit, crawl the site and confirm those elements still work.
Give the AI your real business details for schema, then check its output against them. Follow your platform's announcements, and treat any rendering change like a migration: test everything before you republish.
Hand Routine Updates to Tools Built for Them
If you'll publish posts, products, or news often, use an established CMS like WordPress or Shopify for that part of the site. With a vibe-coded CMS, every content update is a code change, and each prompt can break something. A standard CMS handles sitemaps, redirects, private drafts, and meta tags without touching code.
For forms, use an established form service with built-in validation and spam protection. If your code lives on GitHub, turn on Dependabot alerts. Check the Security Issues report in Search Console on a regular schedule.
One thing I'm thinking about is how confident these tools sound. An AI will tell you the site is ready, and it will sound sure about it.
Building the site got easier, but being responsible for how it performs didn't change. A pass through View Page Source and Search Console catches a lot of what the AI can't see. If you're in doubt, get a second opinion from a technical SEO before you launch.
Till next time,
