"Just A/B test your title changes."

"Just A/B test your title changes."

With what control group, exactly?

You can't split-test SEO the way you split-test a landing page — there's one Google result, not a 50/50 traffic split. Here's how to actually run an on-page test that isn't self-delusion:

— Test in batches, not one page. Change titles on 20 similar pages, keep 20 similar pages untouched as a control.
— Pick pages with stable, comparable traffic. Testing on a page that's already trending up will lie to you.
— Annotate the exact change date. You'll forget. Past-you always forgets.
— Measure CTR at fixed position, not raw clicks. Position drifts; raw clicks confound everything.
— Run it 4 weeks minimum. SEO tests have a long, noisy tail.

The single-page "I changed it and traffic went up" story is not a test. It's an anecdote wearing a lab coat. (Correlation, seasonality, a competitor dropping out — any of these explains your "win" better than your title.)