A lot of SEO advice gets repeated until it sounds like fact. These are the bits I wanted to check myself. Each test runs on sites I control, the method goes up before the result exists, and the prediction is stated in advance so the whole thing can be wrong in public.
How these work
Every test publishes in two stages. The first post goes live when the test starts and contains the question, why it matters, the exact method and what I expect to happen. The second stage adds the result to the same page.
That order is the point. Publishing the prediction first means I can’t quietly rewrite what I expected once the data arrives, and it means you can judge the method before you know whether it produced a convenient answer.
The rules
The method has to be repeatable. If you can’t run it yourself from what’s written, it isn’t finished.
No result gets published before it exists. A results section marked as pending is an honest page. An invented number is not.
The prediction is stated in advance, so the test can fail.
Test properties are mine. These run on my own sites, not on client sites, unless a client has specifically agreed and there’s no risk to them.
Sample sizes are small and get said so. One site is an anecdote. Several are a signal. Neither is proof, and the posts say which they are.
A null result gets published too. The tests that show nothing are usually more useful than the ones that confirm what everyone already assumed.
Running now
| Test | Question | Status |
|---|---|---|
| Noindex and link value | Does a noindexed page still pass value through its links? | Method published, running |
| Answer passages and AI citation | Does a direct answer at the top of a page change how often it gets cited? | Method published, running |
| FAQ schema after the rich result | Does FAQPage markup still do anything now Google stopped showing it? | Method published, running |
| Indexing speed | How long does a new page take to get indexed, and does domain age change it? | Method published, running |
Why I bother
Most of what I recommend to clients rests on something Google has published, something I’ve seen in my own data, or something the industry repeats. The first is checkable and the third often isn’t.
These tests are how I move things from the third category into one of the first two, and where a test shows my own advice was wrong, that gets published as well.
Want the results?
New tests and their results go out in the weekly roundup. If you’d rather talk about your own site than read about mine, get in touch.