How do you prove SEO work actually worked?
Short answer
By leaving some pages alone on purpose. Split similar pages into two groups, improve one and never touch the other, then compare them over a set period. Both groups go through the same Google updates and the same quiet months, so the gap that opens between them is the part your work caused rather than the weather.
Why “before and after” proves nothing
Visitor numbers move for all sorts of reasons that have nothing to do with anybody’s work. Google changes how it ranks things. A competitor’s website goes down for a week. Your busy season arrives. The way results are displayed changes.
So a graph that goes up after someone did some work is equally consistent with three very different stories: the work helped, the work did nothing, or the work actively hurt and was hidden by a rising tide. You genuinely cannot tell.
Most of this industry runs on that ambiguity, which is why so much SEO reporting can never be proved wrong.
Leaving pages alone on purpose
Pick pages that are genuinely comparable — similar type, similar purpose, similar traffic. Split them at random into two groups. Only improve one group. At the end, report the difference.
This costs something real: the untouched pages do not get the improvements, so you are knowingly leaving some value on the table. That is the price of a number you can actually trust, and it is worth paying, because otherwise you can never tell a service that works from one that is simply expensive.
Decide the timescale before you look
Google’s figures arrive about three days late, and a change takes weeks to settle into the results. Anything measured over less than that is measuring random movement.
Deciding the timescale in advance — and not touching the page while the clock runs — is what stops the result being chosen afterwards to make somebody look good.
Last reviewed 6 September 2026