SEO Experiments: Test Every Change Before You Keep It
Most SEO changes are never measured. Every change SERPclimber makes ships as a live experiment against the page's own baseline, runs for a fixed window, and ends in a decision: confirm the winner, or revert to what was there before.
replaces
- Guesswork
- Untracked title edits
- Post-hoc chart reading
- Manual CMS rollbacks
- 14 daysdefault window# Long enough for Search Console to settle, short enough to iterate.
- 1-clickrollback# A losing variant reverts to the original page.
- Baselineas the control# The page's own prior performance, not a guess.
What SEO Experiments does
Why most SEO is unfalsifiable
The usual loop is: change a title, wait, look at a chart, and argue about seasonality. There is no control, so any explanation fits. A page that improves gets credited to the change; a page that drops gets blamed on an update. Neither conclusion is earned.
True split testing — half your users see A, half see B — is not available in organic search. Google crawls one version of a URL. So the control has to come from time rather than from audience.
The page's own baseline is the control
Before a variant goes live, SERPclimber records how the page was already performing: position, clicks, impressions and click-through rate over a comparable prior window. That baseline is the control. The variant then runs live for a fixed window and is measured against it.
At the end of the window you get a decision, not a dashboard: the variant won, the variant lost, or the result is inconclusive. Winners are confirmed and the baseline is retired. Losers revert automatically, so the downside of a failed test is a fortnight, not a permanent regression.
How it works
- 1
A candidate page is chosen
From the SEO dashboard — usually a page whose impressions are holding while its clicks fall.
- 2
The baseline is recorded
Position, clicks, impressions and CTR over a comparable prior window, captured before anything changes.
- 3
A variant is drafted
Written from that page's own search data by the content generator, with a plain-language summary of what changed and why.
- 4
You approve the publish
Publishing changes a live page, so it waits for you unless you have explicitly handed that step to autopilot.
- 5
The experiment runs its window
The variant is live and measured daily against the baseline. Progress through the window is visible the whole time.
- 6
You confirm or revert
A winner is kept and the baseline retired. A loser is reverted to the original page. Then the page cools down before it is eligible again.
Losing is cheap, which is what makes testing possible
The reason teams do not test SEO is that a bad change is expensive to discover and awkward to undo. Automatic rollback removes both problems: the original page is kept, and reverting is one action rather than an archaeology exercise in your CMS.
- The original title, meta and body are retained for the life of the experiment
- A losing variant reverts to exactly what was there before
- A cooldown period stops the same page being churned repeatedly
- Every experiment leaves a record of what was tried and what happened
Every experiment says what it changed, in words
A diff tells you which characters moved. A test is only useful if it also tells you what hypothesis it was testing. Each variant carries a short "what changed" summary — the intent behind the rewrite — so a confirmed winner teaches you something transferable.
- Plain-language summary of the change and the reasoning
- Before and after, side by side
- The query the rewrite was aimed at
- Result recorded against the hypothesis, not just the metric
Everything it covers
Baseline-controlled testing
The page's own prior performance is the control.
Fixed measurement windows
A 14-day default, so results are not read too early.
Automatic rollback
Losing variants revert to the original page.
Title and meta description testing
The highest-leverage change on most ranking pages.
Body content variants
Not just metadata — the page copy itself can be tested.
Daily progress tracking
Position and click deltas visible for the whole window.
Explicit decision step
Confirm winner or revert, rather than a chart to interpret.
Cooldown periods
A tested page rests before it is eligible again.
Experiment history
What was tried on a page, and what happened.
Wix publishing
Variants publish to a connected Wix site directly.
Approval gating
Nothing touches a live page without a yes, unless you delegate that.
"What changed" summaries
The hypothesis behind each variant, in plain language.
Is this the same as SEO A/B testing?
It is the same intent, with the method that organic search actually allows. Classic A/B testing splits an audience; you cannot split Googlebot. Large publishers approximate it by splitting a set of similar pages into test and control groups, which needs hundreds of comparable templated pages to reach significance.
SERPclimber uses the other valid design: a before-and-after test on one page, with a fixed window and the page's own prior performance as the control. It works on sites with dozens of pages rather than thousands, which is where most sites actually are. The trade-off is honest — a single-page time-based test is more exposed to seasonality and algorithm updates than a large split test, which is exactly why the window is fixed in advance and why inconclusive is a permitted outcome.
Frequently asked questions
It is a controlled before-and-after test, which is the design available in organic search. You cannot serve two versions of a URL to Google, so the control is the page's own prior performance over a comparable window rather than a parallel audience.
The default window is 14 days. That is long enough for Search Console reporting to settle and for a position change to show up in clicks, and short enough to keep iterating.
It reverts to the original page. The baseline copy is retained for the life of the experiment specifically so that rolling back is a single action.
A variant is an ordinary content edit — the same thing you would do by hand, but recorded and reversible. There is no cloaking and nothing is served differently to crawlers. The measurable risk is that a change underperforms for the length of the window, which is the reason rollback is automatic.
Wix is supported today. If your site is elsewhere you can still run the discovery and drafting side and publish the variant yourself.
Enough that some are worth testing — in practice a few dozen ranking pages. Because each test uses its own page as the control, you do not need the large sets of near-identical pages that group-based split testing requires.
That is a permitted outcome and it is reported as one. A test that cannot distinguish the variant from the baseline has not earned a change, so the safe default is to revert and spend the next window on a page with more to gain.
Works with
~/serpclimber/pagesSEO Dashboard
Search Console tells you what ranks. GA4 tells you what earns. Neither tells you which page to fix next. SERPclimber fuses both into one page inventory — position, clicks, sessions, revenue and referrers per URL — and ranks it by opportunity.
Read more~/serpclimber/experiments/draftContent Generator
Generic AI writing tools start from a keyword and a tone slider. This one starts from the page you already have: the queries it wins, the queries it is losing, and what is ranking above it. Every draft is a variant of a real page, aimed at a real gap.
Read more~/serpclimber/autopilotSEO Automation
Full autonomy is a big ask for something that edits your live site. Autopilot breaks the loop into four steps — find, draft, publish, decide — and lets you hand over each one independently, from fully manual to running on its own.
Read morePoint it at your site.
A project takes one domain. The first pass tells you what it found — before anything changes on your site.