Website optimizer: Google's product was sunset in September 2023, so here are the three kinds of tool the name now means

Somebody searching for a website optimizer is usually looking for one of two things: the Google product they remember, or a tool that will tell them which version of a page performs better. The product is gone. Google Website Optimizer, the original, was retired in 2012 and its job moved into Google Analytics as Content Experiments. Its successor, Google Optimize, was closed in September 2023, and Google's own help page still carries the reason: Optimize "did not have many of the features and services that our customers request and need for experimentation testing", so the company decided to invest elsewhere. That leaves the category, which contains three different kinds of product, and one question worth answering before you buy any of them.

The question to answer first

A split test needs enough traffic and enough conversions to tell a real difference from noise. The rough test: count the conversions, not the visits, on the page you want to improve. A page producing a handful of conversions a month cannot resolve a small difference between two versions in any reasonable period, and running the test anyway produces a confident answer that is wrong.

At that level the productive work is different in kind: fix the obvious leaks, make one deliberate change at a time with a clean before and after, and spend the effort on finding out why people leave rather than on measuring two variants. How to size a test honestly and what to do when the numbers are too small for one is set out in conversion rate optimisation, and it is the page to read before this one if your traffic is modest.

The three kinds of tool sold as a website optimizer

Kind	What it does	Suits	What it costs you beyond the fee
A/B and multivariate testing platform	Serves two or more versions of a page, splits traffic, reports which won	Sites with enough conversions to resolve a difference	A script on the page, and the flicker risk that comes with client side changes
Analytics with experiment features	Reports behaviour and, in some products, runs simple experiments or audience comparisons	Teams that mainly need to see where people drop out	Setup discipline. The report is only as good as the events you defined
Server side testing and feature flags	Decides which version to build on the server or behind a flag, often inside the product rather than the marketing site	Product teams with developers	Engineering time, and it belongs in the release process rather than in marketing

Two things get sold under the same words and are neither of these. A page speed tool reports how fast the page loads, which is worth doing and is not testing. An on-page SEO scorer grades a page against a query, which is a different job again, and the categories of that tool are covered in what a free audit really tells you.

What happened to the Google products, with dates

• Google Website Optimizer was the free testing product of the late 2000s. It was retired in 2012 and the capability reappeared inside Google Analytics as Content Experiments, which was simpler and did less.
• Google Optimize and Optimize 360 replaced that, with a visual editor and integration into Analytics, and became the default free testing tool for small and medium sites.
• September 2023. Optimize was sunset. Google's help page for it is still published and still headed with the sunset, and the stated reason is that the product lacked features customers needed for experimentation. Google pointed users towards third party testing tools and towards experiment features in its analytics product.
• The practical consequence is that there is no free, full featured, Google made A/B testing tool any more. Any article recommending Optimize is out of date, and any agency proposing it should be asked when they last ran one.

Choosing one, in the order that matters

• Count the conversions on the page first. This decides whether you are buying a testing tool or a research tool, and it is the step people skip.
• Decide where the change happens. Client side tools are quick to install and can flicker on slow connections. Server side tools are clean and need a developer. Pick by who will build the variants.
• Check the conversion definition. The tool has to count the same conversion your business counts, with the same window, or the result will not survive a conversation with finance.
• Insist on the raw data coming out. Per visitor assignment and outcome as an export, so a result can be rechecked after the subscription ends.
• Ask what it does to page performance. A testing script that delays the first render can cost you more than the winning variant gains.
• Price the plan you need, which is usually set by monthly visitors or tested sessions rather than by features, and price it at the traffic you expect in a year.

Running a test so the answer holds

• Write the hypothesis before the variant. One sentence: we believe changing X will improve Y because Z. If Z is missing, the test is decoration.
• Change one thing. Two changes in one variant produce a result you cannot act on.
• Decide the duration and the sample in advance, and do not stop the test the first morning it looks good. Stopping when you like the number is the most common way these projects produce false wins.
• Run for whole weeks. Behaviour differs by day, and a test that starts on a Tuesday and ends on a Friday has measured a slice of the week.
• Record the losers. A test that failed is information about your customers, and a log of them stops the same idea being retried every year by somebody new.
• Check the result against your own order data, not only the tool's dashboard, before rolling the change out.

What to do if you are here because Optimize stopped working

Three steps, in order. Export anything you still need from the old reports, because an archived experiment you cannot read is a decision you will make twice. Remove the old container snippet from the site, since a script that does nothing is still a request. Then decide, using the conversion count above, whether you are replacing it with a testing tool at all, because a good share of sites that used it were never running tests that could resolve a difference and are better served by fixing the known problems.

Questions about website optimizers

Is Google Website Optimizer still available?

No. The original was retired in 2012, and its successor Google Optimize was sunset in September 2023. Google's help page for Optimize is still online and is headed with the sunset, which is the clearest confirmation available.

What replaced Google Optimize?

No single free product. Google pointed people towards third party testing platforms and towards experiment and comparison features in its analytics product. In practice teams either buy a dedicated testing tool or move testing into the product with feature flags.

Do I need an A/B testing tool at all?

Only if the page you want to improve produces enough conversions for a difference to be visible. Below that, sequential changes with a clean before and after, plus talking to customers, produce better decisions than an underpowered test.

Is a website optimizer the same as an SEO tool?

No. One compares versions of a page against a conversion. The other grades a page against a search query. They are bought by the same person and they answer different questions.

Will a testing script slow my site down?

It adds a request and, with client side tools, can briefly show the original before the variant. Ask how the tool handles that, test it on a slow connection, and weigh the delay against the gain you expect.

How long should a test run?

Until the sample you decided on in advance is reached, in whole weeks. Decide the stopping rule before you start, write it down, and keep to it even when the early numbers are flattering.

The version of your page nobody is testing

Every tool above compares two versions of a page in front of the same visitor. There is now a version of your business nobody is testing at all: the two or three sentences an assistant produces when a buyer asks who to use in your category, assembled from whatever has been published about you rather than from the page you optimised. It is the first thing a growing share of buyers read, and it changes without anybody editing it. Worth reading the current wording before the next test cycle. AI Knows Us asks those buying questions across the main assistants and records which companies get named and what the answers were built from.