Back to blog
· Darwin Team

Google Optimize is dead. Now what? The best alternatives for 2026

Google Optimize is gone. It shut down on 30 September 2023, and if you’re reading this, you’re probably still looking for something to replace it.

You’re not alone. At its peak, Optimize was installed on over 2 million websites. It was free, it integrated with Google Analytics, and it was good enough for most teams. Then Google killed it with a blog post and a 6-month countdown.

The official recommendation? Migrate to “partners” like AB Tasty, Optimizely, or VWO. Google’s way of saying: go pay someone else.

Here’s the thing: those tools work. But they all share a fundamental design flaw that Google Optimize had too. More on that in a moment.

What you actually lost

Google Optimize gave you three things for free:

  1. A visual editor: point-and-click changes without touching code
  2. GA integration: experiments tied to your existing analytics
  3. Statistical rigour: proper significance calculations, not vibes

Finding a replacement that does all three without costing £200/month is harder than it sounds.

The usual suspects

Optimizely (from $79/month)

The enterprise pick. Powerful experimentation platform with a steep learning curve. The visual editor works but takes time to master. You’ll need a dedicated person managing it, which defeats the purpose for small teams.

Best for: Large teams with dedicated CRO roles. Not for: Anyone who just wants to test a headline.

VWO (from $199/month)

Feature-rich and well-documented. Visual editor is genuinely good. But the pricing puts it firmly in enterprise territory. You’re paying for heatmaps, session recordings, and surveys you might never use.

Best for: Mid-market companies ready to invest in a full CRO stack. Not for: Startups, side projects, or anyone watching their spend.

AB Tasty

Enterprise only, no public pricing. If you have to ask, it’s not for you. Moving on.

LaunchDarkly

Technically a feature flag tool, not a CRO tool. You can run A/B tests with it, but it’s like using a lorry to do the weekly shop. Overkill and misaligned.

Best for: Engineering teams managing feature rollouts. Not for: Marketing teams testing landing page copy.

PostHog, GrowthBook, Statsig

Open-source or developer-focused options. Solid if you have engineering support. Require self-hosting or significant setup. You’ll write code, manage infrastructure, and interpret your own statistics.

Best for: Dev teams who want full control and don’t mind the setup cost. Not for: Anyone who liked Optimize because it was simple.

The problem nobody talks about

Here’s the fundamental flaw: every tool on this list uses permanent runtime mutations.

When you run an A/B test with VWO, Optimizely, or any traditional tool, here’s what happens:

  1. Their SDK loads on your page
  2. It fetches the active experiment
  3. It patches the DOM: changing text, swapping images, hiding elements
  4. Your visitor sees the modified version

That’s fine for one experiment. But what happens after a year?

After 12 months of testing, you might have 50 concluded experiments. Each one that “won” is now a permanent runtime patch. The SDK loads, fetches all 50 mutations, and applies them to the DOM on every single page view.

This creates real problems:

  • Performance degrades. 50 DOM patches before first paint. Your Lighthouse score drops. Your visitors notice.
  • Mutations conflict. Experiment 37 changed a button’s text. Experiment 42 changed its container. Now one overrides the other in unpredictable ways.
  • Debugging becomes impossible. Is that bug in your code or in a mutation from 8 months ago? Good luck finding out.
  • If the SDK fails, everything reverts. Your “optimised” page depends on a third-party script loading successfully. If it doesn’t, visitors see the original un-optimised version.

Google Optimize had this same problem. They just killed the product before most people noticed.

Test temporarily, commit permanently

There’s a different approach. What if winning changes were baked into your actual source code, instead of living as runtime patches forever?

That’s what Darwin does.

Darwin’s SDK runs experiments the same way: split traffic, show variants, measure results. But when a variant wins, it doesn’t stay as a runtime mutation. The winning change gets committed to your source code: via a Git pull request (for developers) or a direct CMS update (for Webflow, WordPress, Shopify).

The experiment ends. The mutation is retired. Your page is genuinely, permanently better. No runtime debt.

After 50 experiments with Darwin, your page has had 50 real improvements. Zero runtime patches loading on each visit. Clean DOM, fast performance, and changes you can actually see in your codebase.

The AI difference

The other thing Google Optimize never solved: coming up with what to test.

Most teams run 2-3 experiments per year. Not because the tool was slow, but because someone had to:

  • Analyse the data and spot problems
  • Come up with a hypothesis
  • Design and build the variant
  • Set up the experiment
  • Wait for results
  • Decide what to test next

Darwin automates this loop. AI analyses your visitor behaviour, generates hypotheses, creates variants, and launches experiments. You can review everything before it goes live (Review mode) or let it run autonomously (Autopilot mode).

The result is continuous optimisation instead of occasional testing. And because each winner gets committed to code, there’s no accumulating mutation debt.

See it in action

We’re using Darwin to optimise darwin.page itself. Every experiment is public: what we’re testing, why we think it’ll work, and what the results are.

Check the live experiments page to watch it happen in real time. Right now we’re testing two concurrent experiments on the hero section: headline framing and CTA copy: with full transparency on confidence levels and data points.

No other A/B testing tool shows you their own experiments running on their own marketing site. We think that says something.

Who Darwin is for

  • Small teams who lost Google Optimize and don’t want to pay £200/month
  • CMS users (Webflow, WordPress, Shopify) who want testing without code
  • Developers who want experiments committed as PRs, not left as runtime patches
  • Anyone who believes their landing page should get better continuously, not once a quarter

Darwin is currently in early access. Join the waitlist to be among the first to try it.

Google Optimize shut down in September 2023. Since then, we’ve spoken to dozens of teams still looking for a replacement that doesn’t cost a fortune or add permanent performance overhead to their sites. Darwin is our answer.