Can You Run A/B Tests for SEO?

Not really, but we’ll show you how to run a “B/A” test!

A/B testing is a hallmark of good online marketing. Create two versions of something — control case is A, test case is B — and see which one performs better.

A/B testing examples

Testing ad copy and landing pages in PPC campaigns may be the most common use for A/B testing. It’s simple and effective. You can gather meaningful results with no drawbacks.

But there are so many more things you can test:

Lead generation forms: Create two forms, send half the people to one and half to the other, and see which one people fill out more often.
Landing pages for social media ads: Create two landing pages for the same ad, and see which one converts more users.
Button styles: Try variations of button text, color, size, and placement, and see which ones get the most clicks.
Special offers: If you offer free shipping, a free trial, discounts, buy-2-get-1-free, or whatever else, how much does it entice users to convert?
Price: While some companies find this unethical — Amazon, for instance, says they’ll A/B test anything except pricing — there’s no law against offering different prices to different customers.

But can you A/B test your SEO strategy?

It’s the forbidden fantasy of SEO professionals everywhere. If only we could!

Imagine how different SEO would be if we could quickly tell which meta title, meta description, H1, word count, content, page design, and linking structure the Google algorithm rewards more.

Unfortunately, we can’t.

SEO is a gamble. You create your site the best you can, you play your hand, and Google decides who wins.

Why you can’t A/B test for SEO

Google will crawl only one version of a given page, and they will rank it according to the algorithm.

You might want to create two very similar pages and see which one ranks better … but you shouldn’t. Google’s algorithm aims to identify a single page from a given site that best suits a keyword. When it sees two pages doing the same thing, it may grade them both down.

Any tricky method you’ve heard of — “cloaking” for example — DOES NOT work. Google’s algorithm is wise to it.

If you’ve been paying attention to what Google reps say anytime there is an algorithm update, you know this much: Google wants to keep us honest.

So what CAN you do?

You can’t A/B test, but you can “B/A” test: (B)efore and (A)fter.

Keep track of how something is performing now (before), make a change, and then see what the change did (after).

“But wait! If I change my pages to test SEO and I fall down the rankings, I can’t get my old rankings back, right?”

Right, you can’t, or at least you can’t bet on it. It’s much easier to stay on top than it is to get there.

“And wait again! Doesn’t SEO take time? If I make changes, I’ll wait weeks or months to see results! I don’t have time for that!”

Yes, SEO can take time. In addition, your competitors are also working on their SEO game, and Google’s algorithm is changing EVERY DAY. Waiting for rank changes will drive you mad.

So, let’s not use the search results rankings to gauge our tests. Let’s do something else instead…

Test for individual ranking factors

We know that Google pays attention to certain metrics:

Time on page: how long someone stays on the page
Time on site: how long someone stays on the site
Bounce rate: how often someone quickly goes back to the search results after seeing your page
Pages per session: how many pages someone views when visiting your site
Scroll depth: how far down the page users go

These are things you can A/B test, or rather “B/A test,” seeing what happens before and after you make changes.

If you have enough traffic, you can get fast results from testing these elements.

If the test fails, and you get a temporary dip in, say, time on page, it won’t totally ruin your SEO. Change it back once your test results are reliable.

If you see improvements, then great! Congratulations, you’re improving your site and the user experience, which should improve your ranking ability.

“But wait!”

Oh, it’s you again.

“What are these actual changes? What am I testing out?”

Marketers and site managers may want to test out anything. Here are some ideas:

Update the design

Change your colors from green to blue
Make the text larger and use a new font

Add new stuff

Buttons
Images
Videos
Widgets
Links
Ads

Change the content

Write new headings
Write a new meta title
Write new content

Some of these pose more of that SEO risk we mentioned earlier.

If you change your color scheme, Google won’t mind, and you could see better engagement. It’s pretty low risk.

If you change the content, headlines, and especially your meta title or URL, then you’re significantly changing how Google grades the page. It’s higher risk and potentially higher reward, which leads us to…

Test for click-through rate

In a more direct — and risky — example of A/B testing for SEO, you can change your meta title and/or meta description and see if you get more clicks from a search engine.

If you get fewer clicks, change it back.

If you get more clicks, that’s a great sign, and your page is improved in Google’s eyes, assuming you’re still getting good user engagement metrics.

The key here is isolation. When you test something, make sure you’re testing ONLY that one thing.

So, don’t just gather impressions and clicks for the page. Get specific. Gather impressions and clicks from a single keyword.

Let’s say you’ve ranked #7 for a specific keyword for the past six months. Use Google Search Console or another tool to determine the impressions and clicks for your listing on that keyword. After you change your title or description, watch these same metrics over time.

If the keyword moves up or down, the test is over. You’re no longer testing the same thing as you were before.

Be creative

Marketing requires a lot of creativity anyway, so even if you’re looking for hard data to inform your strategy, you can be creative in getting it.

While you can’t technically A/B test SEO, you can definitely find out valuable information, and the right information goes a long way.