SEO A/B Testing for WordPress — Title and Meta Split Tests
Rank Forge includes an A/B testing feature for SEO titles and meta descriptions. It is time-based: the variant depends on the week, not on the visitor. Rank Forge alternates variant A and variant B week by week, counts the visits of each variant on your site and, with Google Search Console connected, compares the two variants in the editor by the clicks, impressions, click-through rate and average position that Google reports for the page. These are real numbers and not simulated clicks. Rank Forge never picks or applies a winner by itself: you decide.
Why SEO A/B testing is hard
Traditional conversion-rate A/B testing (for landing pages, forms, CTAs) works by showing different versions of a page to different visitors simultaneously and measuring which converts better. This is straightforward when you control the traffic.
SEO A/B testing is different. You can’t show different meta titles to different Google crawlers — and you definitely shouldn’t show different content to Google than you show to users. That’s cloaking, and Google penalizes it.
The correct approach is time-based testing: show one variant for a fixed period, then show the other variant for the same period, and compare CTR across the two periods.
Rank Forge works this way: it shows one variant at a time and, with Google Search Console connected, shows you the numbers of each.
How Rank Forge’s time-based A/B testing works
Step 1: Turn the test on in the post editor
In Rank Forge Pro, open a post or a page, go to the Advanced tab of the Rank Forge box and scroll to the A/B Testing section at the bottom. The first time, the section shows an early-access notice: tick the box that says you understand this is an advanced feature and click Unlock A/B Testing. This is asked once for the whole site. Then tick Enable A/B Testing for this post and fill in:
- Variant A: the title and the meta description the page has now. There is nothing to type for it.
- Variant B: an alternative title and an alternative description, in the two fields under Variant B. A field you leave empty keeps the normal text of the page.
There is no setting for the length of the test or for the metric. The variant changes every week, and the test runs until you turn it off.
Step 2: Rank Forge swaps the variants every week
Rank Forge counts the weeks from the moment you save the post with the test turned on. If you turn the test off, update the post, and later turn it on again, it starts over from week 1 with the counts cleared. While the test stays on, editing its texts changes neither the weeks nor the counts. It shows variant A in week 1, variant B in week 2, and so on. Rank Forge serves the same variant to everyone, Google included, in any one week (a page cache can keep sending the page it made earlier). While a variant is live, the box counts the visits to the page on your site for it (the box calls them page views; visits that look like bots are left out, and the count is approximate). The box tells you to run at least 2 weeks per variant before comparing, because Googlebot re-crawls most pages only every 3 to 10 days.
Step 3: You compare the weeks and choose the winner
With Google Search Console connected and a property chosen, the A/B Testing section shows a Search Console comparison under the page views. For each variant it lists the impressions, clicks, CTR and average position that Search Console reports for the page, taken from the full weeks of that variant. It says that a variant is ahead only when the gap is clear and steady, and otherwise says why there is no winner yet. It never picks or applies a winner by itself: you decide. When the box says Enough weeks have passed (4 weeks and 3 days after the start at the earliest), you can also open Google Search Console, look at the CTR of the page for the weeks of variant A and for the weeks of variant B, and pick the better one. Then:
- Untick Enable A/B Testing for this post and update the post. The page goes back to its normal title and description. The box keeps the last comparison in view, headed Last test: Search Console comparison, when at least one full week was compared. Three days after the end, once the last week has settled, those numbers stay as they are. Switching the test on again starts clean
- To keep variant B, copy its title and description into the SEO Title and Meta Description fields of the Rank Forge box
A test has two variants, A and B. The box has no field for a third.
Why the test is time-based
The same content for everyone at a given moment
Rank Forge does not pick the variant by user agent or by visitor. The variant depends on the week, so a visitor and a crawler get the same content at a given moment, and the content changes from one week to the next. A page cache can keep sending the page it made earlier.
Cleaner data
Simultaneous testing on search results is impossible anyway — Google shows one title and description per URL at any given time. Time-based testing gives you actually-comparable data across two real time periods.
Works with cached indexing
Google caches search result entries. Rank Forge alternates the variants by whole weeks, and the box asks you to give each variant at least 2 weeks, so that Google has time to re-crawl the page and to show the variant that was live.
What data Rank Forge uses
Rank Forge itself keeps one number for each variant: how many times the page was loaded on your site while that variant was live. The box shows it as Page views so far. It is a count of visits to your page, not Google’s impressions: visits that look like bots are left out, and the count is approximate (without a persistent object cache, Rank Forge counts one visit in ten, picked at random, and adds ten). The box compares each count with a floor of 2,000 and tells you whether both variants have cleared it.
The clicks, impressions, click-through rate and average position come from Google Search Console, for this page, week against week. They load after the editor has opened, so opening a post never waits for Google, and they are refreshed about once a day. Only the administrators of the site see them: other editors see the last saved result, or one line that says the numbers are visible to administrators. The Search Console comparison needs Google connected (Rank Forge Pro) with a property chosen.
The comparison is careful about what it counts. Only whole days count, in Pacific Time as Search Console reports them. The first 3 days of each week are left out, because Google re-crawls a page every few days and until it has, the result still shows the previous variant. A week counts only once it is over and Search Console’s data for it is final, about 3 days after it ends; the week still running is never counted. A variant is named as ahead only when all of this holds: each variant has at least 2 full weeks and 1,000 impressions in Search Console, the better CTR is at least 10 percent higher than the other and the gap is not down to chance, the same variant is ahead in every pair of weeks (week 1 against week 2, week 3 against week 4, and so on, at least two pairs), and the average positions of the two variants are within 1.5 of each other. If the average positions differ by more than 1.5, the click-through rates are not compared, because the position, not the title, moves the CTR. Otherwise the box says why there is no winner yet.
How long does a test need to run
Rank Forge changes the variant every week. The box asks for at least 2 weeks per variant, which is 4 weeks in all. Its progress note says Early during weeks 1 and 2, then Collecting data, and Enough weeks have passed only once the Search Console comparison below it has counted 2 full weeks of each variant (a week counts 3 days after it ends), so that is 4 weeks and 3 days after the start at the earliest. From then on it points you to the comparison.
Traffic decides whether that is enough. The box says each variant needs roughly 2,000 page views on your site (bots not counted) before the statistics become meaningful, and the Search Console comparison names a winner only with at least 2 full weeks and 1,000 impressions of each variant (the first 3 days of every week are not counted). Those impressions are Google’s: the times Google showed the page in search results, not page views on your site. On a smaller page the result will be noise, so consider whether A/B testing is worth it at all. The floor of 2,000 page views is a check on the counts of the box only. The Search Console comparison has its own rules, described above: it tests whether the difference in click-through rate could be down to chance, and it shows no confidence figure.
What to test
Best candidates
- Title tag keyword order“Keyword | Brand” vs “Brand | Keyword”
- Number of keywords in titleminimal vs keyword-rich
- Description hook stylequestion vs benefit statement
- CTA in description“Learn more” vs “See how it works” vs no CTA
- Emotional vs factual tone
- Numbers and lists“7 Best…” vs “Best…”
- Year specificity“2026” vs no year
Bad candidates (don’t A/B test these)
- Completely different topics (this is just A vs B content, not SEO testing)
- Pages with very low traffic (not enough data)
- Pages during a core algorithm update (confounding variables)
- Pages undergoing content changes (confounding variables)
Several pages at once
Rank Forge has no bulk A/B tool and no dashboard that lists all tests. You set a test post by post, in the Rank Forge box of each one. Every post has its own switch, its own Variant B text and its own counters, so you can run tests on several posts at the same time.
vs external SEO A/B testing tools
SearchPilot, a dedicated SEO A/B testing tool, is built for enterprise teams and lists its Core plan from $37,500 annually (its Commerce and Enterprise plans are priced on request). Competitor prices as of October 2026, from their public pricing pages. Rank Forge’s A/B testing is included in every paid plan at $44/year and uses the same methodology.
For small and medium WordPress sites, this is more than enough. For enterprise sites doing sophisticated multi-variate testing, dedicated tools still have more features.
Ready to test your SEO titles?
SEO A/B Testing is included in every paid plan.