SEO split testing is in early access and requires a connected Google Search Console property. It is not yet available for self-serve activation — tell us about your site and we will onboard you.
Join early accessMeasure Real Impact on Rankings
SearchPilot-style A/B testing for SEO. Connect Google Search Console, bucket URLs with stratified sampling, apply HTML modifications, and measure impact with Causal Impact statistical analysis. Currently onboarding early-access sites.
Built on
Google Search ConsoleCausal ImpactStratified SamplingBootstrap CIScientifically valid SEO experiments
No more guessing whether SEO changes actually work. Measure with statistical rigor.
Feature 01
Smart Bucketing & URL Templates
Wrenda connects to your Google Search Console data and auto-detects URL templates using path tokenization and clustering. URLs like /products/running-shoes and /products/trail-boots are grouped into the /products/{slug} template automatically.
- GSC OAuth integration for real search performance data
- Stratified sampling: top 10% high, next 30% mid, bottom 60% low traffic
- Balanced control/variant groups across all traffic tiers
- URL template auto-detection from GSC query data
- Per-template bucketing prevents cross-contamination
Stratified sampling algorithm
// Traffic tier assignment Top 10% → HIGH (most clicks) Next 30% → MEDIUM (moderate clicks) Bottom 60% → LOW (long tail) // Each tier split 50/50 HIGH: [control: 5 URLs] [variant: 5 URLs] MEDIUM: [control: 15] [variant: 15] LOW: [control: 30] [variant: 30] // Result: balanced groups with // equal traffic distribution
Auto-detected URL patterns
/products/running-shoes
/products/{slug}
/products/trail-boots
/products/{slug}
/blog/seo-guide-2026
/blog/{slug}
/blog/content-strategy
/blog/{slug}
Bucket assignment
Template: /products/{slug}
CONTROL
HIGH tier | MED tier | LOW tier
VARIANT
HIGH tier | MED tier | LOW tier
# Each tier is split 50/50 so both
# buckets share the same traffic mix.
# A balance score is reported per test.Feature 02
Causal Impact Analysis
Observed
What the variant URLs actually did after launch (Search Console clicks / impressions).
Counterfactual
What the model predicts they would have done with no change, built from the control bucket and the pre-launch baseline.
Causal effect
Observed minus counterfactual, reported as a point estimate with a credible interval — not a single p-value.
Illustrative model structure. No customer results are shown or implied.
Wrenda uses Bayesian structural time-series models to build a counterfactual: what would have happened to your variant URLs if no change was made. By comparing actual performance against the counterfactual, we calculate the true causal effect of your SEO changes -- with p-values and bootstrap confidence intervals.
What the analysis reports
- Point Estimates -- Estimated lift in impressions, clicks, and CTR
- P-Values -- Statistical significance for each metric
- Confidence Intervals -- Bootstrap CI so you know the range of impact
- Counterfactual Curves -- See predicted vs actual performance over time
Response headers on variant URLs
HTTP/1.1 200 OK X-SEO-Test: test_abc123 X-SEO-Bucket: variant X-Modified: title,meta_description Cache-Control: no-cache
Four steps to statistically valid SEO testing
How an early-access test runs, from Search Console connection to Causal Impact analysis.
Step 01
Connect Google Search Console
OAuth into GSC to pull real search performance data -- impressions, clicks, CTR, and position for every URL. This is the ground truth for your experiments.
Step 02
Detect URL templates
Wrenda analyzes your GSC data and auto-detects URL patterns using path tokenization and clustering. Confirm the templates you want to test -- /products/{slug}, /blog/{slug}, etc.
Step 03
Bucket & modify
Stratified sampling splits URLs into balanced control and variant groups. Define HTML modifications -- title tags, meta descriptions, h1s, schema markup, or custom HTML injection.
Step 04
Measure with Causal Impact
After sufficient data collection, Causal Impact analysis compares variant performance against the counterfactual. Get p-values, confidence intervals, and clear lift percentages.
Test what actually moves the needle
Stop guessing. Run controlled experiments on the changes that matter most.
Title Tag Testing
Test different title formats across hundreds of pages simultaneously. Measure the exact click-through-rate impact of adding modifiers, reordering keywords, or changing length.
- A/B test title templates at scale
- Measure CTR lift with statistical significance
- Find the optimal title format per page type
Schema Markup Impact
Add or modify structured data on variant pages and measure whether rich results actually drive more impressions and clicks. No more guessing if schema is worth the effort.
- Test Product, FAQ, HowTo schema types
- Measure rich result impression rates
- Isolate schema impact from other variables
Content Enrichment Validation
Validate that AI-powered content enrichment actually improves rankings. Test enriched vs original content with Causal Impact to prove ROI on content optimization.
- Compare AI-enriched vs original pages
- Measure impression and click lift
- Prove content optimization ROI with data
Full API control
Create tests, bucket URLs, and retrieve Causal Impact results programmatically.
# Step 1 — Create an SEO test with title modification POST api.wrenda.ai/seo-testing/tests Authorization: Bearer <token> { "name": "Product Title Optimization Q1", "template_id": "tpl_products_slug", "modifications": [ { "type": "title", "pattern": "Buy {original} | Free Shipping" }, { "type": "meta_description", "pattern": "{original} — Shop now with free returns." } ] } # Step 2 — After the test window, fetch the Causal Impact analysis GET api.wrenda.ai/seo-testing/tests/{test_id}/analysis # Response shape (values are computed per test — none shown here) { "status": "significant" | "inconclusive" | "insufficient_data", "metrics": { "impressions": { "lift": <number>, "p_value": <number>, "ci_lower": <number>, "ci_upper": <number> }, "clicks": { ... }, "ctr": { ... } } }
GSC
Connected Search Console property required
50 / 50
Control / variant split per traffic tier
14–28 days
Recommended minimum test window
SEO Split Testing FAQs
How A/B testing for SEO works and how to interpret the results.
Standard A/B tools (Optimizely, VWO) test what humans see and measure conversion. Wrenda tests what crawlers see — title tags, meta descriptions, H1s, schema markup, internal linking — and measures organic search performance via Google Search Console. The tooling and statistics are SearchPilot-style: causal-impact analysis on time-series clicks/impressions data.
A Bayesian model that compares the variant URLs' actual performance against a counterfactual: what would have happened if the variant had stayed unchanged. The counterfactual is built from the control bucket's post-launch performance plus the variant bucket's pre-launch baseline. The output is a credible interval for the lift, not just a p-value — so you can say "we are 95% confident this change drove a 4–11% click lift."
Stratified random sampling by traffic tier. We bucket your URLs into high/medium/low traffic tiers using your Search Console data, then split each tier 50/50 between control and variant. This guarantees comparable baselines on both sides and works even when a few high-traffic URLs dominate the distribution.
Title tags, meta descriptions, H1, H2, additional H2 subheadings, schema markup (Product, FAQ, HowTo, Article, Organization etc.), and arbitrary HTML modifications via custom selectors. Tests are gated by URL pattern, so you can run separate tests on /products/* and /blog/* simultaneously.
Yes — Wrenda pulls daily clicks, impressions, CTR and average position data via the Search Console API to power the analysis. We use OAuth so you grant read-only access to a single property, and tokens are stored encrypted at rest.
Minimum 14 days, ideally 28+ days. Search Console data lags by 2–3 days, and seasonal/weekly cycles can mask effects on shorter horizons. The dashboard shows the credible interval narrowing over time so you know when to call it.
Early access
Stop guessing. Start measuring.
We are onboarding a small number of sites with a connected Google Search Console property. Tell us about yours and we will be in touch.
Join early accessAvailable on Pro and Scale plans once your Search Console property is connected