Global DeliveryRemote SEO & digital marketing team since 2015 · Serving US businesses from around the world
    SEO Craze Labs & Testing

    We test search strategies instead of blindly repeating SEO advice.

    The search landscape is full of recycled dogma. Through SEO Craze Labs, we conduct live controlled experiments across our own portfolio of websites to evaluate what genuinely moves organic rankings, AI citations, and crawl performance.

    Our Testing Architecture

    01. Hypothesis

    Testable, bounded assumption

    02. Test Setup

    Control vs test URL groups

    03. Data Capture

    GSC, API logs & LLM scrapes

    04. Result Analysis

    Statistically observed delta

    05. What We Learned

    Actionable client application

    Empirical Research

    Published Search & AI Experiments

    All tests conducted on live production environments. Re-tested across quarterly search updates.

    01AI Search / Schema
    August 2026Completed

    Does adding FAQ schema improve AI citations in ChatGPT and Perplexity?

    Hypothesis

    Pages containing clean Question/Answer JSON-LD markup alongside visible on-page answers will be cited significantly more often by LLMs than identical content without structured schema.

    Test Methodology

    Tested 32 comparable commercial pages across 4 test websites. 16 pages received nested FAQPage JSON-LD schema with concise 45-word answers; 16 control pages retained identical on-page text without schema markup. Queried ChatGPT-4o and Perplexity with 80 target prompt variations.

    Data Captured

    80 prompt evaluations across 32 URLs over a 60-day observation window.

    Result: Perplexity cited schema-enabled URLs in 42% of responses vs 18% for control pages (2.3x increase). ChatGPT cited schema pages in 28% vs 21% (modest improvement). When cited, both engines directly lifted answers matching the schema text 74% of the time.

    What We Learned & Practical Application

    FAQ schema provides a machine-readable extraction boundary that generative models prioritize during retrieval-augmented generation (RAG). Keep answers between 35-50 words for optimal extraction.

    02Internal Linking
    July 2026Completed

    Can strategic internal linking move a competitive page without new backlinks?

    Hypothesis

    Re-architecting internal links from high-authority existing pages to an underperforming target page using exact and partial match anchors will increase ranking positions without building a single external backlink.

    Test Methodology

    Selected an underperforming service page stuck on Page 3 (Position 24-28). Without acquiring external links or modifying target page copy, added 14 contextual in-content links from top-traffic blog posts and category hubs with varied anchor text. Monitored GSC for 45 days.

    Data Captured

    Target keyword group: 12 keywords with combined monthly search volume of 4,800.

    Result: The target URL rose from Position 26 to Position 9 within 31 days. Impressions increased by 310% and organic clicks increased by 225% without acquiring any new external referring domains.

    What We Learned & Practical Application

    Internal PageRank distribution is often severely underutilized. Contextual in-body links from pages with existing traffic pass substantial topical relevance and crawl priority.

    03Local SEO / Maps
    June 2026Completed

    Do dedicated service-area pages improve Google Maps 3-Pack visibility?

    Hypothesis

    Publishing localized service-area landing pages with localized schema, landmarks, and geotargeted FAQs will expand Google Maps proximity ranking reach beyond the physical address.

    Test Methodology

    Created 6 distinct service-area pages for adjacent sub-markets (5-15 miles from verified GBP address). Embedded unique driving directions, local landmarks, project examples, and local business schema linking back to the verified GBP CID URL.

    Data Captured

    Local Falcon 7x7 geo-grid scans across 49 data points over 90 days.

    Result: Average Local Rank (AGR) improved from 14.2 to 5.8 across the targeted 15-mile perimeter. Organic ranking for 'service + city' jumped to Top 3 across 4 out of 6 sub-markets.

    What We Learned & Practical Application

    Google uses website relevance to validate proximity trust. High-quality location pages with authentic local proof expand the geographic radius where Google feels confident showing your Map Pack listing.

    04AI Engine Comparison
    May 2026Completed

    ChatGPT vs Gemini vs Perplexity: Which pages do they actually cite?

    Hypothesis

    Different AI search engines rely on fundamentally different retrieval mechanisms: Gemini favors Google ecosystem authority and YouTube; Perplexity favors structured data and recency; ChatGPT favors consensus brand mentions and high-authority publications.

    Test Methodology

    Queried 100 identical commercial buying queries across all 3 engines. Recorded all 540 cited source URLs and analyzed domain age, backlink count, schema presence, content length, and publication date.

    Data Captured

    540 cited URLs analyzed across 100 queries.

    Result: Perplexity cited pages updated within 90 days 68% of the time, and prioritized pages with tables/bulleted data. Gemini cited Google-indexed sources and YouTube videos 3.1x more than ChatGPT. ChatGPT favored established domains (DR 60+) and community consensus sources (Reddit/Quora).

    What We Learned & Practical Application

    AEO cannot be treated as a monolith. A comprehensive GEO strategy must balance fresh, tabular data (for Perplexity), structured Google ecosystem proof (for Gemini), and digital PR brand mentions (for ChatGPT).

    05Entity SEO
    April 2026Completed

    Does verified author and entity information improve AI visibility?

    Hypothesis

    Adding explicit Author schema (with SameAs links to LinkedIn, Wikidata, and published bylines) increases the likelihood of a page being cited in AI Overviews compared to an anonymous byline.

    Test Methodology

    Tested 20 articles in a competitive finance/business niche. Half were attributed to a verified author with full Person schema, SameAs profile links, and credentials; half were published under generic 'Editorial Staff' without Person schema.

    Data Captured

    20 test URLs tracked over 75 days in Google Search Console and AI Overview trackers.

    Result: Verified author pages appeared in Google AI Overviews 44% more frequently and maintained 32% higher average click-through rates.

    What We Learned & Practical Application

    Google's AI evaluation algorithms look for entity validation to mitigate hallucination risks. Attributing claims to verifiable persons provides the required E-E-A-T confidence signal.

    06Programmatic SEO
    March 2026Completed

    How quickly does Google index programmatic SEO pages vs standard pages?

    Hypothesis

    Programmatic pages generated from structured databases will index at the same speed as manually created pages if they possess at least 40% unique data points and proper internal linking.

    Test Methodology

    Generated 200 programmatic location-service pages. Group A (100 pages) had generic boilerplate with only city swapped (thin template). Group B (100 pages) included unique local statistics, custom pricing estimates, localized reviews, and hub linking.

    Data Captured

    200 URLs submitted via XML sitemap, monitored via Google Search Console URL Inspection API for 60 days.

    Result: Group B achieved 92% indexation within 14 days, with 64% ranking in Top 20 by day 45. Group A achieved only 28% indexation, with 54% flagged as 'Discovered - currently not indexed'.

    What We Learned & Practical Application

    Programmatic SEO is not dead—lazy templating is. Google's helpful content systems instantly identify thin boilerplate. When pSEO pages contain genuinely unique structured data and utility, indexation is rapid and durable.

    07Content Strategy
    February 2026Completed

    What happens when AI-written content is substantially human-edited?

    Hypothesis

    Raw AI content will struggle to retain rankings through core updates, while AI-assisted content rewritten with proprietary data and human editorial oversight will outperform 100% human-written content due to comprehensive topic coverage.

    Test Methodology

    Monitored 3 batches of 15 articles each: Batch 1 (100% raw AI generated), Batch 2 (AI outline + 100% human-researched and edited with original quotes), Batch 3 (Traditional 100% manual freelance writer). Tracked performance through a Google Core Algorithm Update.

    Data Captured

    45 articles tracked over 6 months across 3 distinct domains.

    Result: Batch 1 lost 68% of ranking traffic following the core update. Batch 2 gained 42% organic traffic and achieved the highest keyword density without triggering spam flags. Batch 3 performed well (+24%) but took 4.5x longer to produce.

    What We Learned & Practical Application

    AI is best used for exhaustive topical clustering, outline generation, and research aggregation—while humans must provide unique opinions, verified facts, original imagery, and editorial punch.

    08Conversion Rate Optimization
    January 2026Completed

    Do comparison pages outperform standard service landing pages for conversion?

    Hypothesis

    Commercial intent comparison pages ('SEO vs PPC', 'Agency vs In-House') will convert visitors into booked consultations at a higher rate than generic service overview pages.

    Test Methodology

    Split 1,200 commercial organic visitors across standard service pages ('SEO Services') vs structured comparison pages ('SEO Agency vs In-House Team' and 'SEO vs Google Ads'). Tracked form submissions and consultation bookings.

    Data Captured

    1,200 organic visits, 74 conversion events.

    Result: Comparison pages converted at 8.4% compared to 3.2% for standard service pages (2.6x higher conversion rate). Prospects from comparison pages spent 2.8x longer on page and arrived at initial calls with fewer price objections.

    What We Learned & Practical Application

    Buyers seeking comparison content are already in the decision-making stage of their journey. Providing an honest, structured evaluation captures high-intent demand that generic marketing pitches miss.

    Have a hypothesis you want tested on your site?

    We test search strategies for ambitious companies and partner agencies that need rigorous data behind their marketing.