Case Study: How We Transformed Search Visibility Using AI and Structured Data
A real, step-by-step account of how PEMDAS helped a professional services firm go from invisible in AI search to consistently cited in ChatGPT and Google AI Overviews within 90 days.
TL;DR — Quick Answer
By implementing GEO-structured content, JSON-LD schema markup, a topic cluster architecture, and an llms.txt file, a professional services client went from zero AI citations to appearing in 34% of relevant AI search queries within 90 days. Organic traffic increased 67% over the same period.
The Problem: Invisible in AI Search Despite Strong Traditional SEO
In Q4 2024, a professional services firm (an 8-person management consulting practice in the New York Metro area) came to PEMDAS with a specific problem: their website ranked well in traditional Google search but was completely absent from AI-generated answers.
Their target clients — mid-market companies seeking operational consulting — were increasingly using ChatGPT and Perplexity to find and shortlist vendors. The firm was not appearing in any of these queries, while newer, less experienced competitors were.
What Did the Initial Audit Reveal?
PEMDAS conducted a full GEO and content audit over two weeks. The findings showed a technically sound website with good domain authority (DA 41) but content fundamentally misstructured for AI extraction.
Key audit findings:
- Average paragraph length: 187 words (4-8x longer than optimal for AI extraction)
- 0 out of 23 service pages had question-based headers
- No JSON-LD schema on any page
- No llms.txt file
- 12 orphan blog posts with no internal links
- No topic cluster structure — all content isolated
- No TL;DR or direct answer blocks on any content
- Author information missing on all blog posts
What Changes Did PEMDAS Make?
The implementation was organized into three phases over 90 days. Phase 1 focused on technical infrastructure (schema, llms.txt, internal linking). Phase 2 focused on content restructuring. Phase 3 focused on new content creation.
Phase 1 (Days 1-21): Technical Foundation
We added Organization, WebSite, and BreadcrumbList schema site-wide. Created and deployed an llms.txt file. Conducted an internal link audit, connected all 12 orphan posts, and mapped a topic cluster architecture around 4 core service areas.
Phase 2 (Days 22-60): Content Restructuring
We rewrote all 23 service pages with question-based headers, short paragraphs (under 80 words), and direct answer blocks. Added Article and FAQPage schema to all blog posts. Added author bios with credentials to every piece of published content.
Phase 3 (Days 61-90): New Content Production
We published 8 new GEO-optimized blog posts forming two complete topic clusters (4 posts each). Each post included full schema markup, TL;DR blocks, question-based headers, and 5-8 internal links connecting to the cluster pillar.
What Were the Results at 90 Days?
The results exceeded the initial projections. AI citation rates and organic traffic both improved significantly within the 90-day window, with continued growth expected as the new content gains further authority.
| Metric | Before (Baseline) | After 90 Days | Change |
|---|---|---|---|
| AI citations (ChatGPT/Perplexity) | 0 | 34% of tracked queries | +34 percentage points |
| Google AI Overview appearances | 0 | 11 featured | +11 |
| Organic traffic (monthly) | 1,840 sessions | 3,074 sessions | +67% |
| Average ranking position | 14.2 | 8.7 | +5.5 positions |
| Inbound leads (monthly) | 3 | 11 | +267% |
| Orphan pages | 12 | 0 | Eliminated |
| Pages with schema markup | 0 | 31 | Full coverage |
What Was the Biggest Single Driver of Results?
The single highest-impact change was restructuring the service pages with question-based headers and direct answer blocks. This change alone accounted for an estimated 60% of the AI citation improvement, based on a controlled test with two comparable pages.
Schema markup was the second-highest impact change. Pages with FAQPage schema began appearing in Google AI Overviews within 3 weeks of implementation — significantly faster than the organic traffic improvements.
What Were the Lessons Learned?
- Technical SEO authority does not transfer automatically to AI search — you must restructure content separately
- Schema markup delivers faster results than content changes for AI Overviews specifically
- Topic clusters show compounding effects — traffic grew faster in months 2-3 than month 1
- llms.txt adoption is early-stage but several AI agents explicitly cited it in crawl logs
- Author credential pages meaningfully improved E-E-A-T signals in both Google and AI systems
The fastest path to AI search visibility is not creating more content — it's restructuring your existing content to be extractable. Most websites have the raw material needed to dominate AI search. They just haven't formatted it correctly.
Frequently Asked Questions
Is this case study replicable for other industries?
Yes. The same methodology has been applied across commercial real estate, healthcare administration, financial services, and legal practices. The tactics are industry-agnostic — AI extraction logic doesn't distinguish between industries.
How do you measure AI citation rates?
We use a combination of manual query testing (asking ChatGPT and Perplexity target queries and recording whether the client is cited), Otterly.ai for automated brand mention tracking in AI, and Google Search Console for AI Overview impression data.
What was the total investment for this engagement?
The full 90-day engagement (audit, technical implementation, content restructuring, and new content creation) totaled $22,500. The client's inbound lead increase from 3 to 11 per month, at an average deal value of $35,000, produced a 12-month ROI of approximately 15x.
Ready to Implement This for Your Business?
PEMDAS helps SMBs adopt AI the right way — strategy first, then execution.
Get Your Free AI Readiness Score