How We Measure AI Citation Growth
Traditional search engine optimization measures rankings in terms of "blue link" positions. Generative Engine Optimization requires a new metrics framework: **AI Citation Share-of-Voice (SOV)** and **LLM Referral Traffic**. AI platforms retrieve information dynamically using RAG (Retrieval-Augmented Generation), meaning that a website ranking #1 in classic search may get completely skipped by an LLM scraper if it lacks structural citation triggers.
Our case studies document how restructuring existing pages (on-page GEO) and securing third-party co-occurrence placements (off-page GEO) directly drives citation inclusion. Over the last 12 months, we have optimized visibility for over 1,200 brands using our six primary GEO services, compiling clear evidence of how AI visibility translates into business revenue.
Case Study 1: +1,150% AI-Referred GA4 Traffic for a B2B SaaS Brand
The Challenge:
An enterprise subscription scheduling SaaS with excellent traditional Google rankings realized that Perplexity and ChatGPT Search were completely skipping their platform. When users prompted AI engines for queries like *"best team scheduling software for distributed B2B sales teams"*, competitors were consistently cited while the client received zero mentions. Their analytics showed a steady decline in organic traffic as users shifted informational search queries to LLMs.
The Strategy:
Our technical team ran a complete baseline GEO audit, identifying that their feature comparison pages were written in vague, sales-oriented language that LLM scrapers struggled to verify. We started with a baseline GEO audit and implemented a two-step optimization campaign:
- On-Page Restructuring: We optimized 20 high-value feature comparison pages. We restructured paragraph openings into answer-first formats, built custom comparison tables comparing key product specifications (pricing, features, API integrations), and deployed deep JSON-LD SoftwareApplication structured schema maps.
- Off-Page Co-Occurrence: We placed verified product reviews and feature details across Reddit SaaS threads, B2B software directories, and high-traffic comparison listicles. This established the third-party co-occurrence references LLMs check to verify recommendation trust.
The Results:
Within 60 days of deploying the optimized copy and semantic schema, Perplexity's citation engine began pulling data directly from the client's restructured tables. The B2B scheduling brand went from being completely invisible in LLM recommendations to securing a 74% citation share-of-voice across B2B SaaS comparison prompts, generating a 1,150% increase in active AI-referred referral sessions in GA4.
Technical Implementation Details:
Our first step was replacing the vague, benefit-driven product copy with high-density technical attributes. LLM scrapers bypass promotional language (e.g., "our world-class scheduling dashboard helps you win") and search for specific capabilities. We constructed a 5x8 HTML table comparing custom SSO (SAML), automated round-robin routing, Salesforce/HubSpot CRM sync, multi-region time-zone detection, and pricing-per-seat. This table was marked up with semantic table headers (<th>) and table data cells (<td>) using clean CSS class identifiers.
Simultaneously, we built a comprehensive JSON-LD schema file containing the SoftwareApplication type. Within this schema, we defined the applicationCategory as "BusinessApplication", designated the operatingSystem as "All", and nested detailed offers objects representing their pricing plans. Crucially, we populated the featureList array with key target terms discovered during our prompt analysis (e.g., "automated routing", "SSO integration", "Salesforce sync").
LLM Evaluation & Query Pathways:
To verify the citation pathway, we monitored crawl requests from Perplexity's user-agent (PerplexityBot). During testing, Perplexity executed a RAG query matching the user prompt: "compare enterprise scheduling tools with API integrations." The retrieval model retrieved our client's page and extracted the comparative HTML table. The reranking cross-encoder classified this structured data block as the most complete, noise-free representation of tool features available on the web. Consequently, Perplexity generated a response outlining the client's features, citing our client's URL for 8 out of the 10 listed capabilities.
Case Study 2: +420% Citation Share-of-Voice for a New York Corporate Law Firm
The Challenge:
A boutique corporate law firm specializing in B2B litigation in New York City wanted to target corporate clients looking for representation. When users asked ChatGPT Search or Gemini for recommendations like *"who are the best corporate litigation lawyers in Manhattan for tech startups"*, established corporate legal giants dominated the lists. The client had strong local organic visibility but zero recommendation presence inside LLM platforms.
The Strategy:
Startups and founders do not use search engines the way they used to; they prompt AI engines for direct answers, expecting recommendations backed by credentials and proof. Our team executed a targeted Local GEO Campaign:
- Google Business Profile (GBP) & Entity Alignment: Standardized NAP (Name, Address, Phone) data, licensing numbers, and city served regions across all directory platforms to resolve their business entity with Google Business Profile AI.
- On-Page Credential Verification: Restructured their primary Manhattan service pages to lead with direct answer headings, license numbers, credentials (attorneys' bios, education), and direct verification links to state bar associations. Deployed LocalBusiness schema markup.
- Third-Party Listicle Placement: Featured the firm inside high-authority NYC corporate lawyer roundups and startup directory sites (establishing co-occurrence links).
The Results:
By standardizing entity identifiers and placing the firm in startup resources, we increased their brand co-occurrence score. ChatGPT Search began citing the firm's Manhattan landing page as evidence for startups needing corporate lawyers. The firm secured a steady 420% increase in recommendation share-of-voice, bringing in qualified start-up litigation briefs worth over $14,000 in monthly retainer value.
Technical Implementation Details:
Local AI visibility requires perfect entity resolution across the web. If an LLM finds conflicting name, address, or phone (NAP) data, its confidence score drops, and it refuses to recommend the business. We ran a script standardizing their NAP data across 15 high-authority local databases and legal profiles. On the firm's website, we restructured the NYC litigation landing pages to display the firm's license numbers, practice area descriptions, and direct links to the State Bar of New York.
In the header of the NY landing page, we injected a nested JSON-LD schema combining LocalBusiness and Attorney types. We populated the knowsAbout property with explicit entities: "Corporate Litigation", "B2B Contract Disputes", "Intellectual Property Litigation", and "Manhattan Tech Startups". We also included a memberOf node referencing the New York State Bar Association, linking directly to the association's registration page for the firm's partners.
LLM Evaluation & Query Pathways:
We tested the system using OpenAI's GPT-4o model via ChatGPT Search, posing the prompt: "who are the best corporate litigation lawyers in Manhattan for tech startups." The engine queried local indexes and web pages. It discovered the firm's NY landing page and resolved its entity against the State Bar directory and Google Business Profile. Because the NAP data matched perfectly across all external links and schemas, the LLM resolved the entity with a high confidence score. ChatGPT's generated response recommended the firm as a "highly verified corporate litigation specialist in Manhattan," complete with direct citation links.
Case Study 3: 6.4x ROI for a Shopify E-Commerce Retailer
The Challenge:
An e-commerce brand selling eco-friendly pet products faced massive competition in paid advertising. They wanted to capture organic traffic from customers querying AI engines for shopping research, such as: *"what are the safest, most durable eco-friendly pet toys for dogs that chew"* or *"compare eco-friendly pet toy brands by price and material"*.
The Strategy:
AI search shopping engines (like ChatGPT Search and Google AI Overviews) recommend products based on verified user reviews, material attributes, and comparison rankings. We launched a product-level GEO campaign:
- Product Schema Integration: Added Product, Brand, and AggregateRating schema to every product detail page, ensuring materials (e.g., organic hemp, natural rubber) and dimensions were clearly labeled.
- On-Page Attributes Table: Injected direct specification lists and comparison blocks showing product lifespan, price-per-use, and eco-certifications.
- Entity Placement Placements: Placed product reviews and features on pet-care publications, Reddit pet communities, and established pet product listicles (DR30+ to DR50+ authority sites).
The Results:
The schema details and third-party listicle features pushed ChatGPT and Google AI Overviews to recommend their pet toys during shopping research. Scrapers extracted product attributes directly from their tables. The brand achieved 37.6K direct citations across 38 product pages in 3 months, generating a 6.4x return on investment based on direct sales from AI-referred customers.
Technical Implementation Details:
E-commerce GEO depends on the visibility of product specifications. When users ask AI engines to compare products, the engines search for structured product data. We optimized the Shopify theme to inject deep Product, Brand, and AggregateRating JSON-LD schemas. These schemas defined the exact materials (e.g., "100% Organic Hemp Fiber", "Natural Vulcanized Rubber"), dimensions, weight, and chewing durability ratings.
On the front-end product templates, we replaced loose marketing copy with structured specification blocks and a clear comparison FAQ. We detailed how the toys undergo independent chew-durability tests, certified by pet safety organizations. We also secured product placements in off-page publications and pet-owner communities, establishing high-trust co-occurrence signals linking the brand name with keywords like "safest eco-friendly pet toys."
LLM Evaluation & Query Pathways:
We monitored Google AI Overviews and ChatGPT Search for the query: "what are the safest, most durable eco-friendly pet toys for dogs that chew." The AI retrieval engine scanned product pages and external reviews. The JSON-LD schema provided clear, structured evidence of materials and safety ratings. The LLM parsed these attributes and recommended the brand's pet toys as a top eco-friendly choice, citing the product detail page as proof of materials. This direct citation resulted in a 6.4x sales ROI.
Vertical Performance Summary
| Vertical | Primary Services Used | SOV Growth | ROI / Leads Value | Execution Period |
|---|---|---|---|---|
| B2B Subscription SaaS | Content Optimisation + Schema + Community | +1,150% sessions | 3.2x ROI | 60 Days |
| NY Corporate Law Firm | GEO Audit + Entity Building + Listicles | +420% SOV | $14K+ monthly value | 45 Days |
| Shopify E-Commerce | Entity Building + Link Building + Product Schema | 37.6K citations | 6.4x ROI | 90 Days |
Our Rigorous AI Citation Testing & Verification Methodology
Achieving high citation rates is not a matter of luck. It requires a systematic testing loop. When we optimize a client's page, we run pre- and post-optimization query audits to measure performance. We utilize API access to major LLM providers (including OpenAI's GPT-4o, Anthropic's Claude 3.5 Sonnet, and Perplexity's sonar models) to simulate user search behavior. We test a batch of 50-100 target prompts containing variations in phrasing, intent, and local modifiers.
For each query, our scripts analyze the LLM's raw JSON response. We parse the generated text to identify: (1) if the client's brand is mentioned, (2) the position of the recommendation, (3) the presence of citation links, and (4) the exact URL cited. By analyzing this data, we calculate a client's AI Citation Share-of-Voice (SOV). If the client is cited in 30 out of 100 queries, their baseline SOV is 30%.
Post-optimization, we monitor crawler access logs to ensure search bots have fetched the updated page structures. We then rerun the query batch. If the SOV increases and remains stable over a 14-day window, the optimization is marked as successful. If the citation rate does not hit our target tier, we run a comparative RAG audit against the cited competitors, adjusting paragraph semantic density, table layouts, and schema attributes until the citation triggers are successfully activated.
Our 15 Algorithmic Authorship Rules
Every case study success is built upon compliance. We write and audit all deliverables against our strict, proprietary list of 15 Algorithmic Authorship Rules to ensure that copy meets LLM scraper requirements. We remove marketing filler, structure definitions, validate all JSON-LD attributes, and double-check NAP matching to guarantee citation reliability.
Explore Our Managed campaigns →