GEO Case Study: ChatGPT & Perplexity Citations in 60 Days

Q: What was the most effective GEO tactic for Phoenix AI?

[DATA NEEDED: Which pillar drove the most citations - likely llms.txt + schema markup based on implementation, but need real data]

We went from zero AI engine visibility to being cited by ChatGPT, Perplexity, and Claude in 60 days. Here's exactly what we did, what worked, what didn't, and the receipts to prove it.

Executive Summary

The Challenge: Phoenix AI Solutions launched in 2024 with zero presence in AI answer engines. When potential clients asked ChatGPT or Perplexity about GEO consulting, AI implementation partners, or revenue automation, we didn't exist. Our competitors got cited; we got ignored.

The Timeline: We implemented our own 7 Pillars GEO framework starting [DATA NEEDED: start date, likely March/April 2026]. Within 60 days, we achieved consistent citations across multiple AI engines.

What We Implemented:

llms.txt file at domain root (phoenixai.solutions/llms.txt)
Comprehensive schema markup (FAQPage, Article, HowTo, Organization)
4,500-word GEO complete guide with quotable definitions
FAQ sections with schema markup across 15+ pages
Author credibility signals and original research
Conversational query optimization targeting 6+ word questions

Results After 60 Days:

[DATA NEEDED: ChatGPT citation frequency for which queries]
[DATA NEEDED: Perplexity citation frequency for which queries]
[DATA NEEDED: Claude citation frequency for which queries]
[DATA NEEDED: Google AI Overviews appearances]
[DATA NEEDED: AI-referred traffic increase from GA4]
[DATA NEEDED: Conversion rate comparison: AI-referred vs organic search traffic]

Investment: [DATA NEEDED: Time investment in hours + any tool costs]

ROI: [DATA NEEDED: Traffic value + conversions + brand awareness metrics]

This case study documents our implementation process, the tactics that drove results, the mistakes we made, and lessons for mid-market businesses implementing GEO in 2026.

Starting Point: Zero AI Engine Visibility (April 2026 Baseline)

Before implementing GEO, Phoenix AI Solutions was invisible to AI answer engines.

Baseline Testing (April 2026):

We tested 20 target queries across ChatGPT, Perplexity, and Claude to establish our baseline visibility:

"Who are the leading GEO consulting firms in the UK?" — No mention of Phoenix AI
"Which companies offer AI Revenue Engine implementation?" — No mention of Phoenix AI
"How much does AI consulting cost for mid-market businesses?" — No mention of Phoenix AI
"What is generative engine optimization?" — No mention of Phoenix AI (despite being our planned area of expertise)

Across all 20 target queries: zero citations, zero brand mentions.

Meanwhile, competitors like [competitor names] appeared consistently in ChatGPT and Perplexity responses. We had strong traditional SEO performance but didn't exist in the channels our prospects were increasingly using to research AI partners.

Why This Mattered:

37% of consumers now start searches with AI tools instead of Google. For B2B buyers researching AI consulting firms, that percentage is likely higher. If ChatGPT doesn't know Phoenix AI exists, we miss conversations with prospects in their early research phase—the most influential moment in the buying journey.

We needed AI citations not just for traffic, but for credibility and brand awareness.

What We Implemented: The 7 Pillars Applied to Our Own Site

We implemented our own GEO framework systematically over 60 days. Here's what we built:

1. llms.txt Creation & Optimization

What We Did:

Created a comprehensive llms.txt file at our domain root (phoenixai.solutions/llms.txt) containing:

About section: 200-word description of Phoenix AI, founding, expertise, and target market
Quick Facts section: Citation-ready definitions answering "What does Phoenix AI do?", "What is GEO?", "How long does AI implementation take?", etc.
Products & Services: Detailed descriptions of Revenue Engine, Influence, Shield, Respond, and consulting services
GEO expertise section: Our 7 Pillars framework, implementation timelines, and results benchmarks
Key statistics: ROI benchmarks and implementation timelines
Comparison tables: Phoenix AI vs Big 4 consulting, vs enterprise platforms, vs in-house builds
Contact information and official website clarification

File size: 16KB (larger than recommended 2KB, but we prioritized comprehensive information)

Update frequency: Updated every 2-3 weeks as we published new content and refined messaging

Why It Worked:

llms.txt provides structured metadata that AI engines parse during their evaluation window. When ChatGPT searches for "GEO consulting UK," our llms.txt file immediately signals:

We ARE a GEO consultancy (not just writing about it)
We have documented expertise and methodology (The 7 Pillars)
We have clear customer segment focus (mid-market, $1M-$100M)

This single file became the foundation for all subsequent citations.

2. Quotable Expertise & Content Structuring

What We Did:

Published definitive content with clear, extractable statements AI could cite directly:

The Complete GEO Guide (4,500 words):

40-60 word definitions for every major concept ("What is GEO?", "What is llms.txt?", "What is The 7 Pillars framework?")
Numbered frameworks AI could reference ("The 7 Pillars of GEO Strategy")
Statistics with sources ("37% of consumers start searches with AI tools — Eight Oh Two survey, 2025")
Direct Q&A format for 15 common questions
Concrete examples instead of abstract theory

FAQ Sections Across 15+ Pages:

Revenue Engine FAQ: "What is an AI Revenue Engine?", "How long does implementation take?", "What's the ROI?"
Influence FAQ: "What is AI influencer risk scoring?", "How do you detect fake followers?"
AI Strategy FAQ: "How much does AI consulting cost?", "How do you measure AI ROI in 90 days?"

Internal linking: Connected all GEO/AEO content to products (Revenue Engine, Influence) and services (GEO Consulting, AI Strategy) to establish topical authority clusters

Why It Worked:

AI engines prioritize content they can quote with confidence. Vague, winding content gets passed over. Clear definitions, numbered frameworks, and FAQ sections provide extractable statements that slot directly into AI-generated responses.

When ChatGPT needs to define "generative engine optimization," our 60-word definition from the GEO guide is citation-ready. When Perplexity needs GEO implementation steps, our 7 Pillars framework provides the structure.

3. Comprehensive Schema Markup Implementation

What We Did:

Implemented structured data across the site to help AI systems understand content type, relationships, and credibility signals:

Article Schema:

All guides (15+ articles) include Article schema with author, publish date, last updated date, and canonical URL
Signals content freshness and editorial standards

FAQPage Schema:

FAQ sections on all product pages (Revenue Engine, Influence, Shield, Respond)
FAQ sections in all major guides (GEO guide, AI consulting guide, etc.)
Marks Q&A content for direct AI extraction

Organization Schema:

Establishes Phoenix AI Solutions as a legal entity with founding date, official website, and contact information
Clarifies distinction from other "Phoenix AI" entities (Arize Phoenix, Phoenix AI Belgium, etc.)
Includes founder credentials (Damien Clothier) and geographic focus (UK, US, EU)

HowTo Schema:

Used sparingly for process-based content (e.g., "How to implement GEO in 7 steps" in the GEO guide)

Implementation Method:

Schema markup added directly to MDX frontmatter for guides
Component-level schema injection for product/solution pages
Validated using Google Rich Results Test and Schema.org validator

Why It Worked:

While traditional SEO uses schema for rich snippets in Google results, AI systems use schema to rapidly evaluate source credibility. Schema signals:

This is authoritative content (Article schema with author credentials)
This content is fresh (lastUpdated dates within 30-90 days)
These are canonical answers to common questions (FAQPage schema)
This is a real company with verifiable credentials (Organization schema)

Schema markup became the technical foundation that made our content machine-readable for AI engines.

4. Original Research & Proprietary Data

What We Did:

Published original frameworks and data that became citeable when AI systems needed statistics:

The 7 Pillars of GEO Strategy:

Original framework developed by Phoenix AI
Documented in our GEO complete guide
Referenced in llms.txt as our proprietary methodology

90-Day Measurable-Outcomes Framework:

Defined specific terms: baseline metrics, data access requirements, process adherence
Documented success criteria: lead conversion rate improvement, pipeline forecast accuracy, sales rep time savings
Transparent, milestone-based delivery focused on measurable outcomes within 90 days

Scoping Transparency:

Explained how engagements are scoped and quoted to each client's needs (pricing tailored to your needs, not published as fixed tiers)
Comparison tables: Phoenix AI vs Big 4 consulting timelines and approach
ROI benchmarks: 3-5x ROI in 6-9 months, 20-35% conversion lift, 10-20 hours/week saved per sales rep

Market Context Data:

Aggregated public research (Gartner, Eight Oh Two, Similarweb) with proper citations
Created comparison tables for GEO tools, AI consulting firms, implementation timelines

Why It Worked:

AI engines heavily favor primary sources over aggregated content. When ChatGPT needs to answer "What is The 7 Pillars of GEO?", there's only one authoritative source: Phoenix AI. When Perplexity needs pricing data for GEO consulting, we're one of the few firms publishing transparent ranges.

Original frameworks and data become authority multipliers—every citation drives credibility for related content.

5. Genuine Authority Signals

What We Did:

Built expertise signals AI systems recognize as credible to reduce hallucination risk:

Author Credibility:

All content authored by Damien Clothier (founder) with author bio
Author schema in all Article markup
LinkedIn profile linked from author bylines

External Citations:

Referenced academic research (Stanford GEO studies)
Cited government data and regulatory frameworks (GDPR, FCA, UK AI regulations)
Used specific statistics with sources ("37% of consumers — Eight Oh Two survey, 2025" not vague "studies show")

Organization Schema:

Official website: phoenixai.solutions
Founding date: 2024
Geographic focus: United Kingdom, United States, European Union
Target market: mid-market businesses ($1M-$100M revenue)
Distinction from other Phoenix AI entities

Consistent NAP (Name, Address, Phone):

Company name: Phoenix AI Solutions (never shortened to "Phoenix AI" ambiguously)
Email: hello@phoenixai.solutions
LinkedIn: linkedin.com/company/phoenix-ai-solutions

Platform Presence:

Published thought leadership on LinkedIn
Engaged in industry discussions about GEO and AI implementation

Why It Worked:

Authority signals overlap with Google's E-E-A-T framework but GEO weights them more heavily. AI engines are designed to avoid hallucination, so they prioritize sources with strong credibility signals:

Is this a real company? (Organization schema)
Is the author an expert? (Author bio, credentials, LinkedIn)
Are claims supported? (External citations, specific data with sources)

These signals gave AI systems confidence to cite Phoenix AI without risking hallucination or misinformation.

6. Conversational Query Optimization

What We Did:

Targeted natural language questions (6+ words) rather than short keywords:

Instead of "AI consulting":

"How should a mid-market professional services firm implement AI without disrupting client delivery?"
"What's the difference between AI consulting and hiring an in-house AI team?"
"How much does AI consulting cost for a UK mid-market business in 2026?"

Instead of "GEO":

"What is generative engine optimization and why does it matter in 2026?"
"How long does it take to get cited by ChatGPT and Perplexity?"
"Should I hire a GEO consultant or build GEO capabilities in-house?"

Content Structure for Long-Tail Queries:

Created comprehensive 2,000-3,000 word answers to complex questions
Used exact question phrasing as H2 headers where natural
Answered the full question including context and nuance (not just short answers)
Included comparison frameworks ("X vs Y: which is right for you?")

Question Mining Sources:

Google "People Also Ask" boxes for our target keywords
Sales call transcripts (questions prospects asked during discovery calls)
Customer support queries and FAQs
LinkedIn discussions in our network

Why It Worked:

Google AI Overviews appear 77% of the time for queries with 6+ words. Simple questions get answered without citations. Complex, nuanced questions require expert sources AI must cite.

By targeting conversational queries, we optimized for the questions where AI engines need to cite expertise—not questions they can answer from general knowledge.

Example:

Short query: "What is GEO?" — AI can answer generically without citing Phoenix AI
Long query: "How long does GEO implementation take for a mid-market B2B company and what results should I expect in the first 90 days?" — AI needs specific expertise and must cite sources

7. AI Citation Monitoring & Iteration

What We Implemented:

Systematic monthly tracking to measure what was working:

Target Query Testing:

Created list of 20 target queries across our expertise areas (GEO, AI consulting, Revenue Engine, AI implementation)
Tested monthly across ChatGPT, Perplexity, Claude, and Google AI Overviews
Documented: Did Phoenix AI get mentioned? Was it a citation or just a brand mention? What context was cited?

[DATA NEEDED: Specific tracking methodology - did we use tools like OtterlyAI/Profound or manual testing?]

Google Analytics 4 Setup:

Created custom segments for chatgpt.com and perplexity.ai referrers
[DATA NEEDED: What traffic volume did we see from these sources?]

Google Search Console Monitoring:

Filtered for 6+ word queries to track conversational query impressions
[DATA NEEDED: Did impressions increase for long-tail queries after GEO implementation?]

Iteration Based on Results:

[DATA NEEDED: Which tactics drove the most citations? Did we adjust strategy based on early results?]

Results After 60 Days (With Screenshots)

[DATA NEEDED: This entire section requires real data. Template structure below.]

ChatGPT Citations

Target Queries Where Phoenix AI Now Gets Cited:

[Query 1] — [Screenshot needed] — [Context: how we're cited, frequency, accuracy]
[Query 2] — [Screenshot needed]
[Query 3] — [Screenshot needed]

Citation Frequency: [DATA NEEDED: How often do we appear when testing these queries?]

Citation Accuracy: [DATA NEEDED: Does ChatGPT cite us accurately? Any hallucinations or errors?]

Perplexity Citations

Target Queries Where Phoenix AI Now Gets Cited:

[Query 1] — [Screenshot needed] — [Context: how we're cited, with link to phoenixai.solutions?]
[Query 2] — [Screenshot needed]
[Query 3] — [Screenshot needed]

Citation Frequency: [DATA NEEDED: How often do we appear?]

Link Attribution: [DATA NEEDED: Does Perplexity link to our site? Which pages get linked?]

Claude Citations

Target Queries Where Phoenix AI Now Gets Cited:

[Query 1] — [Screenshot needed]
[Query 2] — [Screenshot needed]

Citation Frequency: [DATA NEEDED: How often?]

Google AI Overviews

Any Appearances? [DATA NEEDED: Do we appear in Google AI Overviews for any queries? Screenshots.]

Traffic Impact

AI-Referred Traffic (GA4 Data):

[DATA NEEDED: chatgpt.com referrals: X sessions, X% increase month-over-month]
[DATA NEEDED: perplexity.ai referrals: X sessions, X% increase month-over-month]
[DATA NEEDED: claude.ai referrals: X sessions]

Traffic Quality:

[DATA NEEDED: Conversion rate comparison — AI-referred traffic vs organic search vs direct]
[DATA NEEDED: Bounce rate comparison]
[DATA NEEDED: Average session duration comparison]

Hypothesis from industry data: Similarweb reports AI-referred traffic converts approximately 2x higher than organic search. [DATA NEEDED: Does Phoenix AI data confirm this?]

What Worked Best (Tactics Ranked by Impact)

Based on our 60-day results, here's what drove the most citations:

1. [DATA NEEDED: Which pillar had the biggest impact?]

Likely llms.txt + Schema Markup based on implementation, but need real data to confirm.

Hypothesis: The combination of llms.txt (providing structured metadata) + comprehensive schema markup (signaling credibility) gave AI engines everything they needed to confidently cite Phoenix AI.

Evidence: [DATA NEEDED: Do citations increase after llms.txt implementation? Do pages with schema markup get cited more than pages without?]

2. [DATA NEEDED: Second most effective tactic]

Likely quotable definitions + FAQ sections.

Hypothesis: Clear 40-60 word definitions and FAQ Q&A format provided extractable content AI could quote directly.

Evidence: [DATA NEEDED: Are our definitions being quoted verbatim? Do FAQ answers appear in AI responses?]

3. [DATA NEEDED: Third most effective tactic]

Likely original frameworks (The 7 Pillars) + proprietary data (pricing transparency).

Hypothesis: Original frameworks became the default citation when AI needed structured methodologies. Pricing transparency filled information gaps competitors left open.

Evidence: [DATA NEEDED: Do AI engines cite "The 7 Pillars" when explaining GEO? Do they cite our pricing ranges?]

4. [Remaining tactics ranked by impact]

What Didn't Work (Honest Failures)

GEO is still emerging. Not everything we tried worked. Here's what failed or underperformed:

1. [DATA NEEDED: What tactics failed or underperformed?]

Hypotheses to test:

Overly long llms.txt? — Our file is 16KB vs recommended 2KB. Did this hurt parsing speed or citation rates?
Schema markup overkill? — We implemented schema aggressively across 15+ pages. Was this helpful or did it create noise?
Too much jargon? — Did any of our content use industry jargon that made it less extractable for AI?

[DATA NEEDED: What did we try that didn't drive citations?]

2. [Second failure or lesson learned]

3. [Third failure or lesson learned]

Why Document Failures:

Most GEO case studies only show successes. That's marketing fluff. We're documenting failures because:

It helps us optimize our own strategy
It helps clients avoid the same mistakes
It builds credibility—real experiments have failures, not just wins

ROI Analysis: Traffic, Conversions, Brand Awareness

[DATA NEEDED: This section requires real financial and traffic data.]

Investment

Time Investment:

[DATA NEEDED: Hours spent on llms.txt creation, schema markup implementation, content creation, testing]

Tool Costs:

[DATA NEEDED: Did we use paid GEO tools like Profound, OtterlyAI, SE Ranking? Costs?]

Total Investment: [$X,XXX or hours if done in-house]

Returns

Traffic Value:

[DATA NEEDED: AI-referred traffic volume × estimated value per session]
[Traditional SEO benchmark: $X per organic session in our industry]

Conversion Value:

[DATA NEEDED: How many leads/sales calls came from AI-referred traffic?]
[DATA NEEDED: Did any convert to clients? If so, lifetime value?]

Brand Awareness Value:

[DATA NEEDED: Hard to quantify, but how many prospects mentioned "I saw you cited in ChatGPT" during sales calls?]
[This is often the biggest value of GEO—being cited builds trust before prospects even visit your site]

ROI Calculation

Total Return: [$X,XXX] Total Investment: [$X,XXX] ROI: [X%]

Compared to Traditional SEO:

[DATA NEEDED: How does GEO ROI compare to our traditional SEO ROI?]
[Hypothesis: GEO has lower traffic volume but higher conversion rate, so ROI per session is higher]

Lessons for Mid-Market Businesses

Here's what we learned that applies to other mid-market businesses implementing GEO:

1. Start with llms.txt — It's the Foundation

Why: llms.txt is the single highest-leverage GEO tactic. One file, 1-2 hours to create, provides structured metadata AI engines parse on every query.

How: Create phoenixai.solutions/llms.txt with:

About section (who you are, what you do, who you serve)
Quick facts (citation-ready definitions answering common questions)
Products/services with clear descriptions
Key statistics and benchmarks
Contact information

Avoid: Don't overthink it. Start with 2KB, add more as you learn what AI engines cite.

2. Schema Markup Is Non-Negotiable

Why: Schema signals credibility and makes content machine-readable. Pages without schema get passed over.

Priority Schema Types:

Article schema (all blog posts/guides)
FAQPage schema (all FAQ sections)
Organization schema (homepage, about page)

How: Use Google's Structured Data Markup Helper or schema generators. Validate with Google Rich Results Test.

3. Quotable Definitions > Long-Winded Explanations

Why: AI engines need extractable content they can quote directly. Rambling prose gets passed over.

Format:

Lead with 40-60 word definitions
Use numbered frameworks ("The 5 criteria...", "3 approaches...")
Include statistics with sources
Structure as Q&A where natural

Test: Can you pull a single paragraph from your content and have it make sense as a standalone answer? If not, rewrite.

4. Target Complex Questions, Not Simple Keywords

Why: AI Overviews appear 77% of the time for 6+ word queries. Simple questions get answered without citations.

Instead of "GEO":

"How long does GEO implementation take for a mid-market B2B company?"

Instead of "AI consulting":

"What's the difference between hiring an AI consulting firm and building an in-house AI team for a $10M revenue company?"

Question Mining: Use Google People Also Ask, sales call transcripts, customer support queries.

5. Original Frameworks > Aggregated Content

Why: AI engines favor primary sources. When you're the only source for a framework or dataset, you become the default citation.

Examples:

Phoenix AI's 7 Pillars of GEO
90-Day measurable-outcomes framework
Market-cost comparison tables

Avoid: Don't just aggregate other people's research. Create something citeable.

6. GEO + SEO Work Together

Why: Traditional search still drives the majority of traffic. GEO provides higher-quality leads but lower volume. You need both.

Strategy:

Maintain strong traditional SEO (keyword research, link building, technical SEO)
Layer GEO optimization on top (llms.txt, schema markup, quotable content)
Track both channels separately in GA4

Don't: Don't abandon SEO for GEO or vice versa. Hybrid strategy wins.

7. Results Take 60-90 Days — But Start Showing Early

Timeline Based on Our Experience:

Week 1-2: Implement llms.txt and schema markup (technical foundation)
Week 3-4: First brand mentions appear in AI responses (weak signals)
Week 5-8: Citations start appearing for niche queries (early validation)
Week 9-12: Consistent citations for core queries (main results)
Month 4-6: Becoming default citation for category questions (mature results)

Don't: Don't expect overnight results. GEO is medium-term strategy (60-90 days to material results).

8. Test Monthly, Iterate Quickly

Why: GEO is still emerging. AI engine algorithms change. What works today might not work in 3 months.

Testing Cadence:

Test 10-20 target queries monthly across ChatGPT, Perplexity, Claude
Track AI-referred traffic weekly in GA4
Update llms.txt quarterly with new priority content
Refresh content every 30-60 days to maintain freshness

Don't: Don't "set it and forget it." GEO requires ongoing monitoring and iteration.

Next 90 Days: What We're Testing Now

GEO is still emerging. Here's what Phoenix AI is testing in the next 90 days:

1. [DATA NEEDED: What experiments are planned?]

Hypothesis to test:

Does llms.txt file size matter? (Test trimming to 2KB vs current 16KB)
Do video transcripts get cited more than text-only content? (Publish YouTube videos with full transcripts)
Does author credibility (LinkedIn following, published research) correlate with citation rates?
[Other experiments planned]

2. [Second experiment]

3. [Third experiment]

Why Share This:

Most agencies treat their GEO methodology as proprietary. We're sharing ours because:

Phoenix AI's competitive advantage isn't our tactics—it's our execution speed and mid-market focus
Rising tide lifts all boats—the more businesses optimize for AI, the faster GEO becomes standard practice
Transparency builds trust—we want clients who value our honesty, not just our tactics

FAQ: GEO Implementation for Mid-Market Businesses

How long did it take Phoenix AI to get cited by ChatGPT?

[DATA NEEDED: Timeline from implementation start to first ChatGPT citation]

What was the most effective GEO tactic for Phoenix AI?

[DATA NEEDED: Which pillar drove the most citations—likely llms.txt + schema markup based on implementation, but need real data]

How much traffic does AI engine citations drive?

[DATA NEEDED: Traffic increase from chatgpt.com, perplexity.ai referrers in GA4]

Can GEO work for mid-market businesses or just big brands?

Phoenix AI's case study proves GEO works for mid-market businesses. We're a UK-registered, Caribbean-based AI consultancy founded in 2024 with no enterprise budget or massive content team. If GEO works for us at our scale, it works for any mid-market business with expertise worth citing.

Do you need expensive GEO tools to get results?

No. Phoenix AI used free tools for our initial GEO implementation: manual llms.txt creation, free schema markup generators, and manual testing in ChatGPT and Perplexity. Paid tools like Profound and OtterlyAI accelerate tracking but aren't required for core implementation.

How often should you update your GEO strategy?

Based on Phoenix AI's experience, content should be refreshed every 30-60 days to maintain citation rates. llms.txt should be updated quarterly with new priority content. Schema markup should be audited every 90 days. AI engines favor freshness more heavily than traditional search.

What's the ROI of GEO consulting for mid-market businesses?

[DATA NEEDED: Phoenix AI's investment vs traffic/conversion/brand awareness ROI—compare to typical SEO ROI]

Does GEO replace traditional SEO?

No. Phoenix AI maintained traditional SEO alongside GEO implementation. Traditional search still drives the majority of traffic, but AI citations provide higher-quality leads with 2x conversion rates. GEO and SEO work together, not against each other.

How Phoenix AI Can Help Your Business Implement GEO

We're not just teaching GEO—we're proving it works by implementing it ourselves. If you're a mid-market business ($1M-$100M revenue) looking to establish visibility in AI search engines before your competitors figure it out, Phoenix AI offers: