How Fish.audio Achieved 97% AI Citation Rate: A Case Study in GEO
Eight months ago, Fish.audio asked a simple question: "Does ChatGPT even know we exist?"
The answer was uncomfortable. Their brand appeared in only 8% of relevant AI-generated answers, ranking #12 among voice cloning tools. Today, they rank #1 with a 97% citation rate. Their ChatGPT referral traffic is up 17,306% year-over-year. They generated 448,000 Google impressions in a single quarter.
This is not a fluke. It is the result of systematic Generative Engine Optimization (GEO) — and it is a preview of what every brand will need to do to survive the AI search shift.
The Problem: Traditional SEO Does Not Work on AI
Fish.audio had already done the SEO work. They had strong Google rankings, solid backlinks, and quality content. Their team understood digital marketing.
But AI search does not work like Google. When someone asks ChatGPT "best AI voice cloning tool," the model does not crawl the web in real-time and rank pages by links. It generates an answer based on what it learned during training — citing sources it has determined are authoritative, trustworthy, and relevant.
Having a #1 Google ranking means nothing if you are invisible to AI models.
Fish.audio discovered this the hard way. Their visibility in AI answers did not match their Google performance. They were being outperformed by brands with weaker SEO but better AI presence.
The GEO Framework: Four Quality Pillars
The solution came through a methodology called GEO (Generative Engine Optimization), developed by Topify.ai. Rather than optimizing for search algorithms, GEO optimizes for how AI models select and cite sources.
The framework rests on four quality pillars:
Experience — Does the content demonstrate first-hand knowledge? AI models favor sources that write from direct experience rather than second-hand summaries.
Expertise — Is the content authored by people with genuine domain knowledge? Technical depth matters more than keyword density.
Trust — Does the source have credibility in its field? Consistency and accuracy build trust signals that AI models can evaluate.
Authority — Is the brand recognized as a leader? Industry recognition, citations, and peer validation all contribute.
The insight here is that these are exactly the qualities that make content valuable to humans. GEO is not a hack — it is content quality optimization for the AI age.
The Prompt Categories Framework
GEO also involves understanding the four types of prompts AI models respond to:
Core discovery prompts — "What are the best AI voice cloning tools?" The goal here is to be included in the initial consideration set.
Precision lookup prompts — "How does Fish.audio compare to ElevenLabs for voice interpolation?" The goal is to be cited in comparative analysis.
Capability proof prompts — "Can AI clone a voice from a 10-second sample?" The goal is to be cited when the specific capability being asked about is relevant to your product.
Reputation validation prompts — "Is Fish.audio a legitimate company?" The goal is to appear when brands are being evaluated for trustworthiness.
Each prompt category requires different content optimization strategies. Fish.audio worked through each one systematically, creating content that addressed the specific questions AI models were being asked in each category.
What Actually Changed: The 12x Improvement
The results speak for themselves:
- Citation rate: 8% → 97% (12x improvement)
- Ranking among AI voice cloning tools: #12 → #1
- ChatGPT referral traffic: up 17,306% YoY
- Google impressions: 448,000 in 3 months
The 17,306% traffic increase is particularly striking. When AI models cite you, they are essentially giving you a recommendation to every user who asks a relevant question. That is compounding visibility that Google rankings cannot match.
Why This Matters for Every Brand
Fish.audio is not a special case. The same dynamics are playing out across every industry. Brands with weak AI presence are watching their competitors get recommended by AI while they are invisible.
The GPU cloud company case is instructive: they achieved 25.8% AI visibility improvement in just 20 days from zero. That is how fast the trajectory can shift when you apply GEO systematically.
The Japanese skincare brand went from 10% to 70% visibility in one month. An education consulting firm maintains 70-80% visibility. An e-signature SaaS averages 81.76%.
These are not outliers. They are what happens when brands stop optimizing for Google and start optimizing for AI.
What GEO Requires
GEO is not a magic button. It requires:
Understanding where you stand — You cannot improve what you do not measure. Brands need to know their current AI citation rate across multiple models (ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Google AI Overviews).
Multi-language capability — AI models are trained on content in multiple languages. Brands with global presence need to track visibility across English, Chinese, Japanese, Arabic, and other markets.
Content that demonstrates experience — Generic marketing copy will not cut it. AI models can detect thin content and will not cite it.
Continuous monitoring — AI models update their training data and citation patterns constantly. A one-time effort will not maintain visibility.
Multi-model coverage — Different AI models cite different sources. Optimizing for ChatGPT alone is not enough.
The Strategic Reality
GEO is not the future. It is the present. The question is not whether to do it — it is whether you do it now while you can establish presence, or later when you are trying to catch up to brands that already have AI visibility.
Fish.audio spent months building their AI presence. Their head start is now a structural advantage that their competitors will need significant resources to overcome.
The window for establishing AI presence is closing. Every day that passes, more brands compete for the same citations. The brands that move first will have compounding advantages that are very difficult to replicate.
Tools like Topify.ai are making GEO accessible to brands of all sizes. The technology to track and improve AI visibility is no longer reserved for enterprises with large marketing budgets. The playing field is leveling — but only for brands that are paying attention.
The question every brand needs to ask is simple: "Does AI mention us when it answers relevant questions?"
If the answer is no, the work begins now.
Top comments (0)