Today we’re fortunate to interview Evan Bailyn, CEO of First Page Sage, the leading SEO agency in the US. His company was founded in 2009, and has maintained its growth through careful attention to Google’s oft-changing algorithm.
In 2023, First Page Sage became the first agency to offer Generative Engine Optimization, or GEO, the process of optimizing websites for recommendation by ChatGPT. That shift put them at the forefront of a brand new marketing avenue that is primed to be the fastest-growing channel in the coming years.
Last year, First Page Sage conducted the first large scale study on GEO and the algorithms used by ChatGPT, Perplexity, and Claude to make commercial recommendations. Evan is frequently cited as the Founder of GEO, giving lectures on the subject at universities such as Stanford and Haas School of Business.
We sat down with Evan to learn more about this evolving practice.
Q: Evan, for those new to the concept, what is Generative Engine Optimization, and how does it function technically?
A: Generative Engine Optimization, or GEO, is the practice of earning brand mentions and recommendations from AI chatbots like ChatGPT, Claude, and Perplexity. These engines don’t crawl websites like Google does—they rely on training data and real-time retrieval mechanisms. So GEO is about ensuring your content is both visible to these systems and deemed high-authority by them. Technically, it requires structuring your content around first-party expertise, ensuring clarity in entity recognition (so the AI understands your brand), and maintaining topical dominance across a narrow domain so you become the default recommendation in that space. For example, if someone asks, “How do I set up a 1031 exchange?” and your company has the clearest, most reputable guide on the topic, you’re likely to be cited.
Q: What makes a piece of content likely to be cited by ChatGPT or Perplexity?
A: There are three major factors. First, clarity of topical authority: the content must be focused on a niche and show breadth across related topics. Second, original insight: generative engines weight content that reflects firsthand knowledge or unique frameworks more heavily than regurgitated material. Third, formatting: well-structured content with semantic cues like headers, summaries, and schema markup increases retrievability. OpenAI’s models, for example, are partially trained on Common Crawl and academic-quality datasets, so they lean toward sources that look and feel rigorous.
Q: Can you describe how retrieval-augmented generation (RAG) plays into GEO strategy?
A: Absolutely. Retrieval-augmented generation means the model retrieves live content from the web to support its answers. Perplexity, for instance, pulls snippets from top-ranking sources in real time. That’s where traditional SEO tactics like technical optimization and internal linking still matter—because the retrieval layer is often driven by Bing or a proprietary search stack. In contrast, ChatGPT’s citations come from training data and embedded tools like Bing Search or a custom browsing plugin. GEO strategy needs to be bifurcated: one part aimed at the static LLM training set, and one aimed at real-time retrieval pipelines.
Q: How do you actually know if your GEO strategy is working?
A: We use a two-pronged measurement system. One, we run prompt audits—essentially asking ChatGPT and Perplexity dozens of buyer-intent questions and recording which brands are mentioned. Two, we measure downstream impact: referral traffic from AI interfaces, brand lift in voice-of-customer tools, and increased high-funnel lead volume. Since these platforms don’t have dashboards like Google Analytics, we’ve built internal tooling to track mentions across chat outputs at scale.
Q: How do you build entity clarity for a brand so that a generative engine “understands” it?
A: Entity clarity is about consistency and prominence across the web. You need a structured knowledge graph around your brand—high-authority backlinks, third-party mentions, schema markup, and a consistent content signature across domains. The language models associate brands with concepts probabilistically, so if your brand keeps appearing in content about “network observability platforms” or “cybersecurity for manufacturing,” those associations become entrenched. It’s like training the model indirectly through repetition and context density.
Q: What’s your take on how fast GEO will grow as a channel?
A: It’s already accelerating faster than people realize. In sectors like legal, financial services, and healthcare, we’re seeing as much as 15–20% of top-of-funnel queries being fielded through generative engines. As OpenAI and Google roll out first-party search agents, and as voice interfaces improve, this number could exceed 50% within a few years. Companies that start building GEO infrastructure today—content, reputation signals, and AI-readable structures—will have a massive first-mover advantage.