Millions of people now get answers from ChatGPT, Perplexity, and Gemini without ever visiting a website. If your brand isn’t being cited, you’re invisible to a fast-growing slice of your market.
By Matthis Duarte — Senior SEO professional, 12 years experience
Something significant happened to search behaviour between 2023 and 2025. A meaningful portion of informational queries — the kind that used to flow into Google — started going somewhere else entirely. ChatGPT. Perplexity. Gemini in its AI-native mode. Microsoft Copilot.
These are not search engines in the traditional sense. They don’t return a ranked list of links. They generate a synthesised answer — fluent, often authoritative-sounding — and cite a handful of sources beneath it. Or sometimes no sources at all.
For businesses and brands, this creates a new problem: you can rank number one on Google and still be completely absent from the answers your customers are actually getting. GEO — Generative Engine Optimization — is the discipline of solving that problem.
What GEO actually is — and what it isn’t
Generative Engine Optimization is the practice of building the signals that cause large language models (LLMs) and AI-powered search engines to cite, reference, and recommend your brand when generating responses relevant to your category.
The term gained formal academic grounding in a 2023 research paper from Princeton University, which studied how website content could be optimised to perform better in generative AI responses — coining GEO as a distinct discipline separate from traditional SEO. Andreessen Horowitz (a16z) made the business case even more explicitly in their “GEO over SEO” thesis, arguing that AI-generated answers represent a structural shift in how brands earn visibility online. [verify]
GEO is not:
- Writing content that “sounds like” AI will like it
- Keyword optimisation for a different algorithm
- A replacement for SEO
GEO is:
- Building entity authority — making your brand a clearly defined, well-documented entity that AI systems can confidently reference
- Establishing citation signals — creating a web of authoritative third-party mentions that LLMs draw from when generating responses
- Owning specific topic associations — becoming the recognised source on a defined set of topics so that AI systems default to citing you when those topics come up
“GEO isn’t about gaming a new algorithm. It’s about becoming genuinely authoritative on the topics your customers are asking AI about.”
The distinction from AEO matters. Answer Engine Optimization focuses on structured content signals that help search engines select your page as a featured snippet or AI Overview citation — it operates within the search index. GEO operates at a fundamentally different layer: the signals LLMs use to decide which entities and sources are credible enough to cite when generating free-form responses.
How GEO metrics differ from SEO metrics
This is where most traditional marketers get confused. The KPIs for GEO are not rankings — they are reference rates.
In SEO, you track position. In GEO, you track how often your brand or content is mentioned, cited, or recommended inside model-generated answers. There are no steady rankings in GEO — the same query entered twice can produce different answers with different cited sources. The goal is not to hold position one. The goal is to increase the frequency and accuracy with which AI systems reference your brand.
| Metric | SEO | GEO |
|---|---|---|
| Primary KPI | Keyword ranking position | Reference rate (frequency of AI citations) |
| Measurement | Rank trackers (Ahrefs, Semrush) | AI monitoring tools (Profound, Brandwatch) |
| Stability | Relatively stable rankings | Non-deterministic — same query, different outputs |
| Success signal | Top 3 position | Brand cited accurately in relevant AI responses |
| Traffic type | Click-through traffic | Brand awareness + direct traffic from AI citations |
How LLMs decide what to cite
To optimise for generative engines, you need to understand how they select sources — and it’s fundamentally different from how Google ranks pages.
Training data presence. LLMs like GPT-4 are trained on vast datasets scraped from the web. Brands and sources that appeared frequently in high-quality, authoritative contexts during training have a higher baseline presence in the model’s knowledge. This shapes the long-term importance of consistent brand building across the web.
Real-time web access. Engines like Perplexity and Gemini supplement their training with live web searches. For these engines, your current search visibility, content freshness, and site authority directly affect citation likelihood.
Entity recognition. LLMs reason about the world through named entities — companies, people, products, concepts — and the relationships between them. A brand clearly defined as an entity across multiple authoritative sources is far more likely to be referenced confidently than one that exists only on its own website.
Citation authority. AI systems tend to favour sources that are already widely cited elsewhere — creating a compounding dynamic where brands with strong editorial presence accumulate the citation authority that makes them more likely to appear in AI-generated answers.
| Signal type | What it means in practice |
|---|---|
| Entity clarity | Brand, category, and key attributes consistently described across multiple authoritative sources |
| Third-party mentions | Authoritative publications reference your brand in relevant topical contexts |
| Topical association | Content consistently covers a defined topic cluster with depth and authority |
| Structured data | Organisation, Article, and Author schema help AI crawlers understand what your brand is |
| Recency | Fresh content signals to real-time engines (Perplexity, Gemini) that your information is current |
Why topical association is the core GEO lever
The single most actionable GEO principle: AI systems cite the source they associate most strongly with a topic, not necessarily the source with the best individual page.
This differs from how Google’s featured snippets work, where a single well-optimised page can win a snippet even from a low-authority domain. In generative responses, the model draws on its broader understanding of which sources are authoritative on a subject — built from patterns across many documents, not a single page signal.
In practice, GEO rewards the same strategy that topical authority SEO rewards: a comprehensive, consistent body of content around a defined topic cluster, so that AI systems form a strong association between your brand and that topic.
🔴 Case study — Investopedia: accidental GEO dominance
Investopedia was not optimising for generative engines when it built its content library — the concept didn’t exist yet. But its strategy of publishing comprehensive, authoritative definitions of every financial concept imaginable created exactly the kind of topical coverage that LLMs favour.
When ChatGPT and Perplexity users ask about financial concepts — P/E ratios, compound interest, hedge funds — Investopedia is cited disproportionately. Not because it optimised for AI engines, but because it built the broadest, most consistently authoritative topical presence on financial literacy.
→ Result: Investopedia became one of the most-cited domains in AI-generated financial responses — without any specific GEO strategy — because topical authority at scale is the foundation of AI visibility.
The lesson: the path to GEO is the same as the path to topical authority. Define your topic cluster. Build comprehensive, accurate coverage. Establish consistent entity signals. The AI citations follow.
The GEO action framework
Build entity presence first. Ensure your brand is clearly defined as an entity across Wikipedia (if eligible), Wikidata, LinkedIn, Crunchbase, and relevant industry directories. Consistency of name, description, and category strengthens entity recognition in LLMs.
Create the definitive content on your topic. Identify the 5–10 questions your customers are most likely to ask AI about your category. Build the most comprehensive, accurate, and clearly written content on each. Content that works best is self-contained, factual, structured with clear headings, and easy for AI to extract individual passages from.
Earn authoritative third-party mentions. Guest articles in industry publications, press coverage, podcast appearances, and analyst mentions create the citation signals that LLMs weight. A mention in TechCrunch carries more GEO value than a hundred mentions on low-authority blogs.
Use structured data for entity clarity. Organisation schema, Article schema with author markup, and FAQPage schema help AI crawlers understand what your brand is, what it does, and who stands behind it.
Measure your reference rate regularly. Query the AI engines your customers use with the questions they’re likely to ask. Track whether your brand appears, how it’s described, and which competitors are being cited instead. Tools like Profound are building AI citation monitoring specifically for this purpose. [verify]
Key takeaways
- ✓ GEO is the practice of building signals that cause AI engines like ChatGPT, Perplexity, and Gemini to cite your brand — formalised as a discipline in a 2023 Princeton research paper and validated commercially by a16z
- ✓ GEO metrics are reference rates, not rankings — there are no stable positions in AI-generated answers; the same query can produce different cited sources minutes apart
- ✓ GEO differs from AEO: AEO optimises for answer selection within search engines; GEO optimises for citation in LLM-generated responses based on entity authority and topical association
- ✓ LLMs select sources based on entity clarity, third-party citation authority, topical association, and (for real-time engines like Perplexity) current search visibility
- ✓ The single most powerful GEO lever is topical authority — AI systems cite the source most strongly associated with a topic, not just the one with the best individual page
- ✓ Measure GEO by querying AI engines directly with customer questions weekly — track citations, brand framing, and which competitors appear instead
Matthis Duarte is a senior SEO professional with 12 years of experience. HackingStory.com reverse-engineers how the fastest-growing startups actually grew — with real data, not press releases.
