We invited 40 CMOs and senior marketers to our offices for a working session on Generative Engine Optimisation — the discipline of being the brand AI models cite, recommend and trust. This is the event recap, the framework we presented, and the data behind it.
The opening question
That was the opening provocation. It landed because most of the 40 marketers in the room couldn't remember. We stopped searching. We started asking. The behaviour that built every modern marketing system has fundamentally changed — and most marketing systems have not changed with it.
The session covered three acts. First, the shift: what changed about how people discover brands, and why traditional SEO instincts now under-deliver. Second, the new rules: how AI models actually decide who to cite, what Princeton research on 10,000 queries revealed about citation patterns, and what Algorithm has learned engineering this for clients since 2024. Third, the playbook: a 90-day framework for moving from fragmented channels to a compounding visibility system across Search, AI and Social Engines.
The data behind the shift
69%
of Google searches end without a click
SparkToro / Datos 2025
1.0B
weekly ChatGPT users
OpenAI, March 2026
14.2%
AI referral conversion rate
vs Google organic at 2.8%
61%
CTR drop when AI Overviews appear
Industry research
The framework — Algorithm's position
We don't think about SEO, GEO and social search as three separate channels competing for budget. We think about them as three engines feeding one connected discovery system — and that's how we engineer for them.
Google, Bing
Rankings and authority feed AI sources
ChatGPT, Perplexity, Gemini, Claude
Recommendations drive awareness and consideration
TikTok, YouTube, Instagram
Social signals and citations feed authority
Each engine reinforces the others. Together, they compound. The problem with most marketing systems is not that they aren't doing SEO, or social, or even AI — it's that those efforts are disconnected. Channels, not a system.
How AI decides
01
Does the AI know who you are? Is your brand a recognised entity with structured data, knowledge graph presence and consistent identity across the web? If the model cannot identify you as a distinct entity, you cannot be recommended.
02
Does the web confirm your expertise? AI looks for corroboration — consistent, specific, trustworthy content that positions you as the authority. Multiple independent sources need to agree on your positioning.
03
Do trusted sources reference you? The more high-quality sites that mention and link to your brand, the more confident AI becomes in recommending you. This is not just backlinks. It is mentions, reviews and structured references.
Backed by independent research
Princeton University research published at KDD 2024 tested 10,000 queries to understand what drives AI citation. They found GEO optimisation produced a 40% visibility boost on average, and a 115% uplift for lower-ranked sites— meaning sites ranked #5 in traditional SEO benefited more than sites ranked #1 once AI citation was engineered. That's the inversion that makes this category winnable for challenger brands.
What attendees walked away with
Open ChatGPT, Perplexity and Gemini. Ask them to recommend a company in your category. See where you rank. Screenshot the results.
On Perplexity, click 'Sources' to see what the AI cited. Are any of those pages yours? Are they your competitors'? This is your citation gap.
Search your brand name plus your category. Does Google show an AI Overview? Are you cited in it? This tells you where you stand in the AI layer.
Search your brand on Google and check the Knowledge Panel. Is it accurate? Complete? If AI cannot find a clear entity for your brand, it cannot recommend you.
Run your homepage through Google's Rich Results Test. If your site has no schema markup, AI engines are working harder to understand who you are.
Questions to take back to your team
We closed the session with these. Most attendees told us they couldn't confidently answer half of them about their own brand. That's the starting line.
What does ChatGPT say about us when asked to recommend a brand in our category?
Do we know which sources AI is citing when it recommends our competitors?
Is our brand a recognised entity in Google's Knowledge Graph?
Are we measuring visibility across AI engines, or only tracking Google rankings?
Is our content structured for AI consumption, or only for human readers?
How does our social presence feed back into our search and AI visibility?
Do we have a single, consistent brand narrative across all digital platforms?
What would it take to become the default AI recommendation in our category?
The South African opportunity
One of the strongest reactions in the room came when we ran live queries against ChatGPT, Perplexity and Gemini using categories the attendees competed in. Brands represented in the room were being mentioned — but none of them were doing it intentionally. AI was making accidental recommendations based on whatever signals it happened to have absorbed.
That's the opportunity. The same AI models — ChatGPT, Perplexity, Gemini — serve South African users with the same mechanics they use globally. GEO strategies that work internationally apply directly to the South African market, but against far less competition. The brands that start building entity authority and citation density today will be exponentially harder to overtake in twelve months.
AI recommendation is a compounding game. Once a model has been trained on consistent, authoritative signals about your brand in your category, displacing those signals is hard work for competitors. The first movers don't just win share — they win a defensible visibility moat.
Want to run this for your team?
We run this format for in-house marketing teams, leadership offsites and industry events. 60 to 90 minutes covering the recommendation economy, the three-engine framework, Princeton's findings on what drives AI citation, and a live audit of where your brand sits across ChatGPT, Perplexity and Gemini.
Get in touch