Why an AI assistant would cite you - or not
ChatGPT, Claude, Perplexity, and Google's AI Overviews don't crawl the web live for every answer. They rely on a mix of training data and retrieval over indexed, structured content. If your brand's website has no clear entity definition, no structured markup, and no citable factual claims, there is nothing for the model to retrieve - even if you rank well in traditional search.
What these systems are actually looking for
| Signal | What it does |
|---|---|
| Structured data (Organization, FAQPage, Article schema) | Tells the model unambiguously who you are and what you offer |
| Clear, quotable factual statements | Gives the model a citable sentence rather than marketing prose to paraphrase |
| Consistent entity naming across the web | Reinforces the brand-to-facts association across sources |
| Author and publisher attribution | Supports the E-E-A-T signals that also inform AI training/retrieval weighting |
| Freshness and update cadence | Retrieval-augmented systems favour recently verified content over stale pages |
Why most South African brands are invisible in these answers
SA-specific content is underrepresented in the training and retrieval corpora these models draw from, which means local brands start from a lower baseline than in traditional Google search. Combine that with marketing copy written for humans rather than structured for machines, and most SA business websites are functionally unreadable to an AI system trying to answer "who does X in South Africa."
The fix, in order
- Add Organization, Service, and FAQPage JSON-LD to every core page - not just the homepage.
- Rewrite hero and service copy to include direct, factual, quotable statements about what you do and where.
- Publish content that answers the specific questions your buyers ask an AI assistant, not just what they type into Google.
- Keep author/publisher attribution consistent and visible.
- Refresh cornerstone pages regularly - stale content is deprioritised by retrieval systems.
Frequently asked questions
Is this the same as traditional SEO?
Related but distinct. Traditional SEO optimises for ranking in a list of links. Generative Engine Optimisation (GEO) optimises for being the source a model paraphrases or cites directly in a conversational answer - which depends more on structure and factual clarity than on backlinks or keyword density.
How do I know if I'm currently being cited?
Ask the assistants directly - query ChatGPT, Claude, and Perplexity with the questions your buyers would ask, and see whether your brand appears. An AI readiness audit formalises this into a repeatable score rather than a one-off spot check.