Perplexity AI is a San Francisco company founded in 2022 that runs an AI-powered search engine. Instead of a list of links, it returns a written answer that summarises current web content and cites the sources used. Alongside a free version there are paid tiers with a choice of language models, plus the Comet AI browser (as of October 2026). For businesses, Perplexity is also relevant because it changes how content on the web is discovered.
AI search refers to search methods that use AI to grasp the intent behind a query, weight results and often generate a summarised answer with sources rather than a plain list of links. Technically, they use keyword search, vector search over embeddings or a combination, frequently followed by a language model. Examples include answer engines on the web and internal knowledge search that helps staff find what they need in manuals and policies.
Generative engine optimisation (GEO) covers measures that help content become visible in, and be cited as a source by, AI-powered search systems such as ChatGPT, Perplexity or the AI overviews in search engines. The term comes from a 2023 research paper. Key factors include clear, directly quotable answers, a clean structure, evidenced facts, up-to-date information and identifiable authors. GEO complements classic search engine optimisation (SEO) rather than replacing it.
A zero-click search occurs when users get their answer directly on the results page, for example through featured answer boxes or AI-generated summaries, and no longer click through to a website. For businesses this reduces traffic even when their content serves as the source. Being visible as a cited source therefore becomes more important: with clear answers, supported facts and well-structured pages.
Retrieval-augmented generation combines search with text generation: before answering, the system searches a knowledge collection – such as manuals, policies or FAQs – for relevant passages and passes them to the language model as context. The answer is therefore based on current, company-specific content rather than only on the model's training knowledge. This reduces invented statements, makes sources traceable and allows updates without retraining.
