AI Search Engines

AI search engines are information-retrieval systems that use large language models to compose a single natural-language answer - with inline citations - rather than presenting a ranked list of blue links. They are also called answer engines.^[1]

History

Conventional search engines crawl and index the web, then rank documents against a query. The AI search model adds a synthesis step: the system retrieves candidate passages, and an LLM writes a response grounded in them. Microsoft integrated GPT into Bing in February 2023, Perplexity AI launched its citation-first answer engine in the same period, and Google responded by rolling out AI Overviews - generated summaries above traditional results - from 2024, followed by AI Mode, a dedicated conversational search tab.^[2]

By July 2025 AI Overviews served more than 2 billion monthly users, and Google reported that AI Mode passed 1 billion monthly users in May 2026. ChatGPT's built-in search grew with the assistant: OpenAI reported more than 1.2 billion weekly users in October 2026.^[1]

Key Concepts

Three technical elements define the category:

• Retrieval-augmented generation (RAG). The engine issues searches against its own index or the open web, then conditions the model on the retrieved text, so answers reflect current information rather than training data alone.^[3]
• Citation grounding. Answers carry inline references to source documents, which lets users verify claims - the design decision that distinguishes answer engines from open-ended chatbots.
• Query fan-out and intent rewriting. A single user question is expanded into several sub-queries, run in parallel and merged, which changes how keywords and topical authority are evaluated.

The economics shifted with it. One study found users clicked a traditional result on 8% of visits where an AI summary appeared, against 15% where none appeared; Ahrefs measured the top result's click-through rate 58% lower on keywords triggering an AI Overview.^[1] Publishers therefore treat generative engine optimisation - structuring content so it is quotable by models - as a complement to classical SEO.

Scale remains uneven. In September 2026 ChatGPT accounted for 80.47% of page views referred by AI chatbots worldwide, ahead of Gemini (10.94%) and Perplexity (5.84%), while Google still referred 89.94% of all search-engine referrals - AI assistants remain a small share of total search traffic.^[1]

Applications

Use cases include consumer fact-checking, travel and product research, academic literature scanning, competitive analysis and news summarisation. Perplexity positioned itself around research workflows with numbered sources; ChatGPT Search embeds retrieval inside a broader assistant that can then write, code or analyse what it found; Google's implementation keeps the broadest index and the deepest local and shopping coverage.^[3]

>See Also

• Retrieval-Augmented Generation
• Perplexity AI
• Generative Engine Optimisation
• ChatGPT
• Google DeepMind