The quick version: two flavors of modern search emerged and upset decades of assumptions. ChatGPT (and sibling model-platforms) turned conversational AI into a discovery surface you can interrogate; Perplexity and similar “answer engines” emphasise fast, cited answers with visible provenance. These newcomers didn’t arrive with a full blueprint for the web economy, but by giving users neat, immediate answers they changed user expectations overnight. The result: the old model (crawl → index → click) looks like one option among several, and for the first time in a long while we can honestly call Google a conventional search engine rather than the default definition of “search".

A few years back the idea that a chatbox could replace search was a spicy thought experiment. By late 2023–2024 it stopped being an experiment.

ChatGPT shipped web-aware features and Live Search modes that let users ask complex, iterative questions and get synthesized, conversational answers. That shift matters because the user’s mental model of “search” started to include dialogue, follow-ups, and multi-step instruction execution not merely a ranked list of links. OpenAI’s product moves reflect this transition: ChatGPT is effectively acting as an AI search surface for many users.

Perplexity took a different, citation-first route: rapid web grounding, compact answers, and product features (like Comet) that embed summarisation into browsing. That design trades some of the interactivity of a chat assistant for traceability, users see sources and can verify claims quickly. Perplexity’s growth and browser integrations showed an appetite for answer-first experiences that make research faster and more verifiable.

Faced with that, the incumbents reacted. Google layered generative overviews and integrated Gemini models to avoid being sidelined; Bing and other engines followed with their own AI features and grounding strategies. The consequence is a hybrid search ecosystem: classic index-and-rank still exists (and still drives enormous volume), but generative overlays and RAG (Retrieval-Augmented Generation) systems now mediate how answers are composed and whether your page becomes the cited source or merely a background document.

That hybrid reality creates both opportunity and risk. Market data still shows Google’s scale dwarfs the newcomers in raw query volume, but the fastest-growing traffic segment is “chat/AI prompt” usage, a user behaviour that can reduce click-throughs and redirect attention inside walled assistant experiences. In short: you cannot ignore conventional SEO, but you also can’t treat chat/answer engines as a fad. The two audiences behave differently and expect different signals.

Technically, the bridge that makes chat answers defensible is RAG systems that fetch relevant documents at runtime to ground generations. RAG is now mainstream research and practice: its presence means that a clean, well-indexed, and well-tagged document has a better chance of becoming a ground truth for an answer engine. Conversely, messy content and poor metadata make you invisible or, worse, a source of hallucinated misinformation.

What this means for Google (and whether “Google will fail”)

The dramatic headline “Google is next Yahoo/Nokia” is tempting, but simplistic. Google still controls massive distribution, ad systems, and product margins that fund continuous R&D. However the analogy is useful as a cautionary tale: incumbency doesn’t guarantee immunity. What matters is whether Google can evolve its product and business model fast enough to keep users and creators satisfied simultaneously. Google’s generative features (Gemini, Search AI modes) are an explicit bet to do just that, retain users while adding summarisation and grounding. The outcome depends on product execution and on evolving arrangements around web access, licensing, and creator economics.

Practical impact for businesses, what to change, immediately

First: treat content as data. Canonical URLs, structured data (schema), content provenance (authors, timestamps), and accessible sitemaps are now fungible assets for RAG engines and answer services.

Second: measure differently, track not just clicks but appearances in snippets, answer citations, and referral shifts from assistant surfaces.

Third: decide your goals: do you want clicks, citations, or both? Perplexity-style answers may cite you (helping authority) but reduce CTR; ChatGPT workflows may keep users inside apps unless you’re integrated via plugins/APIs.

Fourth: diversify discoverability, maintain classic SEO health while exploring placements or partnerships with AI platforms and browsers that surface answers.

Finally: operationally, govern your content and internal knowledge stores so any RAG system (internal or third-party) uses high-quality sources. These are engineering and governance problems as much as marketing problems.

We’re in the midst of a structural shift where search is plural: conversational assistants, citation-first answer engines, and traditional indexers now coexist.

For Indonesian businesses (and anyone serious about online value), the sensible play is humble and technical: get your content and infra ready for both worlds. Keep optimizing for Core Web Vitals and classic SEO, yes, but also be rigorous about metadata, provenance, and document hygiene so the new answer systems can find, cite, and use your work. You can’t bet solely on one audience; build for both.

Want a short, pragmatic blueprint for capturing both conventional search traffic and answer-engine citations? Book a 30-minute briefing with Binari Suite sales@binari.co.id