The way humans find information on the internet changed permanently in 2025. Google's traditional ten-blue-links model — the dominant information retrieval paradigm for 25 years — is being displaced by AI-generated answers. In March 2026, AI Overviews appear in 25% of all Google searches. Google's new AI Mode, which gives full conversational AI answers for complex queries, is rolling out publicly across the US. ChatGPT Search crossed 1 billion monthly search queries. Perplexity, which is purpose-built as a cited AI search engine, is growing at 400% year-over-year. Three distinct approaches to AI search are competing in the market — and they are genuinely different tools that produce different results for different query types. Knowing which to use for which question is a meaningful productivity decision in 2026.

Updated September 2026: What's Changed Since March

Six months on, none of the fundamentals from March have reversed — all three tools kept growing, and the models sitting underneath them changed generation more than once. ChatGPT is now running on OpenAI's newest GPT-6 family, which arrived in stages and reached its lightweight and mid-tier variants, Luna and Sol, in September 2026. Google's Gemini models went through several point releases in the same window, and Anthropic — whose Claude powers LumiChats — released Claude Opus 5 earlier in the year and followed it with a faster, cheaper Opus 5.5 update on September 22. None of that changes which tool to reach for first, covered below, but it does mean the raw answer quality across all three has kept improving rather than plateauing.

Google AI Mode itself moved well past the 'rolling out' stage this article described in March. Over the following months, Google kept extending AI Mode from its initial US rollout to more countries and languages, and for a growing share of queries it now behaves less like an optional mode and more like the default way Google Search answers a question. That expansion sharpened a trade-off this guide already flagged: the more Google's AI answers keep users inside Google's own results, the less traffic reaches the websites those answers are built from — a tension publishers and regulators kept pushing back on through 2026, not something that resolved itself since March.

The biggest structural change in AI search since March isn't which tool answers questions best — it's that answering questions is no longer the whole job. Perplexity pushed further into 'agentic' search with its Comet browser, free to download on desktop and mobile, letting Perplexity act on your behalf — booking, filling forms, completing purchases — rather than only citing sources. That shift also introduced AI search's newest real risk: security researchers spent much of 2026 documenting prompt-injection attacks against AI browsers, including Comet, where a malicious webpage or calendar invite can trick an agent into acting on hidden instructions instead of yours — several such vulnerabilities were publicly disclosed and patched over the course of the year. If you use any AI search tool's agent or 'browse for me' mode, treat it the way you'd treat a new employee with your credit card — useful, but not yet something to leave unsupervised around payments or saved logins.

None of this changes the core recommendation from March: default to Google AI Mode for news, local, and shopping; Perplexity for research where you need to verify every source; and ChatGPT Search for open-ended, multi-turn exploration. What's changed is the ceiling on all three tools, and the stakes of the newer agentic features layered on top of them — both worth knowing before you turn a 'do this for me' feature loose on your accounts.

How Each Tool Works — The Core Architectural Differences

Google AI Mode

Google AI Mode is Google Search rebuilt around generative AI responses. Instead of returning a list of links, AI Mode synthesizes an answer to your query from multiple web sources, with inline citations, and allows you to ask follow-up questions in a conversational interface. The critical Google advantage: Google has spent 25 years building the most comprehensive web index in history. AI Mode is built on top of that index — it has access to more sources, more current information, and deeper contextual understanding of web authority than any competitor.

• Strengths: Best index coverage, most authoritative source assessment, real-time news integration, Shopping and Local knowledge deeply integrated, Google's Knowledge Graph for factual accuracy on entities.
• Weaknesses: AI Mode is newer and less refined than Perplexity's citation experience. The transition between traditional Google results and AI Mode can be inconsistent. Privacy concern: Google uses AI Mode interactions to improve its models and ad targeting.
• Best for: Queries that benefit from Google's local and shopping knowledge, current news, anything requiring the depth of Google's index, searches that mix informational and transactional intent.

ChatGPT Search

ChatGPT Search combines OpenAI's language model with web search capabilities, allowing ChatGPT to retrieve current information rather than relying solely on training data. The distinctive ChatGPT Search experience: the conversational interface allows multi-turn search refinement that feels like researching with a knowledgeable colleague rather than typing queries into a search box. You can say 'actually, I meant in the Indian context specifically' and the search adapts without starting over.

• Strengths: The most natural conversational search experience. Multi-turn refinement without starting over. ChatGPT's reasoning quality applied to search results produces more nuanced synthesis than competing tools. Strong for research questions that require understanding and judgment, not just information retrieval.
• Weaknesses: Source citation is less consistent than Perplexity. Index coverage is smaller than Google. More expensive to access (requires ChatGPT Plus or higher for full functionality).
• Best for: Research questions requiring interpretation and judgment, queries where you expect to follow up multiple times, complex topic exploration where you are still learning what you need to know.

Perplexity AI

Perplexity was purpose-built as a cited AI search engine — not a chatbot that happens to search the web. Every answer includes numbered inline citations linked to primary sources. The entire product experience is designed around source transparency. Perplexity's Academic Focus mode searches only peer-reviewed databases. Its Pro tier lets you choose which AI model answers your query (Claude Sonnet 4.6, GPT-5.4, Gemini 3 Pro) while maintaining the citation-first experience.

• Strengths: Best citation experience of any AI search tool — citations are inline, numbered, and linked directly to primary sources. Academic Focus mode for research paper search. Spaces feature for persistent research projects. Multi-model choice on Pro tier.
• Weaknesses: Does not match Google's breadth of index for local, shopping, or long-tail queries. Free tier has daily query limits. PerplexityBot has been involved in web scraping controversy that has caused some publishers to block it.
• Best for: Professional and academic research, any query where verifying sources matters, current events with specific source requirements, competitive intelligence and market research.

Also on LumiChats

AI Guide

Perplexity vs ChatGPT vs Google for Indian Students in 2026

8 min read→

AI Guide

How to Show Up in ChatGPT and AI Search: The GEO Guide

13 min read→

AI Guide

Perplexity AI Is Quietly Replacing Google for Millions

10 min read→

Direct Comparison on 6 Common Query Types

The Optimal Search Stack for 2026

The most efficient search strategy is not choosing one tool and using it for everything — it is knowing which tool to default to for which query type. Based on testing all three tools extensively in March 2026, the recommended default assignments are:

• Default to Google AI Mode for: News, local, shopping, anything requiring Google's breadth of index, queries mixing informational and transactional intent.
• Default to Perplexity for: Research, academic work, professional fact-checking, competitive intelligence, any query where you need to verify the source of every claim.
• Default to ChatGPT Search for: Complex topic exploration, multi-turn research refinement, queries where you expect to build on the answer with follow-up questions.
• The switch trigger: If your default tool gives a poor answer, switch immediately. Do not iterate endlessly on one tool — each tool has different blind spots, and a fresh query on a different tool often produces a significantly better result in less time than trying to refine a bad initial answer.

Pricing Comparison