AI Search: How ChatGPT, Gemini and Perplexity Are Changing Discovery

AI search is search where a language model does the reading for you: it runs the queries, pulls pages from the web and writes one answer with citations. In 2026 it runs inside Google, ChatGPT, Gemini, Perplexity and Copilot — and it decides which businesses get named before anyone clicks a link.

This is the anchor guide for the search and GEO series on this site. It covers what AI search is, how big it has become, how each engine finds its sources, what it does to clicks and customer journeys — including in Dubai and in Arabic — and where it's heading. The hands-on detail lives in two companion guides: SEO vs AEO vs GEO for the definitions, and generative engine optimization for the step-by-step playbook.

Key takeaways

• AI search is mainstream. Google's AI Overviews have more than 2.5 billion monthly users and AI Mode more than 1 billion. ChatGPT has passed 1 billion weekly users, and its search usage nearly tripled in a year.
• Search isn't shrinking — it's splitting. Google says AI Mode is adding to total queries, while discovery spreads across ChatGPT, Gemini, Perplexity and Copilot.
• Every engine searches before it writes. Your question becomes several searches, pages are pulled from an index, and a model quotes passages. If you can't be retrieved, you can't be cited.
• Fewer clicks, earlier shortlists. In Pew's study, people clicked a result on 8% of visits with an AI summary versus 15% without. Being named in the answer is the new page one.
• The UAE audience is already there. 73.3% of the UAE's working-age population used generative AI in Q2 2026, the highest rate Microsoft measured — and Google's AI search works in Arabic.
• Answers are becoming a paid channel. ChatGPT Ads passed a $1 billion annualised run rate and opened self-serve buying in the Middle East and North Africa in August 2026.

What is AI search?

AI search is a search experience in which a generative AI model answers your question directly — in sentences, tables or steps — using sources it retrieves from the web at the moment you ask, and shows which sources it used. Classic search hands you a ranked list of pages to read. AI search reads them for you and hands you the conclusion.

Definition

AI search — also called generative search, or an answer engine — combines a search index with a large language model. The index finds relevant pages; the model reads them and writes a cited answer. Google's versions are AI Overviews and AI Mode; OpenAI's is ChatGPT search.

It now shows up in four forms, and most people already use more than one:

1. Answers on top of results. Google's AI Overviews appear above the classic links on many searches.
2. Search as a conversation. Google's AI Mode, ChatGPT search, Perplexity and Microsoft Copilot let you ask, follow up and refine without starting again.
3. Assistants that search when they need to. ChatGPT and the Gemini app decide for themselves when a question needs fresh information from the web.
4. Agents that search, then act. At I/O 2026 Google introduced information agents in Search that work in the background "24/7, to find what you need at exactly the right moment, and help you take action". Search is becoming the first step of a task, not the whole task.

How big is AI search in 2026?

AI search now reaches billions of people. Google reports more than 2.5 billion monthly users of AI Overviews and more than 1 billion of AI Mode, and OpenAI says ChatGPT has more than 1 billion weekly users. These are the latest figures each company has published:

Product	Latest official figure	As of	Reported by
Google AI Overviews	2.5 billion+ monthly active users	19 May 2026	Google, I/O keynote
Google AI Mode	1 billion+ monthly active users	22 July 2026	Alphabet, Q2 2026 results
Gemini app	950 million monthly active users	22 July 2026	Alphabet, Q2 2026 results
ChatGPT	1 billion+ weekly active users	31 August 2026	OpenAI
ChatGPT search	Usage "nearly tripled in a year"	31 March 2026	OpenAI
Perplexity	200 million queries a day on its search infrastructure; an index tracking 200 billion+ URLs	25 September 2025	Perplexity Research

Monthly, weekly and daily figures aren't comparable, and each company decides what it counts, so read the table as scale rather than market share. The direction isn't in doubt: in about two years, AI-written answers went from an experiment to part of everyday search for billions of people.

Is traditional search dying?

No — the evidence points to search growing and splitting, not dying. In February 2024 Gartner predicted that traditional search engine volume would drop 25% by 2026 as people moved to AI chatbots and virtual agents. Google's reporting points the other way. In May 2025 it said AI Overviews were driving an increase of more than 10% in Google usage for the kinds of queries that show them, in big markets such as the US and India. In July 2026 it said AI Mode, "just like AI Overviews", is "driving an incremental increase in Search queries overall", alongside 17% revenue growth in Search and other.

What is shrinking is the click, and the shape of the query is changing. Google said at I/O 2026 that Search "has become less about individual queries and feels more like an ongoing conversation", and that "when people use our AI-powered features in Search, they use Search more".

Common misconception

"Nobody googles any more, so SEO is dead." Every AI engine in this guide still runs searches behind the scenes: Google's AI features draw on Google's index, ChatGPT search on third-party search providers and OpenAI's own crawler, Perplexity on its own index and Copilot on Bing. Search optimisation hasn't died — it has become the entry ticket to being cited. Google's own guidance says the SEO practices that work for Search also apply to its AI features, with no special optimisation required.

How does AI search work?

AI search works in four steps: it interprets your question, fans it out into several narrower searches, retrieves pages from a search index, then has a language model write an answer from the best passages and link the sources it used. Every major engine follows this pattern, even though their indexes differ.

How AI search answers a question

1 · Ask

One long, specific question

“Which Dubai communities suit a family with a AED 4M villa budget?” — plus location, memory and follow-ups

2 · Fan out

• family villa communities Dubai
• Dubai villa prices 2026
• British schools Dubai
• villa service charges

Many narrower searches, run at once

3 · Retrieve

• Google index
• Bing
• Own crawlers

Pages outside the index are invisible

4 · Answer

A written answer

• 1
• 2
• 3

Only a few sources get named

1. Interpret. The engine reads the whole question plus its context. Questions have grown: Google says early AI Mode users ask questions two or three times the length of traditional searches. OpenAI says ChatGPT may use your approximate location and, if memory is switched on, relevant saved memories when it rewrites a search — so someone who has mentioned they're vegan may get a search for vegan restaurants in their city.
2. Fan out. One question becomes many searches. Google says AI Mode uses a "query fan-out technique, breaking down your question into subtopics and issuing a multitude of queries simultaneously", and that its Deep Search mode can issue hundreds. OpenAI describes the same move: ChatGPT "typically rewrites your query into one or more targeted queries", then may send more specific ones after reading the first results.
3. Retrieve. Those searches run against an index. Which index depends on the engine — see the next section — and pages outside it are invisible to the answer.
4. Write and cite. A language model reads what came back, writes the answer and links its sources. Selection happens at the level of passages, not whole pages: Perplexity's research team says its system treats "the individual sections and spans of documents as first-class units".

What this means

Two things follow for any business. You need to be retrievable — crawlable and indexed wherever each engine looks — and quotable, with self-contained passages that state specific facts. Pages of general marketing copy lose on both counts. The practical checklist is in the generative engine optimization guide.

ChatGPT vs Google AI Mode vs Gemini vs Perplexity: how do they differ?

The major AI search engines differ mainly in where their answers come from and in how much they know about you. Google's products draw on Google's search index, ChatGPT on third-party search providers and its own crawler, Perplexity on its own index and Copilot on Bing — so being findable in one doesn't guarantee being findable in all of them.

Engine	Where answers come from	What it can use about you	Your lever
Google AI Overviews	Google Search's index, summarised by Gemini models	Your search	Classic SEO; Search Console's generative AI report
Google AI Mode	Google's index, via query fan-out	Your conversation and follow-ups	Same as AI Overviews — no special optimisation required
Gemini app	Gemini models, linking the web sources they quote	With Personal Intelligence on: connected apps such as Gmail and Google Photos	Clear, quotable pages it can link
ChatGPT search	Third-party search providers (OpenAI lists Microsoft and Shopify) and sites its OAI-SearchBot crawler can reach	Approximate location; saved memories, if enabled	Allow OAI-SearchBot; ChatGPT Ads
Perplexity	Its own index, crawled by PerplexityBot	Your conversation	Allow PerplexityBot; be the clearest source
Microsoft Copilot	Bing's index, via search queries Copilot generates	Your conversation	Bing Webmaster Tools and IndexNow

Two practical consequences follow. First, Bing matters again: Copilot grounds its answers in it, and Microsoft is one of the search providers OpenAI lists for ChatGPT. Second, crawler access is now a commercial decision. OpenAI's help centre says a site becomes eligible for ChatGPT search by allowing OAI-SearchBot — and making sure its host or CDN doesn't block OpenAI's published IP addresses. Blocking a company's search crawler removes you from its answers; blocking its training crawler doesn't. The full crawler table is in the GEO guide.

What does AI search do to clicks and website traffic?

AI search reduces clicks on the searches it answers. In Pew Research Center's study of 900 US adults' real browsing in March 2025, people clicked a traditional result on 8% of visits to a Google results page with an AI summary, against 15% without one — and clicked a link inside the summary itself on just 1% of visits.

Two more findings from the same study are worth knowing:

• 18% of the 68,879 Google searches in the sample produced an AI summary.
• People ended their browsing session on 26% of pages with an AI summary, against 16% of pages with only traditional results.

Google's counterpoint is scale. Alphabet says it is "sending billions of clicks to websites every week through AI features in Search". Both can be true: the total is enormous, while the share of searches that end in a click falls when an AI answer satisfies the question.

The effect isn't evenly spread. A question that can be fully answered in a paragraph — a definition, a quick fact — no longer needs a visit. A question that ends in a decision — which developer, which agency, which clinic — still sends people onward, but they arrive having already seen a shortlist the AI assembled.

The new discovery path: ask, shortlist, verify, act

The simplest way to plan for AI search is to map the four stages a customer now moves through, and what each stage needs from you:

Stage	What happens	What your business needs
1 · Ask	A long, specific question with constraints, often as a conversation	Content that answers the real questions customers ask, in their words and their languages
2 · Shortlist	The engine fans out, retrieves and names a few options or sources	To be retrievable (crawlable, indexed) and quotable (specific, dated facts)
3 · Verify	The customer checks the names: a cited page, a brand search, reviews, social profiles	Consistent facts everywhere — name, licence, address, services and public prices
4 · Act	A call, a WhatsApp message, a booking or a form — increasingly started by an assistant	A fast response and a clear next step on every page

Classic SEO fought for stage 2 as a list of ten links. AI search compresses stages 1 and 2 into a single answer, so the shortlist forms before the customer reaches any website. Stage 3 is where brands now win or lose trust: someone who sees your company named in a ChatGPT answer is likely to look you up before getting in touch, and inconsistent details at that moment cost you the lead.

What does AI search mean for businesses in Dubai and the UAE?

AI search matters more in the UAE than almost anywhere else, because the audience is already using it. Microsoft's AI diffusion research found that 73.3% of the UAE's working-age population used a generative AI product in the second quarter of 2026 — the highest share it measured anywhere, against 18.8% worldwide.

Google's AI search works in Arabic, too:

• May 2025 — AI Overviews launched across the Middle East and North Africa, and in Arabic globally, on mobile and desktop.
• October 2025 — AI Mode began rolling out in 38 new languages, including Modern Standard Arabic.
• April 2026 — Gemini's Personal Intelligence reached the Arab world, initially for Google AI Plus, Pro and Ultra subscribers.
• August 2026 — OpenAI opened self-service buying of ChatGPT Ads across India, Europe, the Middle East and North Africa.

The opportunity is the content gap. W3Techs' survey of website content languages counts Arabic on 0.6% of websites, against 49.5% for English (September 2026). An engine answering in Arabic can still draw on English pages, but a business that publishes useful pages natively in Arabic competes for those answers against far fewer sources.

A worked example: one property question

Take a question a relocating buyer might type into ChatGPT or AI Mode: "Which Dubai communities suit a family with a AED 4 million villa budget and children at a British-curriculum school?"

The engine fans that out into searches about villa communities, current prices, school locations and commutes, then builds a shortlist from what it finds. A brokerage that has published a clear, dated guide to exactly those trade-offs — with transaction data, service charges and named sources — is a candidate for the citation. A brokerage with only listings pages and a contact form isn't. The same logic applies to investors asking about yields, which is why a well-sourced page on real estate ROI in Dubai is worth more now than it was when the answer was ten blue links.

Where is AI search heading next?

AI search is moving in three directions at once: from answers to actions, from general to personal, and from free to paid.

• From answers to actions. Google's information agents, rolling out first to Google AI Pro and Ultra subscribers, work in the background and help you act on what they find. Google says Search will also build persistent dashboards and trackers for longer-running tasks. The search box is turning into a place where you delegate work — agentic AI explained covers how these agents operate.
• From general to personal. Gemini's Personal Intelligence can draw on Gmail, Google Photos and other connected apps, and ChatGPT can use saved memories to rewrite searches. Two people asking the same question will increasingly get different answers, which makes "what's my ranking?" an even less useful question than it was.
• From free to paid. ChatGPT Ads reached a $1 billion annualised revenue run rate in under 200 days, with tens of thousands of advertisers. OpenAI says ads are "always clearly labeled and separate from ChatGPT's answers" and don't influence them. OpenAI says a pixel and a Conversions API are now the foundations of its ad measurement — the same server-side pattern as Meta's Conversions API.

What should businesses do about AI search?

Businesses should treat AI search as an extension of search, not a replacement for it: make sure every engine can reach and index your site, publish the pages that answer your customers' real questions, keep your facts consistent everywhere, and measure where you're cited. Six steps, in order:

1. Check access. Confirm that Googlebot, Bingbot, OAI-SearchBot and PerplexityBot can crawl your site — in robots.txt and in your CDN or firewall settings.
2. Get indexed by Bing as well as Google. Verify your site in Bing Webmaster Tools, submit a sitemap and use IndexNow, because Copilot depends on Bing's index.
3. Answer the questions, not the keywords. List the 20–30 questions customers ask before they buy, and publish a page that answers each one directly in its first paragraph, with specifics.
4. Make your facts citable. Use dates, prices, figures and named sources. In the research that coined GEO, adding quotations, statistics and cited sources lifted visibility by 30–40% in generative engine answers.
5. Keep your identity consistent. Use the same business name, licence details, address and service descriptions on your site, Google Business Profile, LinkedIn and directories — the verify stage depends on it.
6. Measure citations, not only rankings. Use Search Console's generative AI performance report, Bing Webmaster Tools' AI Performance report, and referral traffic from chatgpt.com and perplexity.ai in your analytics.

Expert takeaway

Start with the questions that end in money. An AI answer to "what is a service charge?" costs you little if nobody clicks. An AI answer to "which agency handles off-plan resales in Dubai Hills?" that names three competitors costs you a client. Map those decision questions first, then make sure the best page on the web for each one is yours.

The detailed version, with crawler settings and a citation-ready page anatomy, is the 12-step GEO playbook. If you're still working out where SEO ends and GEO begins, start with SEO vs AEO vs GEO.

Final takeaway

AI search didn't replace search — it moved the shortlist into the answer. The engines still search, but they read for the customer and name a handful of sources before anyone visits a website. Be reachable by every engine, publish the clearest answers to your customers' real questions, keep your facts consistent and measure where you're cited. In the UAE, where most working-age adults already use generative AI, that shift isn't coming. It's here.