Agentic Search Grew 200%: Your Next Customer Is Already Asking an AI

The data that confirms the shift

On July 28 Salesforce published its Commerce Snapshot with data from April to June 2026, drawing on the activity of 1.5 billion shoppers plus surveys of 3,450 commerce professionals and 4,690 consumers. The headline is a single number: agentic search grew 200% year over year as the first step of the purchase journey.

There is more: traffic arriving at sites from AI chats grew between 150% and 428% per quarter (against single-digit growth in overall traffic), and traffic to brand-owned properties fell 7%. This is not an early-adopter fad either: 28% of organizations already use agentic AI and another 44% plan to adopt it within six months. Only 4% remain in pilot mode.

The report's central idea: customers no longer start with ten blue links. They ask AI what to buy, and they arrive at your site with an opinion already formed — if they arrive at all. That is how Caila Schwartz, Salesforce's director of consumer insights, sums it up.

If you read our article on Agentforce Commerce, this will sound familiar: a month ago we covered how Salesforce put product catalogs inside ChatGPT. This report is the other half of the story — the demand is already there, and it is growing faster than almost anyone expected.

What agentic search is (without the jargon)

Traditional search works like this: you type keywords, the engine gives you a list of links, and you do the work of opening, comparing and deciding. Agentic search flips the roles: you ask an AI agent to do that work for you. "I need a CRM for a 40-employee distributor in Uruguay — which one fits and who should implement it?" The agent researches, compares, discards and comes back with two or three concrete options.

The commercial difference is enormous. In traditional search you competed for a click among ten results. In agentic search you compete to be inside the answer. If the agent doesn't know you, doesn't understand you or can't cite you with confidence, you don't exist for that customer — no matter how polished your classic SEO is.

SEO vs. GEO: what changes when the searcher is an agent

This is where the new discipline comes from: GEO (Generative Engine Optimization), also called AEO. It does not replace SEO — AI engines feed on the same content — but it changes what gets rewarded:

• Direct answers, not keywords. Models cite pages that answer concrete questions ("how much does a Salesforce implementation cost in Mexico?") better than pages optimized around loose keywords.
• Structured data. Schema.org (FAQPage, Organization, Product, BreadcrumbList) gives the model verifiable facts it can repeat without fear of getting them wrong.
• Citable, specific information. Prices, timelines, countries where you operate, case studies with numbers. Vague content doesn't get cited.
• Machine readability. Files like llms.txt, clean HTML and connected catalogs let an agent read your offering without guessing.
• Consistency across sources. Models cross-check: your site, your LinkedIn, directories, reviews. If they disagree, the agent hesitates — and a hesitant agent doesn't recommend.

Why this hits differently in Latin America

First, because of the starting point: discovery in the region was already migrating to conversational channels — business here has run on WhatsApp since long before the word "agentic" existed. The jump from "I'll ask someone I know" to "I'll ask an AI" is shorter than in other markets.

Second, because of the window of opportunity. The report shows 86% of commerce leaders acknowledge AI has raised customer expectations, and 61% admit meeting them is harder than ever. Yet most companies in the region have not adapted their content or their data to this new layer. Whoever structures first keeps the recommendations — and agent recommendations tend to concentrate on very few options.

How to get your business ready to be the answer

The same criterion we use with Agentforce applies here: you don't need to do everything at once, you need to start with what moves the needle.

The uncomfortable question to close

Try it right now, seriously: open ChatGPT and ask which company it recommends for what you sell, in your country. If the answer doesn't include you, that is exactly the size of the problem — and the good news is that almost none of your competitors are working on it yet.

How Zarasa can help

We run GEO/SEO audits of digital presence and we are a Salesforce consulting partner specialized in Latin America: we can tell you how AI engines see you today, what to fix first and, if you sell online, how to connect your catalog to the agentic layer with Salesforce. The first conversation is free.

Frequently asked questions

Does GEO replace SEO?

No. AI engines feed on the same content traditional search indexes, so SEO remains the foundation. GEO adds a layer on top: structuring your information so an AI model can understand it, cite it and recommend it with confidence.

Do I need Salesforce to show up in AI answers?

Not for the content layer: any well-structured site can be cited by an AI engine. To sell directly inside a chat like ChatGPT you do need a connected catalog, and that is where Agentforce Commerce comes in.

How do I know if AI chats are already sending me traffic?

Check the referrers in your analytics: chatgpt.com, gemini.google.com, perplexity.ai and similar. In GA4 you can build a dedicated segment. Salesforce reports 150% to 428% quarterly growth in that traffic; if you are not measuring it separately, you will not see it.

Does this apply to B2B or only to e-commerce?

It applies just as much — or more — in B2B. Corporate buyers also research with AI before requesting a demo or a proposal, and agent answers influence which vendors make the shortlist.