Perplexity AI: What Changes for Businesses in 2026

Perplexity AI is no longer "that search engine with AI" and has become one of the most talked-about companies in the sector in 2026 — with an annualised revenue of around US$500 million, its own browser, and an API platform for building agents. If you run marketing, customer service or technology in a Brazilian company, it's worth understanding what has changed in recent weeks and why it directly affects how your customers will find you.

TL;DR

Perplexity AI is an "answer engine": it responds with text and cites sources, instead of returning ten blue links.

In June 2026 it launched Deep Research on Computer, turning a question into a report, spreadsheet or dashboard in the same place.

The Comet browser arrived on iOS, Android, Windows and Mac and is already integrated with Samsung Internet.

The new API platform (Search, Agent, Embeddings and Sandbox) allows companies to build their own AI agents.

For your business, the game now is to be cited within the answers — not just rank on Google.

What is Perplexity AI and why did it grow so much in 2026

Perplexity AI is an answer engine: you ask a question in natural language and it returns a synthesised text, with numbered sources below. The difference from classic Google is fundamental. The traditional search engine gives you a list of links and leaves the reading work to you. The answer engine reads for you and delivers the conclusion.

This model exploded. According to market data compiled in 2026, Perplexity went from around 45 million monthly active users and 780 million searches per month. Annualised revenue crossed US$500 million in April 2026 — a jump of over 300% in one year. In an interview with CNBC{target="_blank"}, CEO Aravind Srinivas credited much of this acceleration to advances in Anthropic's models, which power part of the product.

The point that matters to the business owner is not the billion-dollar valuation. It's behavioural: part of your audience already asks the AI before opening Google. And when the answer comes ready, with sources, the click on your site only happens if you are one of those sources.

It's worth separating two terms that are often confused. The closed chatbot responds based on what it learned during training and rarely shows where it got the information. The answer engine, on the other hand, searches the web in real time and cites what it used. This difference is why Perplexity AI is taken seriously in professional contexts: the source is there, clickable, and that changes the relationship of trust — and the opportunity for traffic for those who produce quality content.

Comet: the AI-powered browser targeting traditional search

The company's most aggressive move is Comet, the browser with a built-in AI assistant. Instead of opening ten tabs and reading each one, Comet's assistant summarises pages, answers follow-up questions, and performs tasks autonomously — the so-called "agentic" use of the browser.

In 2026, Comet gained scale: after arriving on iOS in March, it expanded to Android, Windows and Mac, and its agentic capabilities were integrated into Samsung Internet, with Perplexity available as an optional default search engine on Samsung devices. Voice Mode also evolved, running on the latest version of OpenAI's real-time voice model.

Why does this matter in Brazil? Because the smartphone is the entry point for most of your customers. A browser that summarises and decides further reduces the space for traditional organic traffic. It's the same movement we saw with Google's assistants — a topic we've covered in Google Gemini: what changed at I/O 2026.

Comet free vs paid: what to consider

Perplexity offers Comet with a free tier and paid layers (Plus, Pro and Max), where more advanced subscribers can even choose the AI model that powers the browser agent. For a small business, the free tier is enough to test the search and summary flow in daily use. The decision to pay usually comes when usage becomes routine — the whole team summarising documents, comparing suppliers, or gathering market data. The practical advice is not to sign up on impulse: run two weeks on the free plan, measure how many tasks the AI actually shortened, and only then choose the tier.

The new API platform: Perplexity as infrastructure

Here the subject gets interesting for developers. Perplexity AI opened its API platform and stopped being just a product to become infrastructure. Today it exposes four fronts:

Search API — real-time web search, with results already ranked and in short snippets, instead of full documents.
Agent API — orchestrates multi-step flows, combining search, tools, models and reasoning.
Embeddings API — large-scale information retrieval, the basis for RAG systems.
Sandbox API — secure code execution (announced as "coming soon").

The official documentation itself describes the Search API{target="_blank"} as the same infrastructure that powers the public product. In practice, this means a Brazilian company can build a customer service agent, an internal research assistant, or its own search engine without building the search and orchestration layer from scratch.

It's exactly the kind of architecture we use to connect AI to real service channels — something we detail in AI Agents: what Gemini Spark changes for businesses.

Deep Research on Computer: the turning point of June 2026

The most recent update came out on 19 June 2026 and shows where Perplexity is heading. Deep Research on Computer allows you to start with a complex question and, from the result, turn the findings into a report, spreadsheet, presentation, dashboard, website or workflow — all in the same environment.

The same batch brought features aimed at businesses:

Command Panel: typing "/" opens a command panel to find modes and skills.
Forking: explore new variations of a search without losing the previous context.
Computer Analytics API: extract usage per member and per day, programmatically.
Custom credit limits: administrators set usage caps per person.

For an operation thinking about adopting AI with governance, these usage controls and analytics are more relevant than any benchmark. They define whether the tool fits within the company's data policy.

What changes for your company's SEO

The practical effect for those selling online is clear: in addition to ranking on Google, you now need to be cited by AIs. This discipline has gained a name — AEO (Answer Engine Optimization) or GEO (Generative Engine Optimization). The logic changes in important ways:

Dimension	Classic Google	Perplexity AI (answer engine)
Result	List of 10 links	Synthesised answer + cited sources
SEO goal	Position on the page	Being cited within the answer
What matters	Authority, backlinks, on-page	Factual clarity, data, reliable sources
Success metric	Clicks and CTR	Brand mentions and citations
Ideal content	Comprehensive and long	Direct, structured, verifiable

It's not that classic SEO is dead — it remains the foundation. But shallow content, without its own data and structure, loses in both worlds at the same time. Those who write with depth, numbers and sources tend to be both well-ranked and well-cited.

How to appear in Perplexity AI answers

There is no magic button, but there is a set of practices that increases the chance of Perplexity AI using your content as a source:

Answer the question right at the start. Answer engines reward the snippet that delivers the direct answer in the first few lines.
Structure with headings and lists. Scannable content is easier to extract as a snippet.
Use verifiable data and cite the source. Numbers with a source are exactly what an answer engine looks for.
Maintain factual consistency across your pages. Internal contradictions hurt your reliability.
Have structured data (schema) and FAQ. FAQ and article markup help the AI understand the context.
Update content. Perplexity prioritises recent information; publication and review dates matter.

The good news: almost all of this also improves your performance on Google. It's not double the work — it's the same work, done right.

Risks and limitations: lawsuits, sources and hallucinations

Not everything is enthusiasm. Perplexity AI faces significant disputes, including a lawsuit filed by the New York Times involving copyright, hallucinations and trademark use. The sensitive point is that answer engines can cite inaccurately, attribute wrong statements to a source, or summarise content out of context.

For your company, this has two readings. First, as a content producer: monitor how the AI cites your brand, because a wrong summary can become a rumour. Second, as a user: never treat the AI's answer as final truth in critical decisions — always open the cited source. The very existence of citations is an advantage of Perplexity over closed chatbots, but it does not replace human verification.

Where Agathas Web comes in

At Agathas Web, the reading is practical: Perplexity AI and other answer engines are not an abstract threat, they are a channel change. We work on three fronts to prepare Brazilian companies for this scenario.

First, content and SEO: we produce material with the density, structure and data that both Google and AIs value — exactly the standard we apply in this blog. Second, AI agents and automation: we use search and orchestration APIs (like those Perplexity just opened) to build assistants that answer customers based on your own knowledge base. It's the same philosophy that drives our Voyia, with unlimited agent service on WhatsApp.

Third, websites and structured data: we ensure that every page has correct schema, FAQ and metadata, so that your brand is readable by both search engines and answer engines. The goal is simple: that when someone asks the AI about your segment, the answer comes with you in the footer of the sources.

Conclusion: the next step

Perplexity AI consolidated in 2026 a model that goes beyond search: an agentic browser, an API platform and integrated deep research. For Brazilian companies, the message is direct — adapt your content to be cited, consider the API to build your own agents, and treat AI answers with the healthy scepticism you always should.

If you want to understand how to position your business in this new map — from citation-oriented SEO to service agents — talk to the Agathas Web team. The best time to prepare your content for answer engines was yesterday; the second best is now.