Evening BriefAI Search & Discovery

Perplexity Built a Research Machine for GDP-Moving Decisions. Your Brand Isn't in the Brief.

Perplexity Computer orchestrates 19 AI models for deep enterprise research — explicitly designed for 'GDP-moving decisions.' Its Draco benchmark outperforms competitors on complex research tasks. Tomorrow its API opens to builders. Here's what that means for how your brand gets discovered before the first call.

Jaxon Parrott|Mar 10, 2026

"You don't hear us talk about MAUs ever."

That was a Perplexity executive two weeks ago, explaining what the company had just built and who they built it for. The product was Perplexity Computer. The audience was not the general public. The audience was the executive making what Perplexity called "GDP-moving decisions."

That positioning is worth unpacking.

What the product is

Perplexity Computer is a cloud-based AI agent that takes a complex research objective, breaks it into subtasks, and routes each subtask to whatever model handles it best. Core reasoning runs on Claude Opus 4.6. Deep research on Gemini. Long-context recall on ChatGPT 5.2. Speed and code tasks on Grok. The system runs for hours, sometimes months, and delivers finished outputs — research reports, competitive analyses, procurement briefings, acquisition rationales.

Perplexity also released Draco, a benchmark for complex multi-step research tasks. Their deep research offering scores higher than every competitor's on that benchmark.

The product costs $200 per month, only available on Perplexity Max. Enterprise Max access is rolling out now. Tomorrow, March 11, Perplexity hosts its first developer conference in San Francisco to open the API to outside builders.

Who this is for

Perplexity's public positioning is explicit: people making decisions with genuine financial stakes. Strategy teams. Senior operators. Procurement leaders. Executives running competitive intelligence before committing to something large.

The $200 price point is designed to filter for this use case. So is the enterprise rollout timeline.

What that means in practice: some portion of your pipeline — the part that matters most — runs their pre-purchase research through a 19-model AI built to outperform every alternative on exactly that task.

How it decides what to cite

Perplexity Computer doesn't read your website before citing you. It doesn't evaluate your landing page copy or SEO-optimized headers. It synthesizes what trusted, third-party publications have written about your company, your category, and your competitors. Its search index is proprietary, built specifically for AI-powered retrieval rather than traditional keyword signals.

A September 2025 study published on arXiv analyzed 1,702 citations across Brave Search, Google AI Overviews, and Perplexity, covering 16 B2B SaaS categories. The finding: AI answer engines "systematically favour earned media — third-party, authoritative domains — over brand-owned and social content, with social platforms almost absent from AI answers." Brand-owned domains were structurally excluded regardless of on-page quality. The deciding variable was whether content lived on a domain the AI already treated as credible from external signals.

Perplexity's enterprise positioning amplifies that dynamic. The system is designed to produce work product that looks credible to a senior decision-maker. That means it needs to cite sources those decision-makers would recognize as credible. Forbes. TechCrunch. Industry trade press. Research publications. Analyst coverage. The same publications that built category authority with human readers for decades.

If that publication record doesn't exist for your brand, the output doesn't include you.

The 19-model problem, restated

Earlier this year, the question was whether your brand passed the basic "does ChatGPT mention us" test. That test measured single-model recall.

Perplexity Computer raises the bar. It routes your category query through whichever combination of models produces the most thorough, most accurate answer. If a model that doesn't know your brand happens to own the relevant subtask, you disappear from the output regardless of your ChatGPT visibility score. The coverage requirement just expanded from one model to an orchestrated research system with its own search index, its own citation logic, and a benchmark it uses to prove the output is better than anything else available.

The timing

The March 11 developer conference matters for a specific reason: when the API opens, third-party builders start deploying Perplexity Computer inside enterprise tools. Due diligence platforms. Procurement software. Strategy tools. Board preparation systems. The research layer doesn't stay inside Perplexity's own interface for long.

The channels through which your brand gets researched just expanded. The mechanism — what gets cited and why — stays exactly the same.

AI has stopped ranking vendors and started shortlisting them. Perplexity Computer is the most explicit engineering statement of that thesis yet. It is not a search tool. It is a research infrastructure layer purpose-built for decisions with dollar signs attached.

What this requires

Perplexity's explicit bet is that the company willing to move away from MAU metrics toward decision quality wins the part of the market that actually moves revenue.

For B2B brands, the implication is simpler than it sounds. The buyers Perplexity is building for are the buyers worth having. And what Perplexity tells those buyers about your category is controlled by one thing: what trusted publications have already written.

That's what Machine Relations means at its core. Not SEO. Not content volume. Earned editorial coverage in sources that carry credibility signals AI systems treat as authoritative. The publication record that precedes your first call now gets synthesized by a 19-model research system explicitly designed for high-stakes decisions.

It's worth knowing what that system currently says about you. The visibility audit shows you.