Model: Perplexity · Deep Research

On 19 June 2026, Perplexity shipped a more powerful Deep Research inside Computer — its agentic assistant — adding a command panel, thread forking, inline actions, analytics APIs and enterprise credit controls. When an answer engine researches deeper, the sources it decides to cite change with it.

Source: Perplexity Changelog — Deep Research, command panel, forking, inline actions and enterprise controls (19 June 2026)

What actually changed

Perplexity has brought its Deep Research mode into Computer, the agentic assistant that can browse, click and complete multi-step tasks on your behalf. Deep Research no longer just returns an answer — it runs a structured investigation, then hands back a sourced report you can keep working inside. The June update wraps it in a faster, more controllable workflow:

  • Deep Research in Computer — multi-step research now runs inside the agentic assistant, so it can gather, act and produce in one flow.
  • Command panel — faster keyboard-driven access to actions, so researchers move through a project with fewer clicks.
  • Forking — branch a research thread to explore a tangent without losing the original line of enquiry.
  • Inline actions — act on a result (refine, expand, export) directly where it appears.
  • Analytics APIs — programmatic visibility into how research and credits are being used across an organisation.
  • Custom credit limits and enterprise controls — admins can cap and account for usage across teams.

Under the hood, Perplexity now runs Deep Research on a frontier reasoning model for its Max and Pro tiers and reports state-of-the-art results on external research benchmarks. In plain terms: the mode reads more, reasons harder and assembles a longer, better-cited answer than a standard query.

Why a deeper research mode changes who gets cited

Perplexity is a citation-first answer engine: every response is built from named sources, and those sources are linked in front of the user. A standard query might lean on a handful of pages. Deep Research is different — it fans out across many searches, reads through far more material, and synthesises a multi-section report. That shift matters for any brand that wants to be named when buyers ask AI about their category.

Infographic

Standard query vs Deep Research: how the citation bar shifts

How much does it read?

Standard querySamples a handful of pages before answering.
Deep ResearchFans out across many searches and reads far more material before it writes.

What clears the bar to be cited?

Standard querySurface-level SEO can be enough to sneak in.
Deep ResearchThin or contradictory content gets filtered out — substance survives the wider read.

What gets quoted?

Standard queryA quick, top-level answer.
Deep ResearchClean, attributable lines — direct answers, defined terms, comparison tables, dated figures.

Where does it look?

Standard queryBroad head terms and the obvious pages.
Deep ResearchSpecific, multi-part “best X for Y” questions — the long tail you may have ignored.

A deeper research mode reads more sources and applies a higher bar — clear, specific, well-evidenced pages are the ones that make it into the final report.

More sources read means a higher bar to be included

When the engine only samples a few pages, surface-level SEO can be enough to sneak in. When it reads deeply, thin or contradictory content gets filtered out in favour of pages that actually answer the question. Depth rewards substance — the brands with clear, specific, well-evidenced content are the ones that survive the wider read and make it into the final report.

Structure and clarity win the citation

Deep Research assembles its report section by section, so it favours sources that are easy to lift a clean, attributable claim from: direct answers, defined terms, comparison tables, dated figures. Pages that bury the answer under marketing copy are harder to quote — and an engine that needs a citeable line will reach for the source that gives it one.

What this means for your brand

  • Answer the question outright. Lead each key page with a plain, quotable statement of what you do, who it is for and how it compares — then support it. Make the citeable line easy to find.
  • Cover the long tail. Deep Research probes specific, multi-part questions. The comparisons, use cases and “best X for Y” queries you have ignored are exactly where it now looks — and where competitors get named instead of you.
  • Keep facts consistent and current. A deeper read cross-checks sources. Stale pricing, conflicting claims or outdated stats across your own pages give the engine a reason to trust someone else.

How to see where you stand in Perplexity answers

You cannot improve what you cannot see. The practical first step is to know whether Perplexity — and the other answer engines — name you today, which sources they pull from, and where rivals are cited in your place. That is exactly what reconnAI monitors: how your brand shows up across AI answers, and how that changes as models and features like Deep Research evolve.

The takeaway

  • Perplexity’s Deep Research is now more powerful and more usable — running inside the Computer assistant with forking, inline actions and enterprise controls.
  • A deeper research mode reads more sources and applies a higher bar, rewarding clear, specific, well-evidenced content over thin SEO pages.
  • The brands that win citations are the ones that answer questions outright, cover the long tail, and keep their facts consistent.
  • Knowing where you currently appear in AI answers is the starting point for fixing it.

Want to know whether Perplexity is naming your brand — or your competitors — when buyers ask? Get in touch with reconnAI and we will show you where you stand across the AI answer engines.