OpenAI Deep Research vs Perplexity Pro: Which Agent Actually Delivers a Usable Market Report?

Most AI search engines are just polite version of Google. You ask them a question, they skim the top four SEO-optimised blog posts on the web, spit back a neatly formatted summary of what they found, and call it a day. If you are trying to settle a bar bet, that is fine. If you are trying to write a serious market report for an investment deck or a product launch, it is completely useless.

However, a new class of "agentic" research tools promises to change this. Instead of a single search-and-response cycle, these systems perform multi-step, iterative research. They search, read, realise they are missing info, search again, dig through PDFs, and compile their findings.

We put the two leading platforms in this space to the test: OpenAI Deep Research (available inside ChatGPT Plus and via the API) and Perplexity Pro (using its "Pro Search" agentic setting).

We wanted to see what makes these research agents tick, so we tasked both with a notoriously difficult research brief: Analyse the current market landscape for solid-state sodium-ion batteries, including key patent holders, active pilot plants in Europe, and realistic production cost-per-kilowatt-hour projections through 2030.

Here is how they performed across the metrics that actually matter to builders and analysts.

Round 1: Depth of Search and "Rabbit Hole" Traversal

A good analyst doesn't just read the first page of Google; they track down the original research papers, press releases, and financial filings.

Perplexity Pro approached the task like a very fast librarian. It ran an initial query, extracted about ten sources, ran a secondary query to clarify some details, and then began writing. It took about 45 seconds. The sources it pulled were decent—mostly trade publications and industry news sites. But it completely missed the academic research papers and patent registries.

OpenAI Deep Research took a completely different approach. It didn't rush. It spent a solid six minutes thinking, planning, and executing dozens of targeted search queries. It explicitly looked for filetypes like .pdf from academic domains and patent databases. When it encountered a paywall on a major trade site, it didn't just give up; it searched for alternative sources hosting the same data or looked for public press releases from the companies involved to cross-verify the figures.

Winner: OpenAI Deep Research. It actually behaves like an agent, digging deep into the web rather than just skimming the surface.

Round 2: Synthesis, Structure, and Writing Quality

Getting the data is only half the battle. If the final report is just a collection of disjointed bullet points, you still have hours of work ahead of you.

Perplexity Pro delivered its report using its standard layout. It was neat, highly readable, and divided into sensible sections. However, the depth was lacking. It gave us broad ranges for the cost-per-kilowatt-hour (e.g., "$40 to $100 by 2030") without explaining why the projections varied so wildly or what chemical compositions (like Prussian blue analogues vs. layered transition metal oxides) drove those differences.

OpenAI Deep Research delivered a massive, multi-page document that read like it was written by a senior analyst at McKinsey. It didn't just list the pilot plants; it organised them into a structured table showing location, planned capacity (in MWh), and partnership structures. It explained the technical hurdles of solid-state sodium-ion interfaces in rigorous detail, drawing on terms you would find in our /glossary rather than watered-down consumer jargon. It even provided a detailed breakdown of the supply chain bottlenecks for precursor materials.

Winner: OpenAI Deep Research. The level of synthesis and academic rigour was in a completely different league.

Round 3: Citation Accuracy and the Hallucination Test

In market research, a fake statistic is worse than no statistic at all. If you present hallucinated data to a client or an investor, your credibility is ruined.

We rigorously fact-checked every single citation in both reports.

• Perplexity Pro had a 100% citation accuracy rate for the sources it did find. Every link went to a real webpage that contained the cited fact. However, because its search was shallow, it occasionally cited opinion pieces and speculative blog posts as if they were established market facts.
• OpenAI Deep Research also had near-perfect citation accuracy. It linked directly to patent filings on Google Patents and academic PDFs. However, because it compiles such a massive volume of data, we noticed one minor error where it attributed a pilot plant capacity figure from a Chinese manufacturer to a European competitor. If you run into odd formatting or API-related citation bugs with OpenAI's outputs, their developer community at https://www.openai-support.com is a great place to troubleshoot structured data schemas.

Winner: Tie. Both tools have successfully eliminated the wild, groundless hallucinations of older LLMs by anchoring their outputs strictly in retrieved search chunks.

Pricing, Limits, and Speed

• Perplexity Pro: Costs $20/month. It is incredibly fast, giving you a comprehensive (if slightly shallow) answer in under a minute. It is perfect for high-velocity, day-to-day queries where you need to get smart on a topic in five minutes.
• OpenAI Deep Research: Included in ChatGPT Plus ($20/month) but with strict usage limits (often limited to a few deep runs per day due to the massive compute costs of running sequential searches). You can also run it via the API using their reasoning models, but be warned: a single deep research run can easily cost several dollars in API tokens depending on how many steps you allow the agent to take. For a detailed breakdown of how this compares to raw token usage, see our guide on /platforms/openai.

The Final Verdict

If you need to quickly understand a new industry or prep for a meeting in fifteen minutes, Perplexity Pro is your best bet. It is fast, clean, and highly reliable for everyday search tasks.

But if you are building an actual business case, writing a whitepaper, or conducting serious competitive intelligence, OpenAI Deep Research is the undisputed king. It is the first AI research tool that doesn't just feel like a faster search engine—it feels like a highly capable junior analyst who went away for an hour and actually did the work.