ChatGPT Deep Research
OpenAI's cited, multi-step research agent inside ChatGPT
Overview
Deep research plans, then autonomously browses for 5-30 minutes and returns a long cited report you can steer, restrict to chosen sites and export. Limited use on the free plan; expanded on Plus ($20/mo) and maximum on Pro ($100, $200 or $500/mo), with OpenAI no longer publishing task counts.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- In-product task counter
- Resets every 30 days
- GPT-6 reasoning models
- Connected-app sources
- Pro reasoning models
- Expanded Codex usage
- Higher usage limits
- Highest usage limits
- Ultrafast mode
Use cases
What you can produce with ChatGPT Deep Research
- Autonomous 5-30 minute multi-source web research runs
- Editable research plan, mid-run interruption and site restrictions
- Long-form cited reports with a table of contents
- Read-only research across connected apps
- Export to Markdown, Word and PDF
- Deep research on every plan, including Free, with a 30-day task counter
Deep research is the most consequential agent mode in ChatGPT: give it an open-ended question and it plans, browses and reads for 5 to 30 minutes before handing back a structured, cited report — work that would take an afternoon by hand. 2026 made it easier to steer, with editable plans, site restrictions, connected-app sources and Word or PDF export, and it now runs on OpenAI's latest model. Every plan includes some deep research, from limited on the free tier to maximum on the new $500 Pro plan, but OpenAI no longer says how many tasks each gets. The honest caveat: it still invents sources. A 2026 study found deep-research agents fabricate citation links more often than ordinary chat, at rates up to 13%. Read every citation before you cite it yourself — it is a strong first-draft generator, not a fact-checked deliverable.
What we found
Deep research is ChatGPT's agent mode for long questions: it proposes a research plan you can edit, then spends from five to thirty minutes, by OpenAI's own estimate, browsing and reading before returning a long report with a table of contents and cited sources. In 2026 OpenAI added more control over that process — you can interrupt a run, restrict it to particular websites or tell it to prioritise them, and let it read from connected apps using read-only access — and reports export as Markdown, Word or PDF. It now runs on whatever OpenAI's latest model is by default, rather than the specialised o3 build it launched with; the older deep-research API models were shut down in July 2026.
Access is on every plan, but OpenAI no longer publishes task counts. The free and Go plans get "limited" deep research, Plus at $20 a month "expanded", and the Pro plans "maximum"; an in-product counter shows how many tasks remain and resets every 30 days. Pro now comes in three sizes: $100, $200 and, since 29 September 2026, $500 a month, with Pro 200 reopened to new subscribers at a lower allowance than before. Business and Enterprise plans can keep data in Singapore, Japan, Korea, India or Australia.
Accuracy is the caveat. A study published on arXiv in April 2026 by researchers including Chris Callison-Burch tested commercial models and deep-research agents and found between 3% and 13% of citation links were fabricated, with deep-research agents hallucinating URLs at higher rates than plain chat. On G2, ChatGPT averages 4.6 out of 5 from about 3,000 reviews, but "AI limitations", context understanding, usage limits and inaccuracy are the most-cited drawbacks. The study tested several commercial systems, so this is not an OpenAI-only problem, but a polished report makes errors easier to miss.
Used well, deep research turns an afternoon of searching into a first draft. Check every citation before you rely on it.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including ChatGPT Deep Research's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with ChatGPT Deep Research unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to ChatGPT Deep Research directly →
Spotted something out of date? Suggest an update →
Alternatives to ChatGPT Deep Research
More in Research & Data
