Skip to main content
ChatGPT's Deep Research tool interface with a built-in viewer, table of contents, and highlighted citations. [theverge.com](h

Editorial illustration for ChatGPT deep research tool adds built-in viewer and jump-to-section TOC

ChatGPT Deep Research: AI Analyst Tool Unveiled

ChatGPT deep research tool adds built-in viewer and jump-to-section TOC

Updated: 3 min read

ChatGPT’s research tool used to be a monologue. It talked, you listened. That’s changing.

The feature, which constructs long reports from web sources, now has a built-in reader and a clickable table of contents. You can read its work directly in the chat and jump to any section. More significantly, you can steer the process, telling the bot to focus on specific websites or apps, adding sources mid-stream, and watching its progress live.

ChatGPT's deep research tool adds a built-in document viewer so you can read its reports You can also open a new table of contents that allows you to jump to specific sections of the report. You can also open a new table of contents that allows you to jump to specific sections of the report. Deep research, which OpenAI first launched last year, has ChatGPT scour the web to compile an in-depth report about the topic of your choosing.

With this most recent update, you'll be able to ask ChatGPT to focus on specific websites and connected apps for its research. You can also track ChatGPT's progress in real-time, as well as edit the scope of its research or add new sources while the chatbot generates a report.

The real news isn't the viewer or the table of contents, though they help. It’s the steering. Letting users add sources or redirect the search while a report is being built turns a batch job into a conversation.

The tool becomes less of an oracle and more of a collaborator you can interrupt. This feels like a quiet correction. Instead of pushing for more autonomous, hands-off AI, OpenAI is building a tool that demands more hands-on time.

The value shifts from the final answer to the process of getting there.

Common Questions Answered

How does the new Deep Research document viewer improve report readability?

The new viewer adds a full-screen document window separate from the chat pane, allowing users to scroll through AI-generated reports in isolation. It includes a table of contents on the left and a list of sources on the right, enabling users to jump directly to specific sections and quickly navigate complex research reports.

What are the usage limits for Deep Research across different ChatGPT subscription tiers?

ChatGPT Pro subscribers receive 250 queries per month (half being 'lightweight'), while Plus, Team, and Enterprise users get 25 queries monthly (15 being 'lightweight'). Free users are limited to 5 'lightweight' queries per month, with the lightweight version based on the o4-mini model instead of the full o3 model.

How does ChatGPT's Deep Research feature perform on complex research benchmarks?

Deep Research scored 26.6% on the 'Humanity's Last Exam' benchmark, significantly outperforming competitors like DeepSeek's R1 model (9.4%) and GPT-4o (3.3%). However, OpenAI acknowledges that the system can occasionally make factual hallucinations or incorrect inferences, and may sometimes reference rumors without accurately conveying uncertainty.

LIVE14:18Claude Cowork Brings AI Skills and Plugins Directly to Chrome Browser