OpenAI Expands Deep Research Controls with Full-Screen Viewer and Source Selection
OpenAI has introduced a series of updates to ChatGPT’s Deep Research feature, giving users greater oversight of how the tool gathers and presents information. The changes include a full-screen document viewer, the ability to specify trusted sources, and real-time progress monitoring.
The update addresses a practical concern for professionals who have struggled to read multi-page AI-generated reports within the standard chat interface. Reports can now be opened in a dedicated viewing window, with a table of contents positioned on the left side and source citations displayed on the right. Users can export finished documents in Markdown, PDF, or Word formats.
According to OpenAI’s release notes, the company has redesigned the feature to produce “more accurate, credible reports with greater control.” Users can now direct the research process toward specific websites and connected applications, rather than relying solely on open web searches. This represents a meaningful shift for researchers who require information from verified databases, institutional repositories, or industry-specific platforms.
The Deep Research tool, which launched in 2024, operates by breaking down complex queries into component questions, then spending between five and thirty minutes gathering and cross-referencing information from across the web. The latest version allows users to intervene during this process, adjusting the scope of investigation or adding new sources while research is underway.
OpenAI has also upgraded the underlying model powering Deep Research to GPT-5.2, which the company says delivers improved accuracy compared to earlier iterations. The feature now supports connections to a broader range of external services, demonstrated during the announcement through integration with financial data from the London Stock Exchange Group.
The updates are currently rolling out to ChatGPT Plus and Pro subscribers, with availability for users on the Go tier and free plans expected in the coming days. Business, Enterprise, and Education accounts will receive access according to OpenAI’s standard deployment schedule.
For African organisations exploring AI-assisted research workflows, the source-selection capability could prove particularly valuable in sectors where local data sources and regional regulatory publications need to take precedence over generalised web content. Legal practitioners, policy researchers, and corporate analysts working across Nigeria and the wider continent have begun testing AI tools for case preparation and market intelligence, though concerns about citation accuracy remain a barrier to broader adoption.
The document viewer update represents OpenAI’s response to feedback that different types of AI output require different interface approaches. While conversational exchanges work well for quick queries, comprehensive research documents demand the navigation tools typically associated with traditional document readers.
The feature’s reliability will likely determine whether it becomes a standard component of professional research workflows or remains a supplementary tool requiring extensive verification. OpenAI has not disclosed specific benchmarks for citation accuracy or indicated whether the updates address earlier criticisms about AI-generated sources that cannot be verified through conventional academic databases.
