ChatGPT and the new Tools interface: six modes to access Agent, Research, Study and Creation

A redesigned interface brings quick access to advanced AI capabilities like Agent Mode, Deep Research, Study Mode, and more—merging assistance, reasoning, and creativity in one panel.

In late July 2025, OpenAI rolled out a visual redesign of the ChatGPT interface across web and mobile. The core of the update is the introduction of a Tools dropdown in the chat input area, giving users direct access to six specialized functions: Agent mode, Deep research, Create image, Study and learn, Web search, and Canvas.

These features were previously accessible through separate workflows or hidden behind less intuitive menus.

The change is not purely cosmetic: it marks a major step in consolidating ChatGPT’s roles as an agent, teacher, researcher, designer, and co-author—within a single conversational flow.

The update is being gradually released to Plus, Pro, and Team subscribers globally, with partial availability on the free tier.

Agent mode can now carry out full tasks on your behalf

ChatGPT takes action by browsing websites, filling forms, organizing files, and generating full reports—inside a virtual desktop.

When Agent Mode is selected, ChatGPT enters a new operational layer: it executes tasks step by step in a sandboxed environment resembling a remote desktop. It can search online, click buttons, extract data from websites, summarize PDFs, populate spreadsheets, and even submit web forms. Every action is confirmed with the user and logged transparently. If an action seems sensitive—like entering payment info or accepting cookies—it asks for explicit permission. This turns ChatGPT into a semi-autonomous assistant capable of completing multi-step workflows rather than just offering suggestions.

Benchmarks released by OpenAI suggest that Agent Mode outperformed human research assistants in time-limited financial tasks, while maintaining safety and rollback options.

Deep research builds professional-grade reports with citations

The assistant browses, reads, filters, and writes—leaving behind a structured file with sources, summaries, and analysis.

With Deep Research enabled, ChatGPT launches a dedicated process that mimics the work of a research analyst. It searches across the web, academic databases, and documents, then assembles the results into a comprehensive answer. Alongside the final report, the user receives footnotes, clickable citations, topic clusters, and optional summaries for each reference. The process unfolds in stages, with ChatGPT sending live progress updates and allowing mid-session keyword refinements.

In this mode, reasoning is explicit: the assistant shows which source backs which idea, avoids unsupported claims, and warns about potential bias in low-quality sources.

Study and learn replaces answers with step-by-step guidance

Instead of giving you a solution, ChatGPT becomes a Socratic tutor, helping you think through the question yourself.

“Study” mode radically alters how ChatGPT replies. The answer is not just revealed—it’s reconstructed through inquiry. The model now returns open-ended questions, asks for predictions, proposes analogies, and offers correction or feedback based on your attempts. This guided process mimics the logic of active learning and retrieval practice. Students can request tailored difficulty levels, while teachers can assign custom learning paths within the same interface.

Once the mode is active (via the new book-shaped icon next to the model selector), every query becomes an opportunity to build critical thinking. Even familiar questions yield new depth through reflection and feedback.

image transforms prompts into coherent illustrations

From quick sketches to detailed visuals, ChatGPT now generates contextual images using DALL·E 3 and GPT-4o.

This feature offers immediate image generation based on textual prompts, tightly integrated with the conversation’s context. Unlike previous image tools, “Create image” lets users build upon earlier messages or describe scenes in multiple iterations. With GPT-4o handling the pipeline, generated visuals are more semantically aligned with the user’s intent—reducing errors in character count, shape consistency, and color fidelity.

Recent updates have also improved in-image text rendering, making it more usable for diagrams, mockups, and stylized educational content.

Web search now functions as a real-time knowledge gateway

Live web results with previews, citations, and smart integration across productivity apps.

The new Web Search tool lets ChatGPT browse in real-time, offering not just links but rich previews with images, summaries, and metadata. Users on Pro and Team plans can go even further: filtering results by domain type (e.g., academic, company sites), or linking results directly to third-party platforms like Canva, Notion, and Google Sheets. This enables seamless workflows such as gathering source material for a report or pulling in recent market data for an investment memo—without leaving the chat.

The system continuously adapts to natural language queries, eliminating the need for search operator syntax and reducing bounce between browser tabs.

Canvas introduces real-time co-editing and project context

An inline editor opens beside the chat, ideal for writing long documents or iterating on code collaboratively.

Canvas offers a new way to work with ChatGPT: instead of writing back-and-forth in pure dialogue, users open a document-style editor beside the chat window. Inside, they can paste drafts, edit code, restructure paragraphs, and annotate with ChatGPT’s assistance. The assistant can highlight weak transitions, fix broken logic, or suggest stronger phrasing—all without interrupting the document structure.

Canvas supports version history, live preview, and an “import from chat” function to bring past conversations into structured form. While still in desktop-only beta, the feature hints at future integrations with IDEs, CMS platforms, and collaborative writing suites.

A unified gateway to a multimodal assistant

This menu redesign is more than a UI update—it reveals OpenAI’s ambition to position ChatGPT as the operating layer of digital cognition.

What was once a question-answering tool is now a modular ecosystem. From the new Tools panel, users can shift between analytical depth (Deep Research), operational autonomy (Agent Mode), didactic structure (Study), visual creativity (Image), real-time facts (Web), and collaborative drafting (Canvas)—all without switching context. This streamlining mirrors a broader move: from AI as a reactive oracle to AI as an interactive, multimodal workspace.

The new menu doesn't just save time. It changes the mental model. The assistant is no longer a passive respondent, but an active partner across research, learning, design, and execution.

____________

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org