Deep Research

Deep Research is Copera's AI research assistant that runs multi-step investigations on complex topics and returns a cited report inside your AI Chat conversation. It breaks your question into sub-questions, searches multiple sources, and writes a structured report with inline citations you can verify — replacing hours of manual research with a single prompt.

Unlike a regular AI response, Deep Research works through several research iterations, similar to how a human analyst would approach a topic. Every claim in the final report includes a numbered citation that links back to the original source.

When to Use Deep Research

Deep Research is ideal for:

• Market analysis — "Research the current state of the remote work software market, including key players, trends, and growth projections"
• Competitive research — "Compare the top five project management tools for teams of 50-100 people"
• Technical investigation — "Research best practices for implementing single sign-on in enterprise applications"
• Policy research — "Summarize the latest data privacy regulations affecting companies operating in the EU and US"
• Literature review — "What does current research say about the effectiveness of asynchronous communication in distributed teams?"

For quick factual questions or simple tasks, regular AI Chat responses are faster and more appropriate.

Starting a Deep Research

1. Open AI Chat from the left sidebar.
2. Click the Deep Research chip in the composer toolbar. The chip highlights to show that Deep Research mode is active.
3. Type your research question. Be as specific as possible about what you want to learn.
4. Press Enter or click Send.

tip

You do not need to say "deep research" in your message. Enabling the chip is enough. Write your question naturally, as you would ask a human researcher.

The Research Plan

After you send your message, the AI analyzes your question and creates a research plan. This plan appears as an interactive card in the conversation.

Plan Card Overview

The research plan card shows:

• Title — A concise title for the research run, generated from your question.
• Sub-questions — A numbered list of specific questions the AI will investigate. These break your broad question into focused, researchable angles.
• Depth tier selector — Three options that control how thorough the investigation will be.
• Action buttons — Start the research or cancel it.

Editing the Plan

Before the research starts, you can customize the plan:

Edit sub-questions:

• Click any sub-question to edit its text inline.
• Press Enter or click away to save your edit.
• Click the + button to add a new sub-question.
• Click the X button on any sub-question to remove it.

Choose a depth tier:

• Click one of the three depth options to select it (see Depth Tiers below).

The auto-start countdown pauses while you are editing, so you have time to make changes.

Depth Tiers

The depth tier controls how thorough the research will be. Select the tier that matches the level of detail you need.

Tier	Description	Best For
Standard	A focused investigation across the main angles of your question.	Quick overviews, time-sensitive questions, straightforward topics
Deep	A more thorough exploration with extra iterations and source filtering.	Detailed analysis, topics requiring multiple perspectives
Very Deep	An exhaustive investigation with recursive follow-ups and source curation. Takes significantly longer.	Comprehensive reports, academic-style research, critical decisions

warning

Very Deep research takes significantly longer than the other tiers. Only choose this option when you need an exhaustive investigation. A warning message appears when you select Very Deep to confirm your choice.

Starting the Research

You have two ways to start the research:

• Click the Start button (checkmark icon) at the bottom-right of the plan card to begin immediately.
• Wait for auto-start — If you do not make any changes for 60 seconds, the research starts automatically with the current settings. A countdown timer shows the remaining time, and a progress ring around the Start button visualizes the countdown.

The auto-start countdown pauses whenever you are editing a sub-question and resumes when you stop editing.

Cancelling

Click the Cancel button at the bottom-left of the plan card if you decide not to proceed. The AI will respond without conducting the deep research.

During Research

Once the research starts, the plan card transitions to a read-only state showing "Research started — we will notify you when it is ready." A progress indicator with a live timer shows how long the research has been running.

Working While Research Runs

Deep Research runs in the background. You can:

• Navigate to other parts of Copera and continue working.
• Open other AI Chat conversations.
• Close the conversation and come back later.

You do not need to keep the conversation open for the research to complete.

Notifications

When your research finishes, you receive a notification through the bell icon. Click the notification to jump directly to the completed research in your conversation. The notification also arrives as a push notification if you have them enabled.

Research Results

When the research completes, the plan card updates to show completion statistics:

• Duration — How long the research took (for example, "Research completed in 3m 42s").
• Citations — The number of unique sources cited in the report.
• Searches — The number of searches conducted during the investigation.

Below the plan card, the full research report appears as a regular AI message. The report includes:

• Structured sections — The report is organized with headings, subheadings, and paragraphs for easy reading.
• Inline citations — Numbered citation badges appear throughout the text. Hover over a badge to see the source website. Click the badge to open the source in a new tab.
• Sources panel — A panel at the bottom of the response lists all sources referenced in the report with their URLs.

Following Up

After the research completes, you can continue the conversation to explore the results further. Ask follow-up questions like:

• "Can you go deeper into point 3 about market trends?"
• "Summarize the key findings in a table format."
• "What are the main takeaways for our team?"

The AI retains the full context of the research, so follow-up questions benefit from all the information gathered during the research run.

Research That Fails

If the research encounters an error, the plan card shows a "Research failed. Try again." message in red. You can start a new research by sending another message with Deep Research mode enabled.

Common reasons for failure include network issues during source fetching or the research taking longer than expected. Simply trying again usually resolves the issue.

Tips and Best Practices

tip

Write specific, focused research questions. "Research the impact of AI on customer support response times in SaaS companies" will produce a better report than "Tell me about AI in business."

tip

Review and edit the sub-questions before starting. The AI generates a good initial set, but you can add angles it missed or remove questions that are not relevant to your needs.

tip

Start with Standard depth and escalate to Deep or Very Deep only if you need more detail. Standard research is faster and often sufficient for most questions.

tip

Use follow-up questions to dive deeper into specific sections of the report rather than running a new research from scratch. The AI remembers the full context of the completed research.

tip

The Deep Research chip automatically toggles off after you send a message, so it does not accidentally trigger on your next regular question. Enable it again for each new research request.

Frequently Asked Questions

How long does Deep Research take?

It depends on the depth tier and the complexity of your question. Standard research typically completes in 1 to 5 minutes. Deep research may take 5 to 10 minutes. Very Deep research can take 10 minutes or more. You will be notified when it finishes.

Can I run multiple Deep Research sessions at the same time?

No. Only one Deep Research session can run at a time. While a research is in progress, the Deep Research chip is disabled and shows a tooltip explaining that a research is already running. Wait for the current research to complete before starting a new one.

Can I edit the plan after starting the research?

No. Once the research starts, the plan card becomes read-only. If you need to change the research direction, wait for the current research to complete and then ask a follow-up question, or start a new research with an updated question.

Are the citations reliable?

The AI searches real web sources and includes citations to the original pages. You can verify any citation by clicking the badge to open the source URL. As with any AI-generated content, it is good practice to verify critical claims against the original sources.

What happens if I close the tab while research is running?

The research continues running on the server. When it finishes, you will receive a notification. Open the conversation again to see the completed report.

Can I use Deep Research with workspace context?

Yes. You can attach workspace documents or Drive files using the context picker before sending your research question. The AI will consider both your workspace context and web sources when conducting the research.

Related Features

• AI Chat — The conversational AI interface where Deep Research runs
• Image Generation — Generate images from text descriptions
• AI in Documents — Use AI directly inside the document editor