How to Use Gemini Deep Research for Thorough Reports

Start a new chat and enable web search

First, let's open Gemini and make sure it has permission to search the internet. An AI (Artificial Intelligence) tool like Gemini is powered by an LLM (Large Language Model) – think of it as the super-smart brain that helps the AI understand and generate human-like text. To perform deep research, this brain needs access to the wider internet.

When you open Gemini, you'll see a clear text box at the bottom of the screen where you type your messages. Look towards the top-right or top-centre of the screen for a toggle switch or button that usually says "Extensions" or shows a puzzle piece icon. Clicking this often reveals a list of tools, including "Google Workspace" and "Google Search." Ensure the "Google Search" extension is switched 'on'.

If it looks different: On some devices or versions, the "Extensions" might already be active or there might not be a visible toggle. If so, don't worry, Gemini often suggests using Google Search automatically when you ask a research question. Just ensure you're starting a fresh conversation, usually by clicking a "New Chat" button (often a '+' icon) in the side menu.

You'll know it worked when: The toggle for "Google Search" appears switched on, usually indicated by a blue or green colour, or Gemini doesn't prompt you to enable search when you type your first question.

Craft a detailed research prompt

A prompt is simply the instruction you give to an AI. To get a high-quality, in-depth report, you need to be very clear with your prompt. Don't just ask a one-sentence question. Instead, tell Gemini who the report is for, what specific areas to cover, and how you want the final document organised. Think of it like giving instructions to a new research assistant.

Type your detailed instructions into the text box. The more detail you provide, the better Gemini can understand your needs and focus its search.

You'll know it worked when: Gemini acknowledges your request (often by saying "Okay, I can help with that...") and begins to indicate it's searching the web or preparing to ask clarifying questions.

Answer Gemini's clarifying questions

One of the cleverest parts of Gemini's deep research process is that it doesn't just guess what you want. It will often reply with two or three clarifying questions to narrow down the scope of its search. This is Gemini using its "brain" (the LLM) to ensure it understands your request perfectly, saving you time later.

Read Gemini's questions carefully. These might be about geographical focus, specific types of technology, or the target audience for the report. Type your concise answers in the text box below.

If it looks different: Gemini might sometimes jump straight into generating the report if your initial prompt was extremely clear. If it doesn't ask questions, that's fine; simply move to the next step.

"Focus primarily on Australian residential properties, especially those in suburban and regional areas, not just cities."

You'll know it worked when: Gemini confirms it has understood your answers, perhaps by saying "Got it," and then transitions to the next stage of gathering information.

Observe Gemini gathering information

Once you've clarified the details, Gemini will begin its thorough search. You'll often see a visual progress bar or a series of messages indicating what the AI is doing in real-time. These messages might pop up within the chat window, showing stages like "Searching for battery lifespans" or "Analysing recycling policies." This is Gemini actively using its Google Search extension to find and process information from various websites, much like a human researcher would.

This step takes longer than a standard AI reply—often up to a minute or two—because the assistant is reading actual web pages rather than relying purely on its pre-existing training memory.

If it looks different: Sometimes, the progress indicators might be subtle, or Gemini might just show a spinning wheel. As long as the chat isn't frozen and you see some activity, it's working.

You'll know it worked when: You see clear messages indicating Gemini is searching the web ("Using Google Search…", "Searching for…"), followed by text beginning to appear that resembles your requested report structure.

Review and verify the generated report

When the report is ready, Gemini will present a structured document, usually complete with clear headings, bullet points, and, crucially, citations or links to the websites it used. This is your chance to put on your detective hat.

Before you copy the text for use, perform a quick spot-check. Click on a few of the provided links to make sure they are active and support the claims made in the report. This protects you against "hallucination," which is the term for when an AI confidently invents facts or links that do not actually exist. While rare with Gemini's search capabilities, it's always wise to verify important details.

If it looks different: Ensure you scroll to the end of the report. The citations are often small numbers or footnotes. If you don't see any, you can politely ask Gemini: "Can you provide the source links for this report?"

You'll know it worked when: You've clicked on a few links and confirmed that the information in the report aligns with the content on the source websites.

Refine or ask follow-up questions

The first report Gemini provides is a great starting point, but you might want to dive deeper or change the focus slightly. This is where the interactive nature of AI truly shines. You can continue the conversation, asking for more details on a specific section, requesting a different format, or even asking Gemini to summarise complex points.

Type your follow-up questions or requests into the text box. This allows you to iteratively improve the research until it perfectly meets your needs.

You'll know it worked when: Gemini provides new information, rephrases existing content, or summarises the report based on your latest instructions, showing its ability to build on previous responses.