The entire article is also available as a podcast episode:
Attention! The podcast was created entirely by my AI-Assistant based on my contribution – no guarantee for incorrect content.
What does deep research mean?
Deep research refers to the ability of AI chatbots, such as ChatGPT, Perplexity or Gemini, to understand complex search queries, break them down into several research tasks, search the Internet independently and consolidate the results in a structured manner. Instead of just listing links to websites – as we are used to from Google – the systems analyze the information, synthesize it and provide the user with a comprehensive answer in the form of a clearly written report.
How can I test and use deep research methods?
The easiest way to test and use the new deep research functions is via the existing AI providers. However, as the Deep Research function requires significantly more computing capacity, this function is currently only available in the paid accounts.
Here are a few examples:
- Perplexity (Pro, 20€/month) -> Start Pro search with R1 model
- ChatGPT (Pro, 200€/month) -> click on the “Deep Research” button in the chat
- Gemini (Advanced, 20€/month) -> select “1.5 Pro with Deep Research” model
How does deep research work?
The in-depth research or deep research method always involves 4 steps, whereby the providers differ slightly (as of 02-2025).
- Planning: The AI processes the search task and independently plans the search process and search queries.
- Information search: The AI searches through numerous sources such as articles, reports and studies and filters out irrelevant information. OpenAI uses web browsing functions, while Gemini relies on Google services.
- Analysis: The AI then “reads” all the collected texts, extracts important facts, compares sources and recognizes contradictions.
- Structuring and preparation: Finally, the findings are presented in a clearly structured report, usually with an introduction, main section and conclusion. Important points (e.g. pros/cons) are highlighted, and references at the end ensure transparency.
Important: Even with source-based research, language models can hallucinate. As always, AI should be used as support and not as a blind substitute. The principle of result control applies in particular to critical facts.
For which applications can I use Deep Research?
Although the deep research function sounds very tempting at first, it is not useful for all AI chatbot applications. This method also requires significantly more computing power and energy, and these should only be used when they are actually needed.
Basically, deep research can help wherever many sources are searched for information and the results need to be put into a structured format.
- Topic research: Creation of a structured report on a topic, e.g. “Introduction to AI agents”
- Technology and innovation scouting: Identification of emerging technologies (e.g. quantum computing, mRNA technology) by analyzing news, blog posts and patents, provided the latter are publicly accessible.
- Trend analysis: Identification of new nutritional or lifestyle trends (e.g. veganism, zero waste).
- Product search and comparison: Search for suppliers in categories such as e-bikes, 3D printers or new smart home systems.
- News overview: Compilation and consolidation of news on a topic.
- Market and competition analyses: Market overview in the medical technology sector in Germany, strengths and weaknesses of the main competitors, key figures, product portfolio, …
- Scientific investigation: Compilation of freely available studies on the topic of “Effects of air pollution on health”.
Practical examples of deep research
Below I will show you how it works using 3 examples with three different providers. I have used simple prompts in all examples to understand how the reasoning models plan the research process.
Example 1 (Perplexity): Quick market analysis
Goal
Get an overview of a specific market or provider.
Prompt
Create an overview of the largest AI language model providers including number of users, feature comparison in tabular form, funding and other relevant information.
Process
In the Research section, you can see the search queries that Perplexity performs to collect the information. In the Considerations section, you can see how the R1 model processes the information to create the result.
Also impressive: a total of 86 sources were searched.
 |
Result
First, a general overview of the top LLM providers was compiled. I find it exciting that very useful criteria such as “Parameters” or “Open Source?” have been added. The problem is that GPT4.5, for example, does not exist, but is listed as a result.
 |
A table with the feature overview was then created. I still find this quite simple, but it could be expanded with a more precise prompt. In addition, the results are not consistent or incorrect: Claude 3.5 Sonnet is heavily used in creative functions and marketing, for example.
 |
Example 2 (Gemini): Topic research on deep research
Goal
Create a report about the Deep Research feature of Google Gemini.
Prompt
Create a report on Gemini’s Deep Research feature with the following content: Definition, functionality, potentials & risks, alternative providers, use cases in knowledge functions.
Process
Gemini first processes my request, creates the research plan including the search queries and asks me for feedback. I now have the option of adapting the plan or starting the research.
 |
Based on the research plan, Gemini searched 22 websites.
 |
In the final step, Gemini consolidates the information from the websites and creates a report. This step took about 5 minutes.
Result
I receive a complete report with the exact structure specified.
 |
I can copy the text or open it directly as a Google Doc. It was noticeable in this example that Gemini strongly favored Google sources. Of course, this may be due to the task, but it also indicates a bias in the model.
Example 3 (ChatGPT): Trend research on AI agents
Goal
A trend report on the development of AI agents and their impact on companies.
Prompt
This time I have made the prompt a little more detailed and added more context and format specifications:
Create a trend report on the development of AI agents. The target group are managing directors in medium-sized companies.
Structure the report as follows:
- Executive Summary
- Introduction to AI agents (definition, mode of operation, …)
- Development of AI agents (history, status quo, future forecasts)
- Fields of application
- Opportunities, risks and recommendations for action
Clarify the following questions in particular in your report:
- To what extent are AI agents already ready for use in SMEs?
- How does the use of AI agents affect the productivity of companies?
- What foundations need to be laid to successfully implement AI agents?
Critically examine AI agents and refer to studies, statistics and sources from thought leaders and scientists from the AI industry, such as Andrew Ng.
Process
ChatGPT first asks questions to refine the search:
 |
After I answered the questions, ChatGPT started the research process. The research took about 5 minutes and included 14 sources:
 |
I find the selection of sources very sobering. Almost exclusively blogs were used and hardly any study results or statistics.
What I find exciting is how you can observe the work steps of ChatGPT live:
 |
Result
I also receive a report from ChatGPT, which at 15 pages is even more extensive than the result from Gemini. I used the o1 Pro model, which further enhances the quality and depth of the results.
 |
Overall, the quality convinces me when I want to get an overview of the topic and the historical developments. I find the writing style amazingly good.
What I am missing are concrete study results and statistics that go beyond general forecasts on management consultant blogs.
It is striking that the majority of the information comes from a few referenced sources, although a total of 14 sources were searched.
Conclusion: Deep research in application
Deep research features enable AI chatbots, such as ChatGPT, to understand complex search queries, divide them into several research tasks, search the Internet independently and summarize the results in a structured manner. Reasoning models, which convert unstructured information into coherent reports, are an ideal complement to this. These models already break down simple prompts into multi-level search queries, but deliver more precise results if additional context such as the target group or desired format specifications are specified.
In a comparison of different AI chatbots, Perplexity searched most sources, but only provided short answers and sometimes contained misinformation. ChatGPT in o1 Pro mode performed best in terms of quality, but was not significantly better than Gemini in relation to the price.
The areas of application of deep research range from topic research to market analyses, with AI-supported research offering significant time savings in the structured preparation of information. Nevertheless, there are risks, including distortions in the models and the danger of relying too heavily on AI without thinking critically about the issues yourself.
Personally, I find it very interesting when AI chatbots like ChatGPT reveal the “thought process” they use to process my prompt. In most cases, however, I am disappointed when I look at the quality of the sources searched. In most cases, few scientific sources and many less trustworthy sources are used. And although the “thought process” sounds so genuine and sensible, the AI chatbots still make significant errors and the deep learning methods also require the results to be checked carefully.
Further questions
Has this article given you food for thought and do you have any further questions? Or are you looking for general support with the use of AI, ChatGPT, Deepseek and chatbots?
I am always happy to receive your messages, preferably by WhatsApp message or e-mail.
This article is also available as a podcast episode
Attention! The podcast was created entirely by my AI-Assistant based on my contribution – no guarantee for incorrect content.
