About OpenAIDeepResearch, and I have some of my own uses and things I want to say about it, so why don't I just do one and come and talk to you about it.
This very awesome, even to the point of making me think it's really worth $200 a month, feature.
DeepResearch.
I'm also going to spend almost 1w words on an article with 10 scenes, and those 10 scenes with thePrompt word, to give you a detailed look at where he can be used, how it should be used, and the last scene, which will surely surprise you, even amaze you.
The article is very long, can not finish can be recommended to collect first, later slowly read.
However, before I look at these 10 scenarios, I think it's important to give a quick and simple talk about what OpenAI, the DeepResearch, is all about.
If you already know in detail or have used it, just skip this party and read backwards.
Zero. What is DeepResearch?
DeepResearch is an Agent feature on ChatGPT launched by OpenAI, in February this year. The Chinese translation is DeepResearch.
Available for Pro and Plus members, 150 visits a month for Pro members and 10 visits a month for Plus members. No quota for free users.
The role is simple, to look deeply into the question you're asking, then spend 10 to 30 minutes searching the entire web for data, and then give you a detailed and very insightful presentation report.
Essentially, it's just an end-to-end Agent model based on o3 fine-tuning that follows a planning - execution - synthesis path to the task, and will also dynamically adjust the task as it goes through the process.
And the quality of the report is about equal to the quality of a report generated by a researcher veteran, working 10 hours to a week.
He is best suited for tasks that involve integrating multiple sources of information, in-depth analysis of complex data, creation of well-documented reports, a multi-step research process (involving planning, finding, navigating, reasoning, analyzing, and synthesizing), and dealing with comprehending and reasoning with large amounts of information.
That, in a nutshell, is DeepResearch.
Next, I'll take you through 10 typical DeepResearch scenarios to show you, how it's used, along with a Prompt for each scenario that I think is great.
I. Analysis of market competition
Before Deep Research came along, it was often painful for many entrepreneurs and market analysts to conduct competitive analysis.
You need to manually collect information, news reports, user reviews, etc. of N competing companies, browse hundreds of web pages, open countless browser tabs, read countless sub-reports, and then use your own human brain to summarize and organize the key points.
With Deep Research, all you need to do is put forward your research requirements (e.g., "Analyze the strengths, weaknesses and market share of major competitors in China's online education industry"), and leave the rest of the work to the AI agent.
It automatically searches numerous sources (official websites, news, market reports, user reviews, etc.), focuses on the areas you specify, and consolidates the information in a single conversation to come up with a coherent analysis.
Entrepreneurs can quickly understand the industry landscape and find market gaps; product managers can gain insight into the strengths and weaknesses of competing features and optimize their own product roadmaps; and analysts can more efficiently collect data for reports.
White-collar workers no longer need to stay up late to do competitive analysis, to save time can sleep more lazy.
Here, I've also put together a Promp template that I think is pretty good for myself, with the blue part changed to your own:
Please help me analyze the current major competitors in the industry. Specific requirements:
1. List at least major competing companies and briefly describe their product positioning and market share.
2. Analyze the of each competitor (in terms of features, price, user reputation, etc.).
3. Combine the above information to give opportunity points and differentiation suggestions in the market.
For example, I could change him to a research PROMPT for the milk tea industry.
The front of the report results look like this:
In 10 minutes, a report covering an overview of the competition, a comparison of strengths and weaknesses, and strategic recommendations, is done.
II. Review of academic literature
Some of my buddies who are doing research told me that the most time consuming part of their project was theLiterature Review.
It is often necessary to read dozens of papers, find various literature reviews and data reports to understand the development of a particular field of study.
Human searches are not only inefficient, they are also easy to miss, perhaps taking days to go through the literature only to realize that a key paper has been missed. Moreover, the literature is often full of difficult terminology, and it takes brain power to digest and understand it.
Now, you only need to enter a research topic, for example, "Review the research progress of deep learning in the field of medical imaging diagnosis in the past 5 years", Deep Research will automatically retrieve relevant abstracts, academic articles, academic blogs and other sources, and integrate the scattered information into a structured review report.
In addition to listing important research findings and conclusions, the best part is that it will beAttach source citation, making it easy for you to track down the original paper.
Prompt template here.Change the blue part to your own:
Help me research: .
The report needs to contain:
- Important in this field (author, year, brief description of results).
- of different (e.g., accuracy, data requirements, interpretability).
- Current Research Recommendations.
Please provide details and cite sources.
Deep Research output report:
Final Literature Reference:
I've also looked at them one by one, and they do exist in the literature.
III. Equity investment research
If you are in the investment business, or in the financial industry, you certainly know that there is just too much information, which is why many people say that they are looking at the long term and going for the noise.
After all, there is so much information out there that researching industry reports, company earnings reports, news announcements, analysts' research reports ...... any one of these details could affect your investment decision.
In the past, to study a stock or an industry, you often had to spend several days going through thick annual reports, checking financial data, collecting market news, and then summarizing and analyzing them manually. Lagging information or mistakes may bring losses.
Ordinary white-collar workers who want to do their own investment research, I'll be honest, the difficulty is really hard as hell, from the complexity of the information to sort out the clues, the energy and time requirements are really too high.
With DeepResearch, you can have it research a company's operations and investment outlook, and it will grab information from multiple sources such as earnings summaries, news reports, and industry analysis.
For example, type in, "Research Tesla's current financial health and future growth points," and Deep Research can automatically summarize key financial metrics from recent quarters, analyze the highlights of Musk's statements on earnings calls, extract Wall Street analysts' opinions (e.g., investing vloggers' comments on X), and so on and so forth. There's even the various sides of the political spectrum he's caused to upset people.
The final output includes both data (e.g., revenue growth rate, change in profitability) and qualitative analysis (e.g., market share, competitive advantage) in an analytical report.
For macroeconomic or industry trends, Deep Research can also integrate multiple forecasts to give you a quick overview of the world for yourself.
It really is no longer a fantasy that everyone has analyst-level intelligence available to them.
I myself recently super doubled down on a certain ETF that I actually had Deep Research pick out after analyzing it.
Prompt template here, change the blue part yourself:
Please help me with the following task using Deep Research:
Research the value of the investment in <XYZ Technologies, Inc. Report needs to contain:
- Overview of financial performance (revenue, net profit and growth rates) and analysis of key drivers for the last .
- for the industry in which the company operates (list major competitors and market share).
- (e.g., stock price target or growth expectations) from authoritative analysts or media and rationale.
Please provide data and sources to support the above analysis and give your overall judgment.
I'll analyze Alibaba with him and see.
IV. In-depth examination of historical events
Anyone who needs to write as an author, screenwriter, journalist, etc., will probably need to consult a lot of historical sources and articles when writing about some historical event or person.
Taking a famous historical event, it may involve newspaper reports, memoirs, academic papers, etc. from multiple eras. The information is scattered in different eras and carriers, making it difficult to collect. To piece together the full picture of the event, it is even more necessary to patiently comb through the timeline and distinguish the credibility of different sources.
But today, whether you're examining an unsolved case or writing a biography of a historical figure, just ask your research question and AI will help you find the clues in a sea of internet and digital libraries.
More broadly, the public's perception of history will be richer and more accurate because AI can help us piece together fragmented historical information and avoid biased listening to a single source.
Although the illusion of AI will still exist, and finally get the authenticity of historical information needs to be screened by people, but Deep Research has dramatically improved the efficiency of information access.
As some have said, AI has opened the doors of time to a wider audience, and information that was previously only accessible to professional historians can now be explored by the general public.
Prompt template here:
I want to research this historical event in depth. Please help me:
- List in chronological order the main stages of the event, briefly describing what happened at each stage and providing time dates where possible.
- Quote at least
Evaluations or accounts of the event (e.g., newspapers of the time, writings of historians, memoirs of those involved) with attribution.
- Synthesize the information and analyze the impact of this event on the then
The impact of the
For example, I want to examine the story of David and Goliath.
You'll find that the story of Goliath seems to be more than just a myth inside the Bible, and may actually be in history, actually happened, only, it's gone through many rounds of embellishment.
V. Fact-checking and rumor-shattering
Frankly speaking, in this age of information explosion, all kinds of rumors and "headline party" are endless.
Fact-checking is an important but difficult task for the media, science writers, or many people who wish to seek the truth. To confirm or refute a claim, it is often necessary to go through a large amount of information, find authoritative sources to support it, and sometimes do experiments yourself or consult experts.
Fact-checking has now become more efficient than ever. GiveDeep Research a to-be-validated statement that it canSearch for both supporting and opposing evidence and present them to you.
For example, you ask, "Is it true that microwaving food for a long time is bad for your health?" .
Deep Research searches the scientific literature, food safety agency announcements, popular science articles, and so on. On the one hand, it may find a statement from WHO or FDA stating that microwave ovens are safe, on the other hand, it may cite sources of small studies or rumors, and then give you a comprehensive judgment: most likely, it will tell you that "there is no reliable evidence that microwave heating causes cancer" or something like that, with key sources.
This way, you not only know the answer, but you can see the chain of evidence behind it.
For more serious investigations, such as news fact-checking, it can also help to list time, place, and character evidence, allowing you to quickly restore the facts.
This usage is significant for journalists, science writers, content reviewers, and more.
Prompt template:
I want to verify a claim:
Is true?
Please help me find and research:
- Basis for supporting this statement (list relevant studies or expert opinion if available).
- The basis for opposing this statement (conclusions given by official health organizations, epidemiological studies, etc.).
- Give a final based on the evidence found (e.g., largely untrue, insufficient evidence, or true).
Please provide a link to the source to prove it.
If I change the question to: does drinking coffee cause osteoporosis, this report looks like this.
Wouldn't it be nice if we could all speak with the truth.
VI. Individual Learning Route Planning
We often face the dilemma of information overload when we want to learn a new skill or knowledge.
Take learning Python programming for example, online tutorials are all over the place, there are free video courses, blog articles, book recommendations ...... where exactly to start making it difficult.
If you just follow your feelings, you may get stuck after learning some skinny stuff, or the information is not systematic leading to knowledge gaps.
Customizing a systematic study plan for yourself is not an easy task, and requires knowledge of what are the necessary foundations, classic textbooks, and practice paths in the field, which is often difficult for beginners to manage.
But for Deep Research, it will synthesize the countless learning resources on the Internet and the experiences of those who have been there to give step-by-step advice.
For example, "I am a marketer and want to teach myself data analytics in 6 months, help me plan my learning path."
Deep Research might recommend: learning Excel and basic statistics in month 1, Python data analysis libraries (pandas, etc.) in month 2-3, practicing a few small projects in month 4, and advancing to machine learning fundamentals in month 5-6, and recommending appropriate course or book titles for each stage.
Some enthusiasts had put together similar study route blogs, and AI will refer to these and then tweak them with your background. The result is a highly personalized study plan that takes you down the road.
In the future, there may be new roles such as "AI Learning Coach" and Deep Research is what I think is in the making.
Promp template here (this will have to be overhauled for your specific situation, I'm just offering a thought):
Background: I am a fresh graduate of who wants to switch to software development but has no programming foundation.
Objective: To achieve the goal of
Competent junior software engineer within.
Please design a study route for me, taking into account my background and strengths:
- List the learning priorities (e.g., programming fundamentals, algorithms and data structures, project practice, etc.) for each phase (e.g., every 2 months).
- Recommend (books, online courses, practice projects, etc.) for each stage.
- Provide some study tips or notes.
In the future, not knowing how to start learning really won't be an excuse anymore.
VII. Social Opinion and User Sentiment Analysis
In the marketplace, brand publicists and marketers need to keep an eye on their company's or product's reputation among the public.
Traditionally, people use opinion monitoring tools to grab social media and news, butQualitative analysis many times still has to be done manually or with partial AI assistance.
For example, when a new product is released, there are thousands of user reviews, and it's not easy to summarize the main positive and negative points. Even for the direction of public opinion in a hot event, you need to read a lot of microblogs and forum posts to get a sense of people's emotions.
Doing this manually is time-consuming and prone to subjective bias.
And with Deep Research, you can just one sentence:
"What are the recent online reviews of our XX product?" , which searches social media posts, product reviews, and relevant news comments to distill common opinions and emotional tendencies.
PR teams can understand the pulse of public opinion more quickly and adjust their strategies in a timely manner; marketers are able to discover users' real pain points and preferences, which can be used to improve products or develop promotional programs.
The voice of each user will not be drowned out, and truly "user-centered" decision-making.and not just patting your head on the back of your feelings.
Template Prompt:
Help me analyze the user opinion of in detail:
- Aggregate the main positive comments (what you like) and negative tweets (what you don't like) about from users on social media.
- Statistics on the approximate percentage of positive and negative ratings to analyze the overall user satisfaction tendency.
- Quote a few representative user comments (with attribution, such as tweets or forums).
- Give suggestions for marketing or product improvements for based on feedback.
For example, I asked it to analyze the public opinion after OpenAI released GPT 4.5.
It's really very accurate.
VIII. Product Comparison
I don't know about you guys, but I have trouble choosing myself.
If you want to buy a cell phone, you have to compare the parameters and reviews of different models; if you want to choose a host computer, you have to look at the differences in features, prices, and user feedback.
In the past, we often needed to open two dozen browser tabs or swipe through Little Red Book for hours, reading review articles, user review posts, and then making our own notes to compare.
Information scattered everywhere and mixed views, tossing a few hours after the neck to break before making a decision, finally made up his mind to click on the payment, backhanded and swiped it on the small red book to avoid the lightning post.
Deep Research, in addition to the normal retrieval of major tech media reviews, user review video summaries, forum word-of-mouth and other information, will be tailored to your information.
For example, if you are particularly concerned about the camera effect, it will focus on comparing the camera ratings; if you are concerned about the price, it will calculate the value for money of each model.
It's really still useful.
Prompt template:
Help me compare the two products:
I'm thinking of purchasing and .
Demand:
- I travel a lot and need laptops with , light weight>, and am also concerned about .
Please compare A and B in terms of specifications, battery life tests, user reviews, etc., list the pros and cons of each and give a recommendation based on my needs.
For example, let me compare Macbook Air and Xiaomi laptop.
What's most interesting is that Deep Research actually captured information about the state supplement, and that's pretty strong.
IX. Integration of public information
Nowadays, there is a sea of news coming out every day, and it's hard for busy white-collar workers to keep up with it all.
Especially when you need to understand the whole picture of a hot event or industry news, it is even more difficult. For example, on the topic of "AI film and television", related news is scattered in technology, finance and economics, and even entertainment boards, to get a comprehensive perspective, you may have to read dozens of reports, turn over the microblogging hot search, read industry reviews, very time-consuming.
And even if you invest the time, it's easy to get into information overload and not know what's reliable information and what's just noise.
And Deep Research can be called aSuper Editor. It quickly crawls the Internet for news, blogs, and social media discussions related to a topic, theFilter duplicate content and distill key information, and culminate in an overview summary report.
Even for hot events (e.g., a corporate takeover, a large-scale accident), it can be able to organize thetimeline, list the context and key points of the event.
Prompt template:
Help me summarize "The latest in the global AI industry."
- Sub-topics covered include: recent news from major vendors (new product releases, earnings reports), major hot AI product releases, works that break the circle, AI-related social opinion, and more.
- Please summarize by topic and provide specific facts and figures (e.g., a company's revenue growth X%, a country's policy points, etc.), listing the sources.
- Finally, it gives an analytical judgment of the AI industry trends for the next 6 months.
The report is very interesting:
Caught tons of domestic AI product developments on product launches.
Even The Heartland Ridge Creeps, arguably a landmark AI sketch in this country, was caught out.
This is really a milestone, because it marks, for the first time, AI sketches have achieved profitability, although it does not earn much, but it also represents, the business model runs through, this is the sketching industry opened up a big event.
But I'm sure that probably a lot of people in the AI industry are completely unaware of this show.
Deep Research, is already more comprehensive than the vast majority of practitioners, in terms of breadth of knowledge.
X. Writing a novel
Maybe I'm one of the few that's writing a novel in Deep Research.
The better the story, the stronger the demands, the more logic is required. There is a very classic theory, especially in movie and TV scripts, called Chekhov's Gun.
Presumably, this means that every element mentioned in the story should make an appearance later in the text, otherwise there's no point in mentioning it:
"Please remove everything from the story that has nothing to do with the story. If you say there's a gun hanging on the wall in the first act, then that gun has to fire in the second or third act or there's no point in hanging it there."
The same goes for AI story writing, which is not a random stroll, but an eureka moment where logic and creativity collide together.
andDeep Research's pedestal o3, in terms of logical reasoning, is really strong as hell.
It's also one of the few novels that can generate tens of thousands of words of logical coupling at once.
I'll show you my stream, a place where it's not really possible to make a template because it's too personalized.
I write novels in two steps, step one: find information and write key plot points, and step two, write the complete novel.
My Prompt for the first step is as follows:
I'm working on an idea for a shelf historical fantasy novel set in a setting similar to the European Renaissance of the 15th-16th centuries, but I'd like to add a bit of alchemy and magic elements. Please use Deep Research and help me with the following tasks:
1. Culture and periodization:
- Find the main cultural features, social structure (status of nobles vs. commoners, urban guild system), and patterns of daily life (clothing, food, etiquette, etc.) of the Renaissance.
- I'd like to add the supernatural setting of "Alchemy", so please search for real historical texts or characters related to alchemy (e.g. Paracelsus) and see what role they played in actual history.
- Also look up important events (e.g., major wars, dynastic changes) and technological advances (e.g., musketry, development of printing) that took place in Europe during this time period and list the key years or milestones that I can refer to.
2. Inspiration for the magic system:
- Search folklore or occult literature for magical or witchcraft material that can be incorporated (e.g., sayings about magical formations, runes).
- Researching the attitudes of religions of the time (e.g., Catholicism, heresy trials) toward the "paranormal" helped me to rationalize the conflict between magicians and the Church.
- If there are similar fantasy novels or movie references (e.g., the worldview of the Witch Hunter, Wizard's Gate series), briefly describe how they deal with the "medieval + magic" setting.
3. Character and plot suggestions:
- I would like the protagonist to be a young alchemy apprentice just starting out, please search for any relevant historical figures or legends that could provide inspiration, such as real-life apprenticeship training processes or anecdotes of famous alchemists.
- Based on the above information, suggest a general story line: include how the protagonist researches alchemy under the patronage of the court or nobility, how he confronts the church or hostile forces of another country, etc. Could help me sort out 5-8 key plot points.
- Provide some possible emotional conflicts (e.g., a noble lady falling in love with an alchemy apprentice, an undercurrent of struggle within the Church, marriages or wars between different countries, etc.), and cite similar court events in real history for reference.
4. Notes:
- Cite the primary historical or fantasy source used; if there is a specific year or person, please cite the source for easy cross-checking by me.
- Please mark anything about magic or alchemy as "speculative" if you don't have enough historical facts to support it, so I can adapt it later.
Output format:
- Start by giving me a condensed version of the worldview background summary (with key settings I can use directly).
- This is followed by a detailed list of references or inspirations for the historical elements of the Renaissance and the elements of fantasy and magic, and a chapter/plot outline based on my needs.
- Finish with a hint of where the possible ending is going or an ambush.
Some of these references and reports in the final output are so awesome and so long that I'm not going to put them all in, I'm going to put some of them in for you to see.
Even if it's a short story, it should have a lot of worldview and historical research, and landing yourself in that era, turning into a cloud of dust to see everything about that world first hand is, well, the foundation of a story.
That's what makes, logic and facts, your story, alive enough.
Once I have that material, I can then talk to theDeep Research said, based on the information in the report above, help me write this 3w word short story, just refer to the style of the Steel Alchemist, and you can incorporate a little bit more of Cthulhu's kind of uncanny fear as a darker line of the story.
After a full 32 minutes, this 6-chapter short story, was finally finished.
I don't know what words to use to describe this one.
This is a, like, 3w word novel.
I'm extremely picky when it comes to written content, especially AI-generated content, and there's almost no fiction that I can read knowing it's AI, but I'm still reading it.
But what Deep Research writes, and I can read it down every time, and it's the kind of reading that, that pulls me down, is that I can really feel the city breathing, feel the characters, the characters' struggles, the characters' growing up in the lone light, and the kind of despair and struggle of the little people underneath the times, underneath the behemoths that are the church.
The ambush from earlier in the book is still useful a few chapters later, as it turns out! It turns out that the element that appeared earlier is so useful! So it was him! That was the exclamation that happened while I was reading.
It's excellent storytelling, it's perfect storytelling, it's well-crafted and designed storytelling, and I just really, really, really, really love it.
The complete full story, because it's so long, I've put it in the Flying Books document, so if anyone wants to read the novel, they can move over and watch it.
https://datakhazix.feishu.cn/wiki/Tsl1wGzr0iKSYJkh4y9c4MWQnvV?from=from_copylink
It's really too good looking.
Final Thoughts
This article, even after I've been using it for a long time forDeep Research is more familiar to write about, and still feels that it can only show the tip of the Deep Research iceberg.
There may be more, more interesting ways to play the game, buried deep down, on territory that I have yet to uncover either.
More and more, I'm getting the impression that bullsh*t AI is like an ocean.
I put my best foot forward and could only manage to get a glimpse of it.
As you learn more, you will only feel a growing sense of wonder as to why you know so little.
One's own life in exhaustion may never again catch up with the pace of AI evolution.
People in heaven and earth, but as small as a mayfly, watching the sky at first light and then dark, it is already a life and death.
Having been born in this age, what then.
Run and study as hard as you can.
On this path of trying not to be outdone by the times.
END.
