What is Google Deep Research?
ChatGPT Search is free, Google's Deep Research feels like a Perplexity agent for CoT Agentic pull requests.
Hey Everyone,
Welcome to the chain of thought (CoT) era of agentic research on the web.
I’m really impressed with Gemini 2’s launch and in particular fascinating by the trio of AI tools known currently as NotebookLM, Learn About and Deep Research. As Google’s models improve these tools can empower Google’s ecosystem and devices in a way that will be hard for others to compete with.
Even as OpenAI is going more into advertising and making Search in ChatGPT free for everyone, Deep Research is what many of us had hoped o1 would be, more accessible and less of a price point issue.
In a nutshell, Deep Research is a CoT tool that does just what you might expect.
Google has added a new "Deep Research" feature to Gemini Advanced that aims to enhance internet research capabilities through its AI assistant.
Given the recent updates to NBLM, we know just what to expect from Google with Deep Research. They will improve the AI tool like a real legit product and seek to bundle their paid sub with more features and AI tools.
For the evolution of the internet, Search with Generative AI will be much better in 2025. Perplexity, Deep Research, o1 and many other upgrades to familiar players. I fancy genspark.AI. Referral traffic from ChatGPT continues to grow every month. ChatGPT search won’t dethrone Google’s search monopoly anytime soon but the Search and research markets are getting more interesting.
This new feature will be available to Gemini advanced subscribers as a new model dropdown called Gemini with deep research.
This tool will get better in 2025, a lot better.
With Deep Research, you can ask Gemini to scour the web on your behalf and write up a report based on its findings.
As it stands today it’s fairly just a CoT prototype a lot like many AI search agents that have been around for many months.
Deep Research is currently only available in English to Gemini Advanced subscribers. If you have access, you can ask Gemini to research a particular topic on your behalf, and the chatbot will create a “multi-step research plan” that you can either edit or approve.
Google is uniquely qualified with all of its search data and analytics to really make such a tool very good, which is likely to happen as soon as 2025. Nobody is going to trust Copilot or ChatGPT to the level as they might Google “Deep Research” eventually.
So for instance this is useful for an initial analysis of a topic: The updated Gemini offers a feature called “Deep Research” that Google says uses “advanced reasoning” and “long context capabilities” to generate research briefs. The briefs are presented in the Gemini apps and can be exported to Google Docs for additional editing. Keep in mind the current product is fairly bare-bones and is not expert level just yet.
Since Gemini 2 isn’t just multi-modal it’s also agentic so we can expect AI tools like these prototypes to get a lot better in 2025 and 2026. This includes how all of these tools work together in Google’s ecosystem, as some of us are lured into using Google workspace a bit more. Google Workspace is essentially Google’s collection of cloud-based productivity and collaboration tools that help people create, communicate, and collaborate that we sometimes take for granted. Google is also trying to find the right upgrades for new kinds of search experiences and human-AI interfaces.
When a Deep Research query has completed, Gemini 2 will product that report of its “key findings” with links to the websites where it found its information. The tool was first previewed at end of Made by Google August, 2024.
The CoT era of agentic AI is here so basically it prioritizes research and accuracy over actually giving you a fast answer, it can take a long time.
How to Use
Deep Research is currently only available for Gemini Advanced subscribers on the web. You can try it by heading to Gemini and then changing the model dropdown to “Gemini 1.5 Pro with Deep Research.”
As ChatGPT races to make search free for all (you still need to sign in), 2025 will make the OpenAI vs. xAI vs. Google era and for Search it will be fairly interesting with even Meta building an in-house Search for its apps. Facebook’s founder is even trying to block OpenAI from going to a for-profit company. This is because of the likelihood that the Search Advertising market is contested. Basically the most lucrative honey pot in the world of the legacy internet.
Google’s Deep Research for now is technically part of the 1.5 Pro model and helps you get research done on a specific topic within minutes rather than hours you’d usually need. But what happens when it’s fully migrated to Gemini 2 and the latest models?
Deep Research is independent from Mariner, NotebookLM, Google About, and its universal-agent Astra among the other various AI tools it’s building and with those future Google glasses (yes they are real).
Basic Query and Fetch User Loop
Here’s how it works:
A user writes a question.
Deep Research creates a “multi-step research plan” for the user to either revise or approve.
Once the user approves, Deep Research refines its analysis over the course of a few minutes — searching, saving potentially interesting pieces of information, and then starting a new search based on what it’s
learned.The process repeats multiple times, and once it’s finished, Deep Research generates a report of the key findings.
However as we have seen with NotebookLM, Google is feisty in terms of customization abilities as they refine a product that gets some traction. It’s too soon to say if Deep Research will get real-world traction just yet.
If Search in ChatGPT is going freemium, this means Ads are coming very soon. This means Google will face its first test in 2025 vs. OpenAI.
Insiders at Google even think Deep Research could impact product research for potential buying journeys (though I find that a bit far fetched). Examples that were given included Gemini deep research helping to plan for home renovation projects, creating reports ahead of the purchase of a new vehicle or even something more complex such as analyzing alternative business propositions ahead of launching a new company. Who knows what we might be depending on AI agents for by 2030.
Product Launch Video
Google is not very good at product-marketing, please excuse them.
“Introducing Deep Research, your personal agentic AI research assistant. Rolling out starting today in Gemini Advanced. With Deep Research, you can create in-depth research reports on complex topics, complete with source links, giving you hours of research at your fingertips in just minutes.”
Google’s Ecosystem
Google Search and Google Search Generative Experiences get stronger with these auxiliary but related tools like Deep Research, NotebookLM, Google About and more.
In 2025 with Perplexity evolving, Google improving and Search in ChatGPT, we’ll just have a better search experience overall. Companies like Twelve Labs should also improve Video search soon. Google acquired a lot of the top talent the immediate access to the models of Charater.AI so it will eventually fuse that into their product (some pending lawsuits).
OpenAI’s ChatGPT base of 300 million weekly users is only going to get bigger. So the likes of Bing (and Microsoft’s Copilot) and others might struggle in comparison. OpenAI historically is likely good for Google to actually innovate under some hypothetical pressure.
As of mid December (the time of writing), both the experimental version of Gemini 2.0 Flash and the Deep Research version of Gemini 1.5 Pro are now available for Gemini Advanced users.
I think Deep Research has higher ceiling than many of you realize. If ChatGPT search adoption accelerates, and if Perplexity’s product keeps improving for white-collar work research, things will get really interesting. OpenAI thinking it’s a great idea to pay $200 or one day $2,000 a month subscriptions is pretty silly compared to how Anthropic, Google and many Chinese players are evolving their models.
The whole appeal of Google was that it was free and the most accessible AI tool of all-time.
Google released the first prototype of Deep Research to the general public on December 12th, 2024.