Summary
Google’s long reign as the go-to search engine took a hit from AI challengers like ChatGPT. Now, Google’s striking back with Gemini Deep Research—an advanced, paid AI tool that may just reshape how we find, learn, and shop online.
What Is Gemini Deep Research and How to Use It?
At the time of writing, Deep Research is currently available to paid plan users through the web version of Gemini—it isn’t currently available with the mobile app versions. You can access it through the model selector in Gemini Advanced by choosing “Gemini 1.5 Pro with Deep Research.”
Once you select the model, you’re able to ask about any topic that you want to research. Gemini will then process your query and create a research plan. You can edit this research plan if you don’t like how the model wants to tackle the topic. You can simply chat with it to add or remove steps from the research plan, and the model will adapt accordingly. Once you’re satisfied with the research plan, give the green signal, and Gemini will start the research process.
Now, it’s worth noting that Deep Research is slow. It takes close to 3–5 minutes to generate the full response. The model first scours the internet to find all potential sources, it thenreadsthe sources, develops an understanding, and finally synthesizes a long-form report.
In my testing, Gemini Deep Research seems to be looking up a different set of websites than what’s showing up on Google Search for the same query. Also, in my opinion, Deep Research’s search index seems to be better and more comprehensive!
It’s Not Perfect—Yet!
The Gemini Deep Research feature can potentially be broken down into these five steps:
At the time of writing, Deep Research does step 1-3 marvelously. However, the final report feels lacking. If you ask me, this is because Deep Research is currently powered by Gemini 1.5 Pro, which is a less capable model compared toClaude 3.5 Sonnet or GPT-4o.
However,Gemini 2.0 Experimental Advancedis currently out in beta (available to Gemini Advanced users), and in my testing, I found it excellent—more comprehensive and thorough than GPT-4o and Claude 3.5 Sonnet. This means that in the coming months when Gemini 2.0 Advanced is released and is powering Deep Research, we can expect seriously incredible results with the search assistant!
Where Gemini Deep Research Fits Into Our Search Habits
According toSEO (Search Engine Optimization),all our web searches can be divided into four categories—navigational, transactional, informational, and commercial. Here’s what that means:
Now, Google still dominates most navigational and transactional search queries. However, it seems that people are moving to AI chatbots or platforms like TikTok or Reddit with their informational or commercial search queries. In fact, Google’s search market sharedropped below 90%during the last quarter of 2024.
There are a ton of underlying issues behind why this is happening, but the core of the issue is that Google search results for commercial and information queries arejust not good anymore. You’re either hit withunhelpful AI Overviewsor you get links toblog posts that are AI-generated. This means you need to do a lot more leg work to find useful sources that’ll help answer your questions.
This is where Deep Research comes in! It can scour the internet on your behalf and look through multiple websites and blog posts. Within a few seconds, it will generate a list of sources that are frankly better than what you get from a basic Google Search.
You can now browse through those sources on your own, or you can wait a few more minutes and let Gemini generate a summarized report with all its findings.
As a result, Google Search is now the platform for navigational and transactional search queries—which it’s still great at—and for commercial and informational queries, you can use Gemini Deep Research.
The magic of this approach is that Google isn’t trying to force all types of searches into a single interface. By separating deep research use cases into Gemini Advanced, they’ve created a more focused experience for both casual and intensive search needs.
It Wasn’t Perplexity or ChatGPT After All!
When ChatGPT launched in late 2022, followed byPerplexity’s AI search engine, many (including me) thought these tools would eventually replace Google Search. The appeal was clear—instead of sorting through multiple blog posts and articles, you could have a conversation with an AI that would gather and explain information to you. Google’s gradual decline in the quality of the search results didn’t help them either.
Eventually, the search giant took notice and responded with AI Overviews—an AI-powered search experience that gave us not-so-helpful and sometimes harmful responses to our search queries, burying the web links below it. Needless to say, it wasn’t the big hit Google might’ve hoped for, and it possibly pushed more users to other platforms. However, with Gemini Deep Research, it seems Google has found the answer to its problems, and it’s poised to be a better solution than any other search product on the market.
While ChatGPT’s search features help it stay current, they’re more focused on supporting chat responses than conducting comprehensive research. And although Perplexity offers a solid AI-powered search experience, Gemini Deep Research stands out with its ability to create and execute multistep research plans that are also editable. This capability already produces more thorough results and will only become more powerful once Gemini 2.0 Advanced is introduced and fully integrated.
The future of search might not be about replacing Google but rather about evolving how we use it. Quick navigational and transactional searches can be reserved for Google—which it’s still great at. Whereas informational and commercial queries that require deep research can be delegated to specific AI research tools, Gemini Deep Research is currently the most promising tool on the market.
