Ever since ChatGPT arrived on the scene, the hype around AI has only intensified. As talk of Artificial general intelligence (AGI) and “superintelligence” — yeah, OpenAI chief, Sam Altman, is now talking about that — heats up, we have another buzzword to deal with.
Say hello to Agentic AI. In simpler terms, AI agents that are supposed to automate a chunk of our digital chores. Think of Gems in the Google lexicon. Custom GPTs by OpenAI. Or Copilot Actions by Microsoft.
The idea is to have an AI do your task, or some portion of it. Qualcomm and MediaTek have already prepped their silicon for the era of agentic AI. But here’s the problem. We don’t have a true agentic AI tool yet. We’ve barely crawled past the inquiry-response transaction flow that most generative AI chatbots offer.
Enter Deep Research, the first agentic AI product in the Gemini family.
A fundamental rethinking of search on the internet
As the name makes abundantly clear, Deep Research is good at research, but in a much more controlled fashion than an average Google Search. With Deep Research, you can lay out the outline for the search quest before the process begins.
You can specify the exact sources (or kind of sources) to get the results. That’s fundamentally different from Google Search, which mostly responds to keywords clubbed together, and shows results that it deems are worthy of a look.
That’s a fundamentally flawed approach, and we often end up in a cesspool of clickbait or AI-generated jargon. Plus, Google’s random changes to its search algorithm often mean the search results for the same query may look different a day, or week, later.
Deep Research pulls up material from a controlled and user-specified knowledge bank. So, let’s say you are trying to find information about the impact of social media on the mental health of young users, but only from peer-reviewed research papers. The results would stick to scientific papers only.
For journalists, students, researchers, or even businesspeople, this approach saves a lot of time. More importantly, it doesn’t put the onus of trusting a source on the user.
You already are familiar with the source, or its veracity, so the material you get doesn’t come with a trust conundrum. Moreover, the chore of skipping past the bad, non-desirable search results or ads is simply non-existent in Deep Research — at least right now.
Deep Research essentially drafts a multi-step search activity, finds the information on your behalf, and repeats the process as the “search agent” moves from one source to the next, hunting for a new piece of relevant information.
Essentially, it saves you the drudgery of running into the same information as you jump between different search results, in hopes of finding the wisdom you seek. In a nutshell, the time-consuming and psychologically infuriating parts of a Google Search are avoided.
That’s not even the nicest part about Deep Research.
Helpful, in just the precise way
Researching and finding information from credible sources is only one-half of the picture here. Deep Research takes the pain point of clicking back and forth between different search result entries, or opening a few dozen tabs. Dealing with a bunch of tabs on a large screen is already a hassle for multiple reasons.
The most important of them all is the hunt for that exact nugget of information embedded within a wall of text, video, or audio. Deep Research not only pulls up reliable information from the sources you’ve cherrypicked, but it also presents all those findings in a non-repetitive, coherent fashion.
Just what you want, from the sources you seek.
Now, unless your search task involves a single-step reference on the internet, you have to break the process into multiple steps. So, let’s say you want to learn about the art of mushroom farming. You would ideally look up information about the seed varieties, weather conditions, pest mitigation, and diseases — separately. It’s hard to find a definitive guide, especially one pulled from reliable sources.
Deep Research does just that for you. All the information that it has crawled across the web to gather, will be presented in the form of a neatly-curated article, with appropriate headlines, tables, and categorical breakdown.
It is the kind of search report that would otherwise take you hours to internalize and flesh out in the form of a document. For anyone who is in the profession of referencing and memorizing knowledge on a daily basis, this tool is a lifesaver.
Take for example this search query:
I am writing a paper on the application and differences between NMP and LFP batteries in the context of electronic vehicles and fire hazards due to battery. Pull up details from research papers and reputed agencies only. Help me understand and clarify the subject.
What I got after roughly 2-3 minutes of research was a comprehensive draft, the way I would write a thesis, legal brief, or research paper. I gave a brief demo of Deep Research to a research student, a lawyer, and a journalist. The overwhelming sentiment was that of “wow” mixed with a sense of relief.
It’s not every day that you see people willing to pay $20 a month for an AI tool that is not even mainstream. Husain Anis Khan, an Alex Chernov Scholar at Melbourne Law School, told me that he loved the premise of being able to find academic research material.
Md Meharban, a multimedia journalist whose work has appeared at outlets like Reuters, NatGeo, AFP, and The New York Times, also tells me that Deep Research could prove to be a valuable tool in their workflow.
“A healthy chunk of my documentary work relies on research. The deeper, the better,” Meharban tells Digital Trends. “If I can narrow down the unexplored areas of an assignment, chances are higher that my work will stand out.”
Hitting the human-machine sweet spot
I’ve embarked on my fair share of overtly optimistic AI adventures. Experimenting with an AI girlfriend (which a few take as far as virtually impregnating), using it for inbox relief, toning down my lazy Gmail conduct, and simply giving up on Apple Intelligence, my experience has been a mixed bag.
Deep Research is the first AI tool that has offered a fulfilling experience, something I can’t say for any other AI tool out there. I’ve paid for more AI products and subscriptions than my gaming, streaming, and reading passions combined, so I feel the sting of paying for a poor product.
For my work as a journalist, a tool like Deep Research has proved nearly indispensable, especially when researching topics like triboelectric nanogenerators on wearables and fabrication complexities for microfluidic sweat sensors.
If I go looking for the aforementioned materials on Google Search, I will essentially be playing a keyboard whack-a-mole spanning across multiple pages of Google Search links. With Deep Research, I am simply narrating what I seek, in natural language.
There’s no guesswork involved. I can specify the exact search route and the knowledge destination. I can tone the whole operation to my specific needs — be it a research-themed task or simply a marketing-related exercise.
Being able to tune it all to your needs, and getting it done without having to stray away from normal human conversational tone is what stands out. It makes my workflow a tad less robotic. A dash of human touch in there, if you will.
Then there’s the value conundrum, which any sane human does a double-take for. With products like Deep Research — or rivals like Perplexity Pro or ChatGPT Plus — the lingering question is just how much value you get out of a $20 monthly subscription.
The best $20 spent for work, and a few unexpected bonuses.
Within Google’s ecosystem, the competition is non-existent. I got access to Gemini Advanced with the Google One AI Premium subscription, which also offers 2TB of cloud storage and Gemini integration across a majority of Google products that we use on a daily basis.
One-click import into Sheets? Add a research brief to Docs? Compose in Gmail? You get all that — alongside Gemini Deep Research — with the bundle. It’s far better value than OpenAI or Perplexity’s products.
Moreover, I would much rather have my workflow concentrated within Google’s universe, than consent to a whole bunch of questionably ethical and privacy-risking T&C of another AI product ecosystem.