How to Use Deep Research in ChatGPT, Gemini and Perplexity to Find Topics
Deep Research in ChatGPT, Gemini and Perplexity can do more than answer questions: it can find topics that actually bring in an audience. The Habr guide's author describes a simple process: give the model a research brief, wait for a table of ideas, then filter the topics for your platform, niche, publishing history and the real interests of your readers. The key point is that a standard chatbot mode fills gaps with generic phrases, while Deep Research works more like a junior researcher.
AI-processed from Habr AI; edited by Hamidun News
A practical breakdown has appeared on Habr showing how to use Deep Research mode to find topics that can bring traffic and subscribers. Instead of an ordinary chatbot that retells the obvious, the author suggests running a deep investigation, collecting a table of ideas and then filtering them for your own audience.
Why you need Deep Research
The article's main idea is that the standard chatbot mode is poorly suited to finding strong content ideas. When the model has no solid material, it fills the gaps with generic phrases, repeats what already sits on the first page of search results, and produces a template list of topics. For a blog, a Telegram channel or a media outlet that is not enough: you need not just ideas but stories built on fresh market signals, discussions, figures and cases.
The author considers Deep Research more useful precisely because it makes the model work like a junior researcher. It reviews more sources, compares different points of view, looks for confirmation, and can admit that the data is insufficient. The piece names ChatGPT, Gemini and Perplexity as working options: they have deep research modes, and these offer enough for at least a few free runs.
«If the bot doubts its answer, it will try to approach from a different angle».
How to write the prompt
The article recommends not asking the AI for an abstract "10 viral topics" but giving it a clear research brief. The basic prompt should describe the subject of the research, the content type, the publishing platform and the audience, and then ask the model to study trends, reviews, events and other authoritative sources from the past six months. The output should be not a stream of ideas but a structured report with prioritization and an explanation of why these particular topics could take off.
- the topic or market to analyze
- the publishing platform: Habr, VC.ru, DTF and others
- a description of the audience and its pain points
- the period to search for signals
- the report format: table, sections, previews, evidence
The author also advises feeding the model your own internal data: statistics on your best posts, successful competitor publications or examples of material that has already worked in the niche. This matters because Deep Research looks for external signals, but it is the channel's historical data that helps determine what from the findings will actually resonate with your readers and what will remain just a nice-looking topic on paper.
How to choose topics
After launching the research, wait roughly 5–10 minutes: during that time the model gathers sources, consolidates ideas and usually produces a table with priorities. The work does not end there. The author suggests first writing down the topics you like, then checking through Yandex search how overused they already are, and only then opening a new regular chat. In this second session you can feed the AI the report and the audience description and ask it to rank the ideas for your specific task.
This two-step process keeps research separate from editorial decision-making. First, Deep Research answers where potential interest exists and which arguments support it. Then a separate chat helps narrow the choice for a specific product, blog or channel. The author also points to the block under the table where the model explains why a topic is worth pursuing: it often contains not only reasons for virality but also drafts for a future article outline, headline and angles of presentation.
What this means
Deep Research is gradually becoming not just a chatbot feature but a working tool for editors, marketers and authors who need quick content research without a week of manual preparation. But the model does not replace the editor: it helps assemble a field of options and evidence, while the final choice still depends on understanding the audience, the platform and how genuinely new the topic is for your market.
- Nov 19, 2026The Word "Video Game" Raised GPT-6 Astra's Willingness to Pick Targets From 6% to 92%
- Nov 19, 2026OpenAI's AI Improved an AMD GPU Design by Up to 14% in Power, Speed and Area in 5 Months
- Nov 19, 2026OpenAI's agents now complete 46% of 2–4 hour research tasks on their own, up from 23%
- Nov 18, 2026AWS adds Dynamo to HyperPod: Qwen3-235B delivers 207 tokens/s per GPU instead of 101
Key Questions Answered
Why you need Deep Research?
The article's main idea is that the standard chatbot mode is poorly suited to finding strong content ideas. When the model has no solid material, it fills the gaps with generic phrases, repeats what already sits on the first page of search results, and produces a template list of topics. For a blog, a Telegram channel or a media outlet that is not enough: you need not just ideas but stories built on fresh market signals, discussions, figures and cases.
How to write the prompt?
The article recommends not asking the AI for an abstract "10 viral topics" but giving it a clear research brief. The basic prompt should describe the subject of the research, the content type, the publishing platform and the audience, and then ask the model to study trends, reviews, events and other authoritative sources from the past six months. The output should be not a stream of ideas but a structured report with prioritization and an explanation of why these particular topics could take off.
How to choose topics?
After launching the research, wait roughly 5–10 minutes: during that time the model gathers sources, consolidates ideas and usually produces a table with priorities. The work does not end there. The author suggests first writing down the topics you like, then checking through Yandex search how overused they already are, and only then opening a new regular chat. In this second session you can feed the AI the report and the audience description and ask it to rank the ideas for your specific task.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.