Research With AI Tools

Key takeaways

• AI research tools combine an LLM with web sources retrieved in real time.
• You should always open and verify citations before reusing a claim.
• A private RAG system lets you search sensitive documents without publishing them online.

Using AI tools to research a blog article

Starting a creative task such as writing a blog article or course lesson about a technical topic like “How do image generation models work?” can feel intimidating. Traditionally, it means searching Google for academic papers, well-written blogs, Wikipedia sources, or even books, opening dozens of tabs, reading redundant or irrelevant content, and manually extracting useful information. It is a time-consuming and often frustrating process.

But there is a better way to approach research: tools such as Perplexity, ChatGPT Search, or Ai2 OpenScholar, which combine language models (LLMs) with real-time information retrieval. Today, I will show you how to use them to make your research faster, more precise, and more effective, whatever the creative task. We will focus on the concrete example of creating a course lesson, but these tips apply to any similar writing task.

Step 1: Understand the limits of traditional research

As writers, we are used to researching online iteratively: reading articles, saving relevant information with its links, meaning our sources, and repeating the process until we have a thorough understanding of the topic. Although this method works, it is quite inefficient. It takes a lot of time to sort through useless or redundant content to find the needle in the haystack.

Using LLMs to learn a new topic is useful, but you need to watch for potentially inaccurate information. They are generally reliable for learning popular subjects. For example, if you want to learn a programming language like Python and write an article about it, AI models are often well trained because so many tutorials exist online. For a newer subject, such as a neural network paper published last week, however, the AI could provide incorrect information. In that case, it is better to give the paper to the AI so it can use it when answering.

AI tools like Perplexity, ChatGPT Search, and Ai2 OpenScholar simplify this process by combining web search with an LLM. They retrieve the most relevant information, summarize it, and even provide citations, helping you quickly extract key information while checking where it came from.

Step 2: Research with AI tools

Let us start with Perplexity. Perplexity connects LLMs to web search, making it a useful tool for efficient research. To begin, enter your main question in Perplexity, for example:

“How do image generation models work?”

Perplexity retrieves information from reliable sources such as research papers, technical blogs, and reputable websites, then summarizes the important points. From there, you can refine your query to explore specific aspects, such as:

• “What is the difference between autoregressive and diffusion models?”

• “What are the real-world applications of image generation models?”

Perplexity also suggests related questions based on your query. This feature is useful for exploring the topic in more depth and discovering perspectives you may not have considered.

To target your search, use Perplexity’s focus categories feature. For example, selecting “academic” ensures that your results come from detailed, credible academic papers. Every answer includes citations, allowing you to check the sources and validate the claims.

An academic-focused search displays research papers among the sources used.

Perplexity works well for different questions across many fields. ChatGPT Search works similarly and is also optimized to provide fast, almost immediate answers like Google search results.

Ai2 OpenScholar, on the other hand, focuses only on research papers, which can produce more precise and complete answers. This can come at the cost of speed, with some answers taking more than 10 seconds to generate. The system spends more time finding the best available documents. There are different ways to manage this speed-quality tradeoff, which we explain in our course on developing with LLMs.

Step 3: Avoid hallucinations with real-time retrieval

Using an LLM without a web search tool can lead to hallucinations, meaning plausible but false answers. LLMs rely on their training data, which may be outdated or incomplete. They almost never state that they do not know. They provide an answer even when they are unsure.

Tools such as Perplexity and Ai2 OpenScholar reduce this problem by retrieving information in real time from external, up-to-date sources.

In our examples, these tools retrieved the latest papers and technical publications when researching autoregressive and diffusion models, helping provide credible information. Still, you should always verify the citations to validate the information and make sure your article reflects current knowledge.

Step 4: Search local documents

Sometimes your research involves understanding specific documents you already have. If those documents do not contain sensitive data, many LLM platforms can help you extract the key information. Major tools such as ChatGPT, Perplexity, and Claude let you upload files like PDFs directly into their conversations.

For sensitive data, you can consider building a custom RAG, or retrieval-augmented generation, system. This involves:

1. Splitting your documents into chunks and storing them in a vector database.

2. Using APIs such as Bing or Perplexity to collect relevant public information.

3. Combining the private and public documents, ranked by relevance.

4. Providing the final set of documents to an LLM to produce a complete and precise answer.

Conclusion

By using tools such as Perplexity or custom RAG systems, you can optimize your research process, reduce inefficiencies, and create high-quality articles or courses. Real-time retrieval helps keep your content accurate and current, while citations build trust. Whether you are researching public topics or private data, these tools offer a faster and smarter workflow.

Happy researching and writing!

FAQ

Why use an AI tool for research?

It can search multiple sources, summarize the relevant passages, and provide citations faster than a workflow built around dozens of manually opened tabs.

What is the difference between Perplexity and Ai2 OpenScholar?

Perplexity covers several types of sources and responds quickly, while OpenScholar focuses on research papers and may take longer to find the best documents.

Do citations from an AI tool eliminate hallucinations?

No. Retrieval reduces the risk, but you still need to open the sources and confirm that they actually support the sentence produced by the model.

How can you search a PDF with an LLM?

Tools such as ChatGPT, Perplexity, and Claude accept PDFs in a conversation and can extract or summarize the information you request.

When should you build a custom RAG system?

Consider one when documents are private, when you need precise control over retrieval, or when the workflow must combine internal and public sources.

How do you verify AI-produced research?

Review every citation, compare multiple sources, check their dates, and remove any claim you cannot clearly connect to reliable evidence.