AI for Research

AI can make you a faster, more effective researcher. It can also lead you to cite sources that don't exist. Understanding the difference between where it helps and where it breaks is essential.

What AI Is Actually Good At in Research

Explaining concepts quickly. Need to understand a topic you've never encountered before? AI is excellent at giving you a clear, plain-language overview in minutes. It synthesizes large amounts of information into digestible explanations. This is one of its genuine strengths.

Summarizing long material. Paste in a research paper, long article, or report and ask for a summary of the key findings. AI handles this well and can save significant reading time.

Generating search terms and angles. If you're not sure how to start researching a topic, ask AI: "What are the key debates around X?" or "What search terms should I use to research Y?" It's good at mapping the intellectual landscape of a subject.

Identifying what questions to ask. AI can help you figure out what you don't know yet. "What aspects of this topic should I research before drawing conclusions?" surfaces angles you might have missed.

Structuring your thinking. Once you've gathered information, AI can help you organize it, identify gaps, spot contradictions between sources, and build arguments.

The Critical Problem: AI Fabricates Citations

This is not a minor caveat — it's a fundamental limitation you need to internalize.

When you ask an AI to provide specific sources, studies, papers, or quotes, it will often hallucinate them. It will invent plausible-sounding author names, journal titles, publication years, and paper titles that do not exist. It does this confidently, with no indication that anything is wrong.

This happens because AI language models don't look up facts — they generate text that statistically resembles what a real answer would look like. A realistic-sounding citation is what a real answer would contain, so it produces one.

Never paste an AI-provided citation into academic or professional work without verifying it exists. Many people have made this mistake and damaged their credibility as a result.

Use Perplexity When You Need Real Citations

Perplexity is built differently. It searches the web in real time and provides actual, clickable sources alongside its answers. It's not perfect, but it's dramatically more reliable for sourced research than a standard language model.

For anything where you need to trace back to a real source — journalism, academic work, professional reports — Perplexity is a much better starting point than ChatGPT or Claude for that specific task.

Knowledge Cutoffs Matter

Every AI model has a knowledge cutoff — a date after which it has no training data. Ask about recent events, new studies, or anything from the past year or two, and the model may give you outdated information or simply not know.

Cutoffs vary by model and are updated with each new release. Check the model's documentation if you're researching time-sensitive topics. For recent information, use Perplexity, Grok, or direct web searches.

A Practical Research Workflow

A good AI-assisted research workflow separates what AI does well from what requires real sources:

1. 1.Start with AI for orientation. Ask it to explain the topic, identify the key debates, and suggest what to research. Use this to build your mental map, not your citations.
2. 2.Search for real sources using the angles and terms AI helped you identify. Use Google Scholar, reputable databases, and credible publications.
3. 3.Use AI to help you process sources. Paste in papers or articles and ask for summaries, key arguments, and contradictions.
4. 4.Verify any specific claim you plan to use before including it. If AI told you "a 2021 Stanford study found X," find that actual study before citing it.
5. 5.Cite your real sources, never the AI.

AI makes you a faster researcher. It doesn't replace the need to verify.