Using AI for Research and Summarisation

One of the most popular uses of AI is research — finding information, summarising long documents, and making sense of complex topics. AI can be extraordinarily helpful for this. It can also be extraordinarily misleading if you do not know how to use it carefully.

This lesson will teach you how to use AI effectively for research and summarisation, and — just as importantly — how to avoid the traps that catch people who trust AI output without verification.

What AI Is Good at in Research

AI excels at certain research tasks:

Summarising Long Texts

If you paste a long article, report, or paper into an AI chat, it can produce clear, concise summaries. This is one of the most reliable AI use cases because the source material is right there — the model is working with the text you gave it, not making things up.

"Here is a 3,000-word article about renewable energy policy. Summarise the main arguments in 5 bullet points, each no longer than 2 sentences."

Extracting Specific Information

You can ask targeted questions about a document you have pasted:

"What methodology did this study use?"
"List all the statistics mentioned in this article"
"What are the three main counterarguments the author addresses?"

Comparing and Contrasting

AI is excellent at comparing sources when you provide them:

"Here are summaries of two studies on the effect of screen time on children's sleep. Create a comparison table showing where they agree and where they disagree."

Explaining Complex Topics

When you encounter a concept you do not understand during your research, AI can explain it at any level:

"Explain 'p-hacking' in simple terms. I am a first-year undergraduate."

Generating Research Questions

One underrated use is asking AI to help you figure out what to research:

"I am interested in how social media affects political opinions. What are the five most important research questions in this field that I should read about?"

The Critical Rule: Always Verify

Here is the most important rule for using AI in research, and it deserves to be in bold:

Never cite an AI-generated claim as fact without verifying it against a primary source.

This is not because AI is always wrong. It is because AI can be wrong in ways that are very hard to detect. It produces text that sounds authoritative, uses proper academic language, and structures arguments logically — even when the content is partially or entirely fabricated.

The Citation Problem

One of the most notorious failure modes is fabricated citations. If you ask an AI to provide references for its claims, it will often generate plausible-looking academic citations — correct-seeming author names, realistic journal titles, believable dates — for papers that do not exist.

Example of a fabricated citation: Smith, J. & Johnson, A. (2019). "The Impact of Social Media on Adolescent Political Engagement." Journal of Digital Communications Research, 14(3), 287-305.

This looks perfectly real. The author names are common, the journal title sounds legitimate, the formatting is correct. But if you search for this paper, you will not find it — because the AI made it up to satisfy your request.

How to Verify AI Research Output

Check specific claims against primary sources: If the AI says "a 2021 study by Harvard researchers found that...", search for that specific study. Does it exist? Does it say what the AI claims?
Search for cited papers: Copy any citation the AI provides and search for it in Google Scholar, PubMed, or your university library. If it does not appear, it is likely fabricated.
Cross-reference statistics: If the AI claims "73% of teenagers use social media daily," search for that specific statistic. Who published it? When?
Be especially sceptical of specific numbers: AI is more likely to fabricate precise statistics ("a 23.7% increase") than general trends ("a significant increase"). Specific numbers should always be verified.