Lesson 129 · 2 min
Fake citations from AI
Chatbots sometimes invent sources that look perfect, with real-sounding authors, titles and links, so open every source before you use it.
A language model writes what fits the pattern of an answer. When you ask for references, it may produce a paper title, a journal and a year that look right but do not exist.
This has embarrassed students, journalists and even lawyers who submitted invented court cases. The fake looks just as confident as the real thing.
Tools with web search or deep research modes do better, because they quote pages they actually found. Even then, a link can be real while the claim attached to it is missing from the page.
Open each source, find the sentence that supports the claim, and check the author and date. If you cannot find it, leave it out.
Do it now
Test a chatbot's sources
- 1. Ask a chatbot a question about your field and request three sources.
- 2. Search each title in a search engine or a library database.
- 3. Count how many you could open and confirm.
Answer this question: [QUESTION]. For each claim, give the source you used with a link. If you are not sure a source exists, say so clearly instead of guessing.
Quick check
Keep going
Next →Misleading chartsWords from this lesson
- Hallucination
- When an AI states something false with confidence, such as a made-up fact, quote or source.
- Grounding
- Tying an AI's answer to real sources you supply or it looks up, so claims can be checked instead of invented.
- Deep research
- A chatbot mode that searches many web pages over several minutes and writes a longer report with sources.