FAQS

While generative AI tools, like the one available to students in Grammarly, can be real life savers in many aspects of writing, it is important to know that gen AI can also make up the information it provides. This includes creating fake references for sources like books and journal articles. It might generate details or quotes from sources that don't actually exist. These are referred to as “hallucinations”.

Generative AI tools may have the data in its knowledge bank to know the name of a real author from the profession and the names of journals in the field, but it might have to make up the rest of the information of a reference to fill in the gaps. Article titles are often fake, made up of words that have appeared in similar publication titles in different combinations.

It's important to develop your critical thinking skills so you can evaluate and question information you read from all sources, not just content produced by gen AI.

How to Check if a Source is Real

One quick way to validate sources is to search using the resource’s title. Try copying and pasting just the title into:

• Google Scholar
• The library’s Discovery search
• Google web search

If the source is real, a match to the publisher of the information should show up in the result list. If you can’t verify that a source exists in one of the resources above, try another in the list. You can also try adding additional pieces of information to your search (such as the author’s last name) to see if that changes your results.

If it turns out the source isn’t real, it is important that you remove the reference from your paper and any content that was credited to that source. You would then want to search for a new source to include in your assignment. Have questions about doing research? Ask a Librarian!

Fake Reference/Hallucination Research

Studies are being published that looked at the prevalence of fake references in Gen AI created content. In one study of 115 references, only 7% were both authentic and accurate (Bhattacharyya et al., 2023). Another study found that the Gen AI suggested references only existed 14% of the time (Zuccon et al., 2023). A third study looked at 59 references and 69% were fabricated (Gravel & Osmanlliu, 2023).

While the prevalence of fake references varies across studies, this recent literature tells us that generative AI is not able to consistently integrate legitimate sources into the content it produces.  Writers and readers alike hold the responsibility of evaluating the information they produce and consume to ensure its accuracy.

References

Bhattacharyya, M., Miller, V. M., Bhattacharyya, D., & Miller, L. E. (2023). High rates of fabricated and inaccurate references in ChatGPT-generated medical content. Cureus, 15(5), e39238. https://doi.org/10.7759/cureus.39238

Gravel, J., D, A.-G. M., & Osmanlliu, E. (2023). Learning to fake it: Limited responses and fabricated references provided by ChatGPT for medical questions. Mayo Clinic Proceedings: Digital Health, 1(3), 226–234. https://doi.org/10.1016/j.mcpdig.2023.05.004

Zuccon, G., Koopman, B., & Shaik, R. (2023). ChatGPT hallucinates when attributing answers. In Proceedings of the Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific Region (SIGIR-AP '23). Association for Computing Machinery, 46–51. https://doi.org/10.1145/3624918.3625329