Key points:
- Users must not heavily rely on generative AI tools for research
- Why agentic AI matters now more than ever
- AI prompt engineering: A critical new skillset for 21st-century teachers
- For more news on generative AI, visit eCN’s AI in Education hub
Generative AI continues to cause some existential crises at all levels of education. There are concerns around using generative AI, with a few studies focusing on the fact that ChatGPT and other tools are reducing our critical thinking capacity.
These studies include The Memory Paradox: Why Our Brains Need Knowledge in an Age of AI, and another from MIT’s Media Lab focusing on how AI-assisted work impacts critical thinking and memory.
In the same way that Nicholas Carr raised concerns about the internet’s impact on how our brains work in The Shallows: What the Internet is Doing to Our Brains, some veteran educators view the controversy over generative AI to be a replication of the controversies around calculator use in the 1980s and early 1990s. Others are convinced the entire foundation of education will have to be rebuilt in light of generative AI tools.
Among the more emerging types of tools popular with researchers and all levels of students are those categorized as deep search research engines. According to ChatGPT, “deep search generative AI tools have emerged as a response to the limitations of traditional keyword-based search engines, offering users the ability to pose complex, natural language queries and receive synthesized, context-rich responses. These tools combine the capabilities of large language models with expansive, real-time data retrieval systems to deliver answers that are both relevant and nuanced.”
One question concerns how effective these tools are, and what these tools are using to complete their searches. Will students or other researchers begin to rely on these new tools to the exclusion of more traditional sources?
Most common K-12 topics and many technology-related topics are easily answered through the use of current internet or web sources. So, we identified a difficult and somewhat obscure topic to ask these deep search tools to investigate. Each of the deep search tools was asked: Can you find all the information possible on the 75th Regiment of Foot in the British Army that was raised in 1778?
Those tools that asked for clarification were further prompted with: I am looking for when and by which officers it was formed, where it served, and when it was disbanded. Notable events and uniform specifics would be helpful. Please search historical archives, military records, and regimental histories, if available.
The 75th (Prince of Wales’s) Regiment of Foot was raised during the American Revolution in 1778 and remained an active regiment until the end of the war in 1783. It remained in England throughout most of the war. It did not serve in America. However, several of the search tools articulated otherwise.
Google Gemini’s Deep Search tool cited four sources in its final two-page summary, but identified 32 sources that it reviewed. Gemini covered the information of the regiment available via Wikipedia, that it was raised in Wales, and generally served in Wales. Some of the officers were sent to serve in America. Although none of the officers served in America after joining the 75th Foot, a few officers served in America prior to serving with the 75th. Gemini wrote its summary in such a way that is raised several questions about the 75th that should be further investigated.
Gemini’s basic tool gave an entirely inaccurate answer confusing the regiment with the Gordon Highlanders that were raised in 1794 and amalgamated with a different 75th Foot raised in 1787, four years after the regiment that was to be the focus of the research had been disbanded. It identified several battles that the 75th fought in during the American Revolution in America. In fact, the 75th Foot participated in no battles in America during the war.
Claude provided nearly no information. It did state: I want to be transparent that while I have general knowledge about British military organization of this period, the specific details about this particular regiment might require verification from primary historical sources. If you’re looking for precise information about the 75th Foot’s history, I’d recommend consulting military historical archives or specialized historical research materials. This might actually be one of the best answers by directing researchers to better sources.
Elicit did a particularly poor job in that it provided an overview of the U.S. Army’s 75th Ranger Regiment. It was 100 percent inaccurate information. Bing responded with information from Wikipedia. Deep Seek confused the 75th Foot raised in 1778 with the 75th Regiment raised in 1787. Deep Seek did correctly identify that the regiment was disbanded in 1783, but otherwise conflated information from the 1787 regiment, including having the regiment raised in Scotland, which happened in 1787 and not in 1778. Co-Pilot actually responded that the regiment was raised in 1787 and not 1778, and then provided the history of the regiment raised in 1787.
However, Co-Pilot’s Researcher tool embedded within Co-Pilot provided a strong summary and included information on a 1783 mutiny of the regiment that was put down by Capt. Thomas Picton, an officer who would later earn fame in the Napoleonic period. Co-Pilot’s Research tool was the only tool to mention the younger Picton. It shared eight sources, including a Dictionary of National Biography entry on Thomas Picton, and an auction website, which sold a relic of the 75th Foot from the 1778 regiment. Co-Pilot’s Researcher tool suggested searching specific archival document sets within the British National Archives.
The Liner Pro tool used three Wikipedia articles, a military wiki page, and one document uploaded to Scribd. The Scribd paper included several significant errors in a brief review. It repeated common misconceptions about the regiment, including that some of the officers served in America and that the regiment remained in Wales during the Revolution. Liner Pro properly identified the three colonels who served with the regiment, but failed to identify any of the other officers who served with the regiment.
Perplexity.ai was able to distinguish that the 75th Foot raised in 1778 was not related to the 75th Foot raised in 1787 and known as the Stirlingshire Regiment. It used only three sources: two from Wikipedia and a page from the British National Army Museum. Due to inaccurate information in the Wikipedia pages, Perplexity replicated an error that appears to originally derive from an inactive website cited by Wikipedia through www.archive.org as the source for its information on the 75th Regiment of Foot (Prince of Wales’s Regiment).
Another tool, Jenni.ai, which offers to draft academic articles, provided a brief narrative titled The Service of the 75th Regiment of Foot: A Historical Overview.However, outside of some broad truisms such as “its formation was directly linked to the Crown’s imperative need to augment its forces engaged in the increasingly protracted and challenging conflict across the Atlantic,” it generally failed. Jenni.ai incorrectly identified that the 75th served in the West Indies.
This experiment raises a couple of important considerations in using these tools:
First, the accuracy of these tools is apparently reliant on the accuracy of information available on the internet. None of these tools cited any Google Books, which are often a good starting point for historical information. In some other searches, Google Books and JSTOR resources are cited. Because there was information about the 75th Regiment of Foot (1778) easily available on Wikipedia, that might have impacted the overall search process of the tools. This might be particularly true of obscure topics. It is easy to see that a number of the tools replicated errors from Wikipedia.
Second, it seems essential not to take the search tool results at face value, but to use the results to access the sources consulted. Gemini’s Deep Search probably did the best job of providing a complete source list of all the sources it reviewed. Co-Pilot’s Research tool provided the strongest and most comprehensive response, including directing researchers to archival resources.
Overall, users must not heavily rely on generative AI tools for research. Currently, the tools seem to provide a way to gather source lists in a bit more robust way than traditional search engines. Jenni.ai, though not a true deep search tool, seems to have produced the most dangerous results for a novice researcher, as it presented a strong summary of the regiment that was nearly 100 percent inaccurate.
It is essential that users are aware of the limitations of these deep search tools and make sure they use such tools in conjunction with traditional library databases and other more established tools, such as JSTOR, Google Scholar, etc.
Baule’s scholarly interests focus on online student engagement, educational technology-- particularly the impact of 1:1 implementations, social-emotional learning, and the history of education. Baule led several efforts to improve student emotional health and reduce discipline issues prior to moving into higher education. He also writes on aspects of early American history.
Baule has held memberships in the American Association of School Administrators, the American Library Association, the American Association of School Librarians, the Association for Supervision and Curriculum Development, the Consortium for School Networking, the International Association of School Librarians, the National Association of Secondary School Principals, the National Staff Development Council, and many of their state affiliates. He has served as a consultant in the areas of educational technology, facilities design, library program development, team building, and communications.
- Deep search: How do emerging research tools handle ambiguous topics? - July 9, 2025
- 5 places to find dissertation topic ideas - June 12, 2025
- It’s time to rethink the foundations of higher education - June 2, 2025