AI Resilience Report forHistorians

If you're worried that AI will replace historians, here's the honest picture: today, AI is mostly augmenting the slow, repetitive parts of historical research, not the thinking that makes history meaningful. The biggest leap is in reading old handwriting. A May 2026 IEEE Spectrum report on a Wilfrid Laurier University study found that general-purpose large language models pushed character error rates on 18th- and 19th-century documents below 2 percent — beating the specialized tool Transkribus while running 50 times faster and at roughly 1/50th the cost ^{[1] [1]}.

Museums and archives are using AI to make collections more accessible too: the National Archives Museum is using AI to power a "choose-your-own-adventure" exhibition ^[2] that surfaces documents tailored to visitors' interests. But for interpretation, the picture is different. A historian testing Microsoft Copilot on his own Tudor-era research found it did not fully grasp what historical analysis is ^[3], confirming that synthesis and judgment remain human work.

Adoption is moving fast on the "grunt-work" side because tools are cheap, commercially available (ChatGPT, Claude, Transkribus), and the economic benefits are obvious — collections that were "preserved but functionally hidden" suddenly become searchable. But cultural and ethical brakes are strong. The American Historical Association warns that generative AI "regularly hallucinates content, references, sources, and quotations" ^[4] and stresses that historians' skills will be more in demand, not less, in a messy information landscape.

Consulting research agrees the broader pattern is reshaping rather than replacing: BCG argues that AI will reshape more jobs than it replaces ^[5], and the BLS now explicitly incorporates AI impacts into occupational projections ^[6] — treating it as a workflow shift, not a wipeout. For young people curious about history: AI handles the squinting; you'll still do the meaning-making.