For this experiment, I picked Mary Shelley’s Frankenstein; or, The Modern Prometheus from Project Gutenberg, and downloaded the txt file so I could use it across tools.
Method 1: Voyant Tools
I uploaded the raw .txt into Voyant Tools and explored trends and topics Trends let me track how specific words rise and fall across the narrative. The main payoff was “theme timing”: spikes in words like man or eyes gave me a map of where to return for close reading, instead of relying on memory. Topics gave me rough term clusters that behaved like mini theme bundles, helping me see repeated conceptual groupings I might not have named on my own.
I asked Gemini to pull frequently recurring phrases. Its list differed from Voyant. Voyant gave transparent word frequencies. Gemini returned longer thematic phrases rather than strictly repeated ones. This comparison shows AI can drift from measurable evidence, so I verified motifs with Voyant, adding screenshots, prompts, and proper tags in my WordPress post for clarity.
What does this suggest about the “Age of AI”
Voyant feels methodologically reliable because it clearly shows what is being counted and where patterns appear across the text, but it cannot explain why those patterns carry meaning. Gemini can generate interpretations and relationship networks quickly, yet it sometimes blurs the boundary between evidence-based inference and confident speculation. The main concern in this Age of AI is not that machines replace reading, but that interpretations risk becoming detached from textual proof. In practice, this means saving prompts, screenshots, and tool settings, treating AI outputs as tentative hypotheses, and tracing every interesting claim back to specific passages in the original work. A responsible workflow pairs Voyant’s transparent measurement with AI’s creative synthesis, while preserving the human role of justifying arguments through clear citations rather than intuition alone.
I think your insights about questioning Gemini (and AI in general) in terms of its analysis of the computational analysis from books are very important to consider. I think the more disconnected from the actual text Gemini gets, the more prone to mistakes it could get, so inferring yourself based on the visual analysis Gemini creates could be more reliable.
I really like your interpretation of the certain tools available to us in Voyant. I like how I looked at the tools and thought in a completely different way than you did but, both of our points have reasoning. I came up with the same sort of results as you did with the AI, I don’t believe that is it completely reliable.
I had fun looking at your textual analysis of “Frankenstein” because I just watched the new movie and was able to compare your findings on Voyant and Gemini to my experience of the movie. I think you’re right that Voyant feels a lot more methodologically reliable than Gemini but I do see the appeal of using Gemini as a more fluid tool for finding themes or bigger picture context. I like the idea of using them in tandem, as long as we have enough expertise to make sure that Gemini isn’t making anything up.