Tablet centered in a softly lit blue-and-peach corner displays the text “(I do not think it means) what you think it means,” while shadows of a flower and a dagger appear on opposite walls, suggesting meanings beyond the words on the screen.

What You Think It Means

A third of recent websites now contain AI-generated text; Google is increasingly making AI summaries the default mode of knowledge retrieval; and third parties are creating “sloppelgangers” of artists’ web pages with AI. What might this trend mean for cultural heritage, and what should we do about it?

Artists Janet Cohen, Keith Frank, and Jon Ippolito have created an interactive artwork that shows how AI can rewrite the same website according to different biases. Entitled (I Do Not Think It Means) What You Think It Means, the work was exhibited at the University of Southern California’s Future of Writing gallery last spring and is featured in a special session for American University’s MyFest26 on June 22nd entitled “Personalization or Ensloppification? When AI Rewrites the Web Just For You.

Screenshot of a dark-themed interface describing the 1993 artwork “Manhattan Project,” with photographs, a data visualization, and AI-generated interpretive text discussing offense, defense, childhood rivalry, and power dynamics in the performance.(I Do Not Think It Means) What You Think It Means (2026) presents an unsettling vision of a future where even critical discourse becomes algorithmically tailored to the reader’s preferences. As the viewer interacts with this website, AI rewrites written documentation of past works by the artists in real time to align with biases suggested by the viewer’s choices. The description of a single work may morph from academic formalism to psychological introspection to political critique, suggesting a web that is no longer fixed, but atomizes into words that are recombined to tell the audience what it wants to hear.

You can interact with the work here and watch a two-minute introduction here.

After a quick demo of this work, MyFest facilitators will lead an exercise to see how AI could rewrite the web to match their own biases. Participants will then workshop practical responses—like requiring source-checking across multiple outlets or via aggregators like AllSides, building class activities that compare AI summaries to original texts, and even a return to Web 1.0 protocols like RSS and web rings—to help protect public knowledge in a more personalized web.

Four translucent blue panels filled with blurred text stand in perspective against a dark background, illuminated by overlapping red, green, blue, and white spotlights that create a layered, floating effect. Facilitating the session is University of Maine New Media professor Jon Ippolito, who will lead a discussion inspired by a work created with his longtime collaborators Janet Cohen and Keith Frank.

Since forming their collaboration in 1992, the three artists have explored the tensions and contradictions inherent in working together. Their projects have examined the adversarial dynamics of installations and books, as well as the Internet’s capacity to fuel flame wars and other clashes of perspective. Their latest work turns to AI, asking how it might foster subtler and more insidious forms of division.

You can watch them haggle, argue, and throw things at each other at three.org.

Scroll to Top