Cartographer Warns AI-Generated Maps Spread Falsehoods Without Responsibility

The author, a 23-year blogger on maps who identifies as having no cartographic skills, opens by rejecting the label of "gatekeeper," "Luddite," and "AI hater" that pro-AI advocates in cartography have deployed against critics. She notes that the ICA's Commission on Map Design has been posting examples of AI-generated maps and promoting websites that use ChatGPT to generate maps from natural language prompts.

The author acknowledges the historical merit of the "gatekeeper" argument—tools like iMovie and affordable digital SLRs did democratize their respective fields. However, she observes that the professionals promoting AI-generated maps are not themselves non-professionals seeking entry; they are established cartographers. For the author herself, despite lacking mapmaking skills and envying those who possess them, an AI mapmaking tool would be useless: "I wouldn't touch it with a ten-foot pole." She explains that beyond her concerns about generative AI in her other field (science fiction, where AI trained on unconsented author work has flooded fiction markets with auto-generated content), she objects to using AI because "I wouldn't be the one making the map: the AI is doing the work. I couldn't claim it as mine, nor would I be able to enjoy the satisfaction—the joy—of having created something."

The author questions the commercial sense of professionals using AI-generated maps. Unlike hiring a contractor (where a human takes responsibility) or using tools like Google MyMaps, AI-generated maps raise a critical legal problem: they are not copyrightable. "You don't own the work," she writes. "So why should anyone pay you money for an AI-generated map based on your prompts, or employ you to make one? Anyone can use the same prompts and get the same results." She speculates that AI-generated maps would likely not be accepted for publication in professional venues like the NACIS Atlas of Design or the GeoHipster calendar, and raises concerns about Amazon and Apple being "flooded with AI-generated fake books designed to fool buyers"—a precedent for AI-generated content flooding and degrading creative markets.

The author predicts negative industry-wide consequences if AI-generated maps were widely adopted. Citing experience in other fields, she notes that corporate savings from AI adoption have not translated into leisure time or job preservation for professionals. Instead, "AI adoption in other fields has not meant more leisure time for professionals. It's led to mass layoffs, and precarious employment for the survivors whose work has shifted to checking and correcting AI outputs." She invokes the actual Luddite movement—opposition to low pay, poor working conditions, and bad quality of work—as a model for why professionals should resist this technology.

The author then addresses what she calls "bad quality of work." Large language models like ChatGPT are "stochastic parrots: what they produce is not necessarily true. That is to say, they make stuff up." She cites real-world consequences: "AI hallucinations in the form of fictitious citations have proliferated in academic publications and legal filings (for which lawyers have been sanctioned)." She initially hoped AI-generated cartography would be less prone to hallucinations; the evidence suggests otherwise. Cartographer Daniel Huffman documented that "some of the AI-generated maps shared by ICA's Commission on Map Design contain 'serious inaccuracies' (as in, a U.S. map with two Ohios)." The author notes that despite these acknowledged errors, the maps' promoters have "continued to post them on social media without including any sort of caution."

The author criticizes this as unethical. "While geography enthusiasts such as you and I will spot some of the errors, I do not doubt that some lay readers are going to take the maps at face value and be misled." She distinguishes between errors that slip past correction efforts and knowingly sharing false information: "it's one thing to let a mistake slip past, and quite another to knowingly share false information." She grants the Commission's good intent—their brief includes exploring AI errors—but judges their public sharing without warnings as "professional negligence." Cartographer Dylan Moriarty and cartographer Daniel Huffman both argue that AI-generated maps should come with disclaimers and should not misrepresent their subject or leave readers with "a significantly inaccurate understanding."

The author points out that AI-generated maps are visually distinctive: "they have the same graphic design language shared by a lot of AI-generated artwork, in terms of text and colour choices and general over-sharpening." She notes that people are already pushing back against flyers using these design defaults, and maps will likely face the same skepticism. She frames the core issue as accountability: "take responsibility for the work, because a computer can't." She critiques the argument that users simply gave "a better prompt"—if prompting correctly and debugging outputs require specialist knowledge, then the claim that AI "opens the field to non-professionals" is dishonest.

The author then challenges the typewriter analogy often used by AI advocates. She notes that publishers historically insisted on typewritten manuscripts on legibility grounds—contradicting claims that no gatekeeping occurred. More fundamentally, "writers as a group will not shut up about their tools": George R. R. Martin is known for using WordStar; Larry McMurtry thanked his typewriter at the Golden Globes. Artists and writers openly acknowledge their tools. The key difference is control: "in each case you are in complete control of what appears on the page: no one is going to question your authorship if you use an IBM Wheelwriter instead of an Olympia SM9." With generative AI, the human does not maintain that control—the algorithm does. The author argues that generative AI's best analogy is not the typewriter but cocaine: a tool that becomes addictive, progressively impairing judgment, with users unable to stop despite professional sanctions.

The author acknowledges her critique extends beyond maps because "the problem of generative AI can't be seen in isolation." Mapmakers focused narrowly on their field may not recognize these patterns from other industries. She observes that it is "in the proponents' interest to be laser-focused on what it can do for their specific field... because the problems we've heard about haven't really struck home yet." The article concludes by noting that vocal resistance to generative AI and the angry defensiveness of AI boosters attacking resisters should themselves be telling signals.

The article frames the debate over AI-generated maps within a broader critique of generative AI adoption across creative and professional fields. The author positions herself as a non-cartographer who might logically benefit from AI mapmaking tools—yet explicitly rejects them, both because she works in science fiction/fantasy writing (where generative AI trained on unconsented author work has flooded markets with low-quality content) and because using an AI tool would mean she does not own or control the output. This personal stance grounds her skepticism about the professional case for AI cartography.

A key tension the article identifies is that promoters of AI-generated maps appear to be professionals themselves, not non-professionals seeking to bypass gatekeeping. This undermines the "democratization" argument. The author cites cartographer Daniel Huffman's documentation of serious errors in ICA-sponsored AI maps—such as a U.S. map with duplicate Ohios—which were shared publicly without caution labels despite being known to contain mistakes. The author calls this "professional negligence" because lay readers may take false information at face value. Legally, AI-generated work cannot be copyrighted, raising the question of what cartographers gain by promoting a tool that commodifies their labor: anyone with access to the same prompts can produce identical maps, eliminating the value proposition of hiring a professional.

The author also draws on patterns from other industries where AI adoption has not benefited workers as promised. Rather than delivering leisure, AI adoption has produced mass layoffs and precarious roles focused on validation and error-correction—work that still requires domain expertise to perform reliably, making the "anyone can use this" claim hollow.