Kentucky School Sends Home AI-Generated Maps Full of Gibberish

Farnsley Middle School, part of Jefferson County Public Schools in Louisville, Kentucky, distributed printed student agendas containing AI-generated graphics riddled with errors. A US map labeled Illinois as "Vitoiis" and Kentucky as "Venecky." The periodic table was nonsense. An extra planet called "Marc" appeared in the solar system alongside "Sature" instead of Saturn. The school district told local news outlet WDRB only that it was "communicating with parents," and has not identified which AI tool was used or whether the problem extends beyond one school.

What happened

Detail	Fact
School affected	Farnsley Middle School, Louisville, Kentucky
District	Jefferson County Public Schools
Errors found	Garbled US state names, nonsense periodic table, invented planet “Marc”, “Sature” for Saturn
Illinois labeled as	“Vitoiis”
Kentucky labeled as	“Venecky”
Story first reported by	WDRB (local Louisville outlet)
District response	Communicating with parents; no tool identified
Gizmodo test date	August 18, 2026

A bound student agenda, printed and handed to pupils, contained AI-generated images that no human appeared to have reviewed before distribution. The errors were not subtle. State names were replaced with invented strings of letters, the periodic table bore no resemblance to the real one, and the solar system gained a new planet.

WDRB reported the story first and published video of the agenda. The school district did not give the outlet a clear explanation, nor confirm whether other schools in Jefferson County received the same materials.

Which AI tool made the map?

That question remains open. Gizmodo tested the major AI chatbots by asking each to generate a labeled US map. ChatGPT (OpenAI), Grok (xAI), and Claude (Anthropic) all produced accurate results, though Claude used state abbreviations rather than full names. Google’s Gemini, which uses a model called Nano Banana for image generation, produced a map with garbled state labels strikingly similar to the school’s version. Gizmodo’s test took place on August 18, 2026.

Three other explanations are possible. The school may have used an older AI image generator from roughly a year ago, when this type of text-rendering error was far more common, and stored the graphics for later use. It may also have used a niche software package sold specifically to school districts, which could be running an older or less capable image model under the hood. The district has not responded to questions about the tool.

For context, text rendering has been a known weakness in AI image generators. Models trained primarily on visual data often treat words in an image as shapes rather than readable characters, producing plausible-looking but meaningless letter sequences. The best current models have improved significantly, but errors still appear, as this case shows.

Why it matters

The district’s own website states that it views generative AI “not as a replacement for human intellect or the invaluable role of educators, but as a powerful tool to enhance learning, teaching, and operational efficiencies.” The agenda is a direct contradiction of that stated policy.

Gizmodo noted that the district’s website also features AI-generated imagery of children with too many fingers, phones with nonsensical designs, and tablets that appear to float in mid-air. The problem appears to be systemic, not a single slip by one staff member.

This is less a story about AI failing and more one about a missing review step. As we have covered, AI tools often add friction when there is no quality gate in place. A person who looked at that map for five seconds would have seen the errors. The issue is that, apparently, no one did.

The US State Department recently published a map of Africa with incorrectly placed country names, so this is not a problem limited to underfunded school districts. Any organisation pushing AI-generated content out the door without a human sign-off is exposed to the same risk.

Our take

The AI did exactly what it was designed to do: produce an image that looks like a map. The humans in the chain failed to do what they were designed to do: check the work. Blaming the chatbot here is like blaming a search engine for a bad source you printed without reading.

What concerns us more than this one agenda is the pattern. A district that publishes floating-tablet stock art on its own homepage is not running a considered AI workflow. It is using AI as a shortcut, full stop. That is fine for low-stakes internal tasks, but not for materials handed to children as factual references.

If you are building AI into any content or document workflow, whether in a school, a business, or an agency, the minimum viable process is a human checkpoint before anything goes out the door. If you need help structuring that kind of AI integration with proper review steps, that is exactly the kind of work we help businesses set up at Lumien.

What to do about it

1. Audit any AI-generated content in your pipeline before it reaches a customer, student, or client.
2. Set a policy that no AI-generated image or document skips a human review, even for internal use.
3. Test your image generation tool specifically on text-heavy outputs (maps, tables, labels) before relying on it for anything factual.
4. If you use a third-party software package that includes AI features, check which underlying model it uses and when it was last updated.

The simplest safeguard here would have been one person looking at a map of the United States and noticing that Illinois does not exist.

Source: Bing News · Claude AI