AI and the School Search: Why Human Guidance Still Matters
/By Audrey Fleischner
It seems like every day brings a new headline about artificial intelligence stepping into a traditionally human role: AI-managed stores, AI music labels, AI tax preparers (spoiler: don’t use AI to file your taxes). Depending on your perspective, these stories are either inspiring (“think of all the possibilities!”) or terrifying (“what is the fate of humanity in a robot-run world?!”). But wherever you fall on the luddite-to-AI-enthusiast spectrum, there is often an element of humor. We’ve taken a particular liking to Luna, the bot that ran Andon Market in San Francisco, who had an inexplicable obsession with candles and also ordered 1,000 toilet seat covers for the employee bathroom.
Naturally, we had to ask: can AI serve as an educational consultant? We went down the rabbit hole with ChatGPT, Claude, and Gemini, testing everything from broad questions (“Where should I send my child to kindergarten in NYC?”) to highly specific prompts about admissions odds and school fit. The results were consistently baffling, occasionally helpful, and often reminiscent of Luna’s toilet seat purchase. We are not immune to the existential threat of bots one day performing our jobs, but that day is not now…and, honestly, that day doesn’t feel particularly close.
The AI School Consultant:
How It Falls Short.
Across all bot types, AI had a lot to say. Our prompts consistently generated extremely long, partially bolded, emoji-sprinkled responses that offered three categories of guidance:
Words of confident comfort: “I’ll break it down so you can actually make a smart decision (not just panic like most parents do).”
Potentially helpful, but very general, procedural information: “public schools in New York City are partly zoned, but it depends on the grade level.”
Misleading, incomplete, or simply incorrect information: Eleanor Roosevelt Middle School would be a “reach” for a Group 1 applicant...considering that no such school exists.
Given its eloquence, it is easy to mistake the volume of information for true expertise. But the quality rarely matches the quantity. One chatbot highlighted Battery Park City School and The East Village Community School as NYC’s “top public zoned schools.” Both are excellent programs, but why these two over countless other beloved options? This is akin to Luna ordering dozens of candle varieties. People do love candles, but the fixation on that product over, say, scented oils or artisanal salts, beats me.
Its “top private kindergarten” recommendations were even more puzzling: Marymount School of New York and The Goddard School in StuyTown. Marymount is a reasonable, if context-free, suggestion (AI failed to mention it is an all-girls Catholic school). Goddard? That’s 1,000 toilet seat covers. It is a preschool with multiple locations, and not a kindergarten option at all.
These same patterns appeared across prompt tests. Claude recommended District 2 middle schools to a District 3 family without noting that admissions are largely district-based. Gemini fixated on The Calhoun School for a theater-loving student without offering meaningful alternatives. We could go on.
To be fair, the bots occasionally provided glimmers of useful insight. Kudos to Claude for its explanation that the contrast between Trinity and Dalton “is one of emphasis and tradition rather than absolutes.” But, as a parent conducting genuine research, it would be nearly impossible to distinguish fact from fiction– or, to use the technical term, hallucination.
The AI School Consultant:
Why It Falls Short.
We have some best guesses as to why AI is so flawed as an educational consultant. It is the same reason this process is difficult in the first place– the right information is hard to find. When we asked ChatGPT about P.S. 6, a highly regarded public school on the Upper East Side, one of its first selling points gave us pause: “Known for Exceptional Writing Instruction (Teachers College Model).” P.S. 6 does not use the Teachers College model. In fact, very few schools do since the DoE’s 2024 overhaul of the curriculum. We clicked the source: a 2009 Time Out New York article. These bots are working with what is readily available online, and that information is often incomplete, outdated, or missing the nuance families actually need.
But AI’s biggest limitation is not its sourcing. It is its lack of humanness. In every prompt we tested, AI failed to ask meaningful follow-up questions or acknowledge when it didn’t have a clear answer. It didn’t share telling anecdotes. It didn’t provide examples of coursework in action. Most importantly, it did not, and could not, get to know a child or uncover a family’s educational philosophy and values.
School culture. Inspiring teachers. Homework load. Administrative responsiveness. Student stress levels. The subtle differences between two schools that look nearly identical on paper. These are the insights that shape a child’s experience, and they are deeply human. They are best uncovered through tours and conversations– with educators, school experts, current families, and those who have gone through the process before. There is no replacing that.
Confession: we are not on the luddite side of the AI-fear spectrum. AI has organized our spreadsheets, helped us design slide decks, and summarized volumes of Chancellor's Regulations from the DoE archives. There is a lot that AI can do. Yet, despite our prompting, AI can’t seem to do the actual work of an educational consultant– partnering with families to make informed decisions about their child’s schooling. These bots are evolving quickly, and, believe us, we will be the first to know when they can do our job. But, for now, when it comes to making sense of options, identifying true fit, and feeling confident that you are focusing on what matters most (and not getting distracted by the candles and toilet seat covers), leave that to the humans.
