School is Back in Session! So is AI.

Aug 18

Every August now brings the same anxious headline: kids are back in school, and so is the worry that AI is doing their thinking for them. A student opens a chatbot, pastes in the prompt, and hands in a polished essay or a fully worked problem set without ever wrestling with the material. Schools respond by banning devices, running detection software, or retreating to in-class blue books — all of it a version of the same instinct: treat AI as an intrusion to be kept out of the room.

The data suggests that instinct is a few years too late. Students, teachers and parents need Conversational AI and Conversational Intelligence skills, now. Here is why -

The numbers behind the panic

‍AI use among students isn't a fringe behavior anymore — it's the default. According to Pew Research Center's February 2026 survey of U.S. teens, 54% have used a chatbot for schoolwork, and 10% now do all or most of their schoolwork with AI's help. RAND's tracking of student AI use tells the same story from a different angle: usage jumped from 48% to 62% of students between May and December of 2025 alone, with 71% now using at least one AI tool for something school-related.

The concern isn't just that students are using AI — it's what they, and the adults around them, think it's doing to them. Pew found that 59% of teens say students at their school use AI to cheat at least sometimes, and among teens who've used a chatbot for schoolwork themselves, 76% think that kind of cheating is routine. RAND found something arguably more telling: 67% of students now believe AI is actively harming their own critical thinking skills, up from 54% earlier in the same year — and the students most convinced of the harm (78%) are the ones who use AI the least.

Teachers are, if anything, more alarmed. A national NPR/Ipsos poll of 545 K-12 teachers, conducted in the spring of 2026, found that 54% believe AI is making it harder for students to develop critical thinking skills, and 55% see it as mostly a shortcut for avoiding real work. Principals are even blunter: in an EdWeek survey, 87% said AI tools could prevent students from developing critical thinking, and 100% said they were worried about cheating. Perhaps the most striking number in that same NPR/Ipsos poll: 59% of teachers say AI is actively eroding the level of trust between them and their students.

That last statistic is the one worth sitting with. Cheating detection is a symptom. Erosion of trust is the disease.

What teachers are actually saying

The statistics get sharper when you hear how individual teachers describe the problem. Josh Kauffman, who teaches English at a virtual school in Alabama, put it bluntly to NPR:

"I would rather deal with all of your typos... than wonder how much you're standing on other people's shoulders."

Christa Corricelli, a special education teacher near Boston, framed it as a long-term erosion rather than a one-time offense — worrying that students' critical thinking skills will "atrophy over time" if they aren't self-motivated to do the harder cognitive work themselves. And Michele Naber, a California biology teacher, told NPR that her real job now is teaching students not to take AI output "literally" — to push them to "figure out things for themselves, not rely on software."

Notice what none of these teachers are describing: a detection problem. They're describing a diagnosis problem. They can't tell, from the artifact alone, what a given student actually understands anymore.

Meanwhile, students see a different picture of the same moment. Jessica Howell, a vice president at the College Board, described the divide to EdWeek this way:

"I worry that there's a little bit of a perception gap [with] the students thinking 'this is grand!'"

That gap is visible in the numbers, too. Teachers rank cheating as one of their top fears about AI; students, when asked directly, rank it much lower — only 42% of teens in that same survey named cheating as a downside of AI at all, well behind concerns about memory and overreliance. Two groups, staring at the same technology, worried about almost entirely different things.

Rethinking the response

If AI use is now the default rather than the exception, the realistic question isn't how to keep it out of the classroom — 78% of teachers themselves now say responsible AI use should be taught in schools, per that NPR/Ipsos poll, and yet only a third report their school has any formal policy on it. The question is how to make sure students come out of school able to use AI well, and how to make sure teachers can still tell, underneath the AI-assisted output, whether real learning happened at all.

That's really two separate literacies, and they aren't the same skill wearing two hats — one is about talking to a machine, the other is about reading a human being.

The first literacy is conversational AI. This is the technical term for chatbots and assistants like ChatGPT or Gemini — AI systems built to hold a natural, back-and-forth conversation with a person rather than just return a search result. The skill students need isn't just access to conversational AI, it's fluency in using it well. A student who can only paste a question and copy the answer hasn't learned to use the tool — they've outsourced to it. A student who knows how to interrogate an AI's answer, ask it to show its reasoning, push back when something looks wrong, and treat it as a sparring partner rather than an oracle is developing something closer to real fluency. That's teachable, directly, the same way research skills or source evaluation once were.

The second, harder literacy is conversational intelligence. This is a human skill, not a technology — the ability to read what's actually happening inside a conversation as it unfolds: how meaning is being shaped, how a response lands, what's really being communicated versus what's merely being said. It's the capacity to sense, in real time, whether the person across from you genuinely understands what they're saying or is just repeating something fluent-sounding. For generations, the artifact a student handed in — the essay, the problem set, the lab report — did that diagnostic work for teachers automatically; it was a reasonably reliable proxy for understanding. That proxy is breaking down, and no amount of AI-detection software is going to fully restore it; detection is already a losing arms race, and the teachers quoted above aren't describing a software gap, they're describing a conversational-intelligence gap. What still works is a few well-aimed follow-up questions — in the hallway, in office hours, in a quick check-in after class — that can surface in minutes what a stack of essays can't: can this student explain the idea in their own words, extend it to a new case, defend a claim under mild pressure?

Most teachers were never trained to run that kind of diagnostic conversation at scale, because they never had to. The paper used to do the work. Now it can lie, convincingly and at no cost to the student — and the conversation has to pick up the slack.

Where Conversational Intelligence and Clarion Fit

Getting better at that second literacy isn't just a matter of teacher willpower — it's a skill, and skills can be supported by tools. CivilTalk's stated mission is "to help individuals, teams, educators, and organizations better understand the conversations that shape their lives and work," built around what it calls conversational intelligence — "the ability to recognize and navigate the dynamics of a conversation as they unfold: how meaning is shaped, how responses land, and what's created between people in real time." Its product, Clarion CI, is described as "an AI conversational intelligence agent that joins your meetings, analyzes conversations in real time, and delivers insights that help you communicate better, decide faster, and lead with confidence."

As CivilTalk currently positions it, that's aimed at organizational meetings and leadership conversations, not classrooms specifically. But the underlying idea maps onto teaching more directly than it might first appear. A teacher doesn't have one conversation with a student — they have dozens, scattered across a semester: in class discussion, office hours, hallway check-ins, comments on drafts. Individually, each one is a snapshot. Looked at together, they're a record of whether a student's understanding is actually deepening or just producing better-dressed answers. An agent built to sit in on conversations and surface what's really happening in them — rather than just what was said — is exactly the kind of tool that could help a teacher notice, across all those exchanges, where a student's explanations are getting genuinely stronger, and where they're staying suspiciously flat despite increasingly polished submitted work. That flatness is the fingerprint of a student leaning on AI instead of learning from it.

None of this replaces the teacher. It's closer to what game film does for a coach: the conversations were always the real signal. A tool like this just makes them legible enough to act on.

What conversational intelligence actually looks like

"Conversational intelligence" can sound abstract until you see it broken into parts. CivilTalk's own CI Model — built on the foundations of emotional intelligence but extended specifically to how conversations unfold between people — names 20 distinct skills, organized into four clusters, that together make someone genuinely good at reading and steering a conversation rather than just having one.

Self-awareness comes first, and it's easy to skip. Before a teacher (or a parent) can accurately read a student, they need self-assessment — an honest check on their own assumptions walking into the conversation. It's tempting to treat a suspiciously polished essay as a verdict already reached; self-assessment is what keeps that from turning into an accusation before a single question has been asked.

Self-management is what keeps the conversation useful once it starts. Self-control means not reacting to a defensive or evasive answer with visible frustration, which is usually what makes a student clam up for good. Accountability means the adult follows through — a check-in that happens once and is never revisited teaches a student that the performance, not the understanding, was what mattered. And trustworthiness is the skill that sits underneath the "erosion of trust" statistic cited earlier: a student only admits how they actually used AI if they believe the adult asking will respond fairly rather than punitively.

Social awareness is the listening half of the equation. Active listening is the difference between hearing a correct-sounding answer and hearing the gaps underneath it — the over-rehearsed phrasing, the hesitation when asked to go one layer deeper, the vocabulary that doesn't match how this particular student normally talks. Empathy is what makes a student willing to let those gaps show in the first place, instead of doubling down to protect themselves from an adult who already seems to have decided the verdict.

Relationship management is where the diagnostic conversation turns into something that actually improves the student's learning instead of just catching them. Developing others reframes the whole exchange from "prove you didn't cheat" to "let's find where your understanding is thin and build it up" — the same conversation, aimed at growth instead of punishment. Conflict management matters for the harder moments, when a grade or an accusation is on the table and the conversation could easily become adversarial instead of instructive.

None of these are AI-specific skills. They're the same relationship-building skills that make any teacher, parent, coach, or mentor … classroom more effective. What's changed is how badly schools now needs them, at scale, in a moment when the paper trail can no longer be trusted to do this work on its own.

Parents belong in this conversation too

Teachers aren't the only adults trying to read whether a student's understanding is real. Parents are having a version of this conversation at home — or, often, not having it at all.

Pew Research Center's February 2026 survey found a real blind spot here: 64% of teens say they've used a chatbot, but only 51% of their parents believe they have, and roughly 40% of parents say they've never talked with their teen about AI use at all. That's not a minor gap — it means a large share of parents don't know how deeply AI is already woven into their kid's homework routine, let alone whether it's helping them learn or cheat.

A car ride or the dinner table doesn't come with a grade attached, which makes it a lower-stakes place for a kid to be honest about how they actually did an assignment.

But a low-stakes setting only helps if the adult has the same Conversational Intelligence skills that a teacher needs. Active listening and empathy turn "did you use AI on this?" — a question almost guaranteed to get a defensive one-word answer — into "walk me through how you got to this answer," which invites an actual explanation. Self-control keeps a parent from reacting to an uncomfortable answer in a way that guarantees the next one is a lie. Trustworthiness is what makes a kid willing to say "I used AI to get started, and then I was stuck" instead of hiding it entirely. And developing others is the difference between a conversation that ends in a punishment and one that ends with the kid actually a little better at the material than they were an hour before.

None of this asks parents to become AI-detection experts or homework police. It asks them to have the kind of conversation they're often already inclined to have with their kids — just with more intention about what they're listening for. Teachers can build the habit of this kind of check-in during the school day; parents are positioned to reinforce it every evening.

That parental blind spot isn't new. Long before AI, researchers in digital literacy and digital citizenship were already describing this same dynamic: parents can see the outline of a teen's online life — the apps installed, the hours logged, the general shape of it — without ever really seeing inside it. A chatbot a teen confides in or leans on for homework is just one more layer of that private world, and arguably a harder one to see into than a text thread or a social feed, since the whole exchange lives inside a single, personalized conversation no one else is copied on.

The answer isn't more surveillance; it's building conversational intelligence to enable. - what's often called digital agency — in the teen, and in the teacher and parent guiding them.

The idea, familiar from decades of media-literacy work, is that no one should be a passive consumer of whatever a feed, or now a chatbot, hands them. We should all learn to question it, redirect it, and decide deliberately how to use it instead of just absorbing it. That is the exact posture this piece is arguing for with conversational AI: not banning it, not trusting it blindly, but exercising real agency over how it gets used. Technology, in the end, is just the tool — it's the humans on either end of the conversation with conversational intelligence skills who decide what it becomes, and, increasingly, who have to decide how to use it well.

The actual back-to-school story

The panic over AI and cheating is really a panic over losing a signal teachers and parents used to get for free through tests and papers. The fix isn't to fight a tool students are going to use for the rest of their working lives — it's to build the fluency in conversational AI they need to use it well, and to invest at least as heavily in the conversational intelligence the kids and adults around them need to see past it. Students need to learn to talk to AI. Teachers and parents need to get better at reading students. Given that 59% of teachers already say trust is eroding, a third of parents aren't even sure whether their own kid uses AI, and most families and classrooms have no shared policy to work from, that second skill isn't a nice-to-have this year. It's the one thing standing between "AI helped me learn" and "AI did it for me" — and right now, hardly anyone has a reliable way to tell the two apart.

Critical thinking is the outcome that connects these two literacies.

The goal is not simply to teach students how to use AI, nor is it simply to help adults determine whether AI was used. The goal is to develop students who know how to question an AI-generated answer, test its assumptions, examine its evidence, compare it with other sources, explain the issue in their own words, and ultimately decide for themselves what they believe and why.

CivilTalk is the Conversational Intelligence Company.

We make conversations visible.

We know Better Conversations lead to Stronger Relationships. Stronger Relationships lead to Better Outcomes.

📅 Schedule a demo: https://www.civiltalk.com/schedule-call#clarion-demo

Sources

CivilTalk