Organizations that work with youth are being asked to say something about artificial intelligence, and most would rather say something accurate than something alarming. That turns out to be harder than it sounds. The research on youth and generative AI is young, the headlines move faster than the evidence, and a session built on fear tends not to reach teens at all.
Justin and I have spent the last two decades presenting to students, parents and guardians, educators, clinicians, and law enforcement, and the questions we are asked now are different from the ones we were asked even two years ago. This page sets out how we would evaluate a speaker on this topic, written for the principal, conference planner, or program director who has been handed the task and a short deadline.
AI Sessions Should Rest on Data
A great deal of what circulates about youth and artificial intelligence traces back to media coverage instead of research. Some of it is accurate. A fair amount of it is extrapolated from a single case, or from a study with a sample too small to support the claim being made.
A speaker who works from data will tell an audience where the evidence is thin, and will say clearly when something is a reasonable concern that has not yet been measured. In our national study of teen AI chatbot use, for instance, 47.1% of teen chatbot users reported at least one risky or harmful interaction, and the risks were greatest among 13-year-olds. That is a figure that holds up when a board or a leadership team asks where it came from, and it is also a figure with a method behind it that anyone can read.
Anyone evaluating a speaker should ask what their claims rest on. A good answer names studies. A weak answer names news stories.
Sessions Should Reach Teens Without Fear Tactics
Students are often quite weary of the one-sided advocacy they hear from adults about technology. A session that opens by telling them these tools are dangerous and they should be careful will lose the room in the first few minutes, and the students who most need the material are usually the first to disengage.
What tends to work instead is treating teens as capable of reasoning about their own lives. That means acknowledging what is genuinely useful about these tools, being specific about where the harm actually occurs, and giving students something to do rather than something to fear. Moreover, a speaker who is pro-technology and pro-youth has far more credibility with a student audience than one who arrives sounding like a warning label.
Content Should Be Current
This subject moves faster than almost any other topic an organization will address with youth. A session built in early 2025 would not cover AI companion apps the way the evidence now supports, and would likely miss the sexualized deepfake question entirely.
Hosts should ask when the material was last revised, and what changed. A speaker who cannot answer that specifically is probably delivering a deck that has been sitting unchanged for a while.
Separate Audiences Deserve Separate Sessions
The questions a 14-year-old brings to a session on AI chatbots are not the questions a parent brings, and a faculty audience wants guidance that would lose a student assembly entirely. Most organizations that get real value from a visit plan it as more than a single hour, often pairing a student assembly with a parent and caregiver evening and a faculty session on the same day. The same holds outside of schools, where a clinical team and the families they serve need quite different sessions.
Audiences Outside K-12 Schools
Schools ask for this material most often, though they are far from the only ones asking. The requests I receive now come from hospitals, clinical practices, law enforcement agencies, technology companies, and professional associations, and the session changes quite a bit depending on who is in the room.
Clinicians and mental health professionals generally want the material framed around what surfaces in a session with an adolescent client, including how AI chatbots figure into a teen’s emotional life, what is worth asking about during an intake, and how to respond when a teen describes a sustained relationship with a conversational system. I deliver a session built for clinicians and school counselors on exactly that.
Pediatricians, hospital staff, and children’s advocacy center teams usually want something closer to a briefing, covering what the data can inform about prevalence and risk, what a provider might reasonably raise during a well visit, and how sextortion and image-based abuse schemes present when a family finally comes in. Similarly, law enforcement and child protection professionals need the investigative picture, including how generative tools are being used in harassment, impersonation, and exploitation cases, and which youth the evidence suggests are most at risk. That is a separate session as well.
Technology companies and platform trust and safety teams come at the subject from the other direction. They generally want to know what independent evidence shows about youth experience on their services, and where the research is strong enough to inform a policy decision. Conferences and professional associations are a category of their own, since the audience is usually mixed and the content has to work for educators, clinicians, and administrators at the same time. I give those talks regularly, and I always welcome your input and suggestions so that we optimize our ability to meet the specific needs you have identified.
Attendees Should Leave With Something They Can Use
A session that ends without anything in people’s hands rarely changes what happens the following week. Hosts should ask what attendees take home, whether that is handouts, tip sheets, classroom activities, clinical conversation guides, or discussion prompts for families, and whether those materials cost extra.
Relatedly, it is worth asking whether the speaker created those materials or is pointing to someone else’s. Both can be useful, though a speaker who has built their own tends to know the subject at a different depth.
Kinds of AI Speakers and What Each Is Good For
Speakers on this topic come from several different backgrounds, and the background shapes what an audience actually receives. None of these categories is better than the others in the abstract. They simply fit different goals.
Researchers Who Present
These speakers conduct and publish original research on youth and technology, and they present from their own data. Organizations that need to defend a choice to a board or a leadership team, or cite the material afterward, tend to look here first. Justin and I fall into this category, and our studies are published openly at this site so that anyone can read the underlying work before bringing someone in.
Classroom Technology Trainers
These speakers come from teaching and instructional design, and they are strongest on how artificial intelligence changes daily practice in a classroom. They are often the right choice for an in-service day where teachers want hands-on practice instead of an overview, though they usually address instruction more than student safety.
Digital Parenting Voices
These speakers write and speak primarily to families, and their sessions are built around the decisions parents and guardians are making at home. They can be quite effective at an evening event. It is worth asking whether the speaker works for a company that sells monitoring software, since that shapes the advice given and is worth knowing in advance.
Technologists and Industry Speakers
These speakers come from building the systems rather than from studying their effects. They can explain how the technology works with real authority. However, they are often less equipped to speak to adolescent development, and a student audience may find the material abstract.
Former Investigators and Law Enforcement
These speakers bring case experience, which carries weight with some audiences and can be quite compelling. Hosts should ask how recent the cases are, since investigative experience from several years ago may not reflect how these harms occur now.
Whatever category you are looking at, the vetting advice in our companion piece on assembly speakers applies here as well. Ask for references from organizations similar to your own, hold a call before inviting, and tell the speaker what has actually been happening in your community.
Questions Hosts Ask About AI Speakers
Who is a good AI safety speaker?
Organizations that want the conversation grounded in evidence usually look for someone who conducts their own research on youth and technology. I (Sameer) often present on AI risks and harms to youth, covering AI chatbots and companion apps, sexualized deepfakes, AI-facilitated harassment, and the mental health questions attached to them. I also offer separate AI literacy sessions for educators and school leaders, and versions of the session adapted for clinical, healthcare, law enforcement, and industry audiences.
What should an AI session for students actually cover?
A session for a student audience usually works best when it covers how generative AI actually functions, where the real risks occur, and what students can do when something goes wrong. That typically includes AI chatbots and emotional reliance on them, deepfake and image-based abuse among classmates, catfishing and sextortion schemes that now use generative tools, and the media and AI literacy that helps students think critically about what they encounter.
How is an AI safety session different from an AI in education session?
An AI in education session addresses instruction, including lesson planning, academic integrity, detection tools, and classroom adoption. An AI safety session addresses what happens to youth outside of assignments, including AI chatbot use, deepfakes, harassment, and well-being. Some speakers cover both, though most are considerably stronger at one.
Should students, parents and guardians, and staff attend the same session?
Usually not. Each group brings different questions and needs a different level of technical explanation. Most speakers who work regularly in this space offer separate versions, and many organizations arrange two or three sessions during a single visit.
What should a host ask a speaker?
Hosts should ask what the speaker’s claims rest on, when the material was last updated, what attendees take home, and which audience the session was built for. It also helps to ask for references from organizations of a similar size and population served, and to hold a short call before committing to anything.
How current does the material need to be?
More current than for almost any other topic in this space. The tools youth use changed substantially between 2024 and 2026, and sessions built before AI companion apps became common among teens will miss a significant part of what youth are actually experiencing.
Are these sessions appropriate for elementary students?
They can be, with age-appropriate adjustment. Sessions for lower school audiences generally focus on what artificial intelligence is, why something that sounds human is not a friend, and when to tell a trusted adult about something that felt strange. The AI chatbot and deepfake material that works in a high school assembly is not suitable for younger students.
Does an AI safety speaker only present to schools?
No. Hospitals, pediatric practices, children’s advocacy centers, clinical teams, law enforcement agencies, technology companies, and professional associations all have asked for sessions on youth and artificial intelligence, and the material is adjusted considerably for each audience. We present to all of these, including a session built for clinicians and school counselors and one for law enforcement and child protection professionals.
What does the research say about teens and AI chatbots?
Our national study of US teens found that 47.1% of teen AI chatbot users reported at least one risky or harmful interaction, with risks greatest among 13-year-olds. The full study is published in the Journal of Adolescence, and we have written a plain-language summary of it here.
Should the session address academic integrity and AI detection tools?
That depends on the goal. Academic integrity is a legitimate topic, though it is a different session from one on risks and harms, and detection tools remain unreliable enough that schools should be careful about what they promise staff. An organization that wants both should say so early, since few speakers prepare both as a matter of course.
How do organizations pay for a session like this?
Speakers differ widely in how they handle fees and travel, and an out-of-state visit carries costs a virtual session does not. Where a single organization cannot carry the full amount, a common solution is to split the cost with a neighboring district, a parent-teacher organization, a hospital foundation, or a local prevention coalition, and to combine additional sessions into the same trip.
A Closing Thought
The organizations that get the most from a session on this topic tend to treat it as a starting point. They tell the speaker what has been happening in their community, they plan who owns the follow-up before the visit happens, and they give the material somewhere to go afterward, whether that is an advisory period, a counseling team, a staff meeting, or a newsletter to families.
Artificial intelligence is going to keep changing what youth encounter, and no single hour is going to settle it. A good session gives a community a shared vocabulary and a reasonable place to start. Justin and I always welcome your thoughts and experiences in this area, and the comments below are open.
The post What the Best AI Safety Speakers Do appeared first on Cyberbullying Research Center.