K-12Issue 545 records
K–12
A JED Foundation survey of more than 5,500 middle and high school students found 57% turn to AI when sad, stressed or lonely, and some uses were linked to worse mental health, though causation is unclear; NPR's coverage adds early signs that companionship use deepens loneliness in middle schoolers. Fayetteville, Arkansas, is merging a state-required AI policy that bars AI as the sole basis for grades and requires tools that avoid emotional relationships with students. Google reports self-reported classroom gains from Gemini lesson co-design in 31 Italian schools.
- Evidence records
- 5
- Cross-source patterns
- 2
- Evidence classes
- 3 independent reporting1 independent research1 vendor claim
- Outcomes
- 3 emerging2 cautionary
- Source freshness
- 5 new this fortnight
- Research completed
- 2026-10-31
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Student emotional use of chatbots is reaching district AI policy
JED's school-based survey found 57% of middle and high school students use AI when emotionally struggling, with some uses associated with worse mental health, and NPR's coverage adds a UNC finding that middle schoolers using chatbots for companionship grew lonelier. Fayetteville's state-driven draft policy already requires tools that limit human-like behaviour and avoid emotional relationships with students. The evidence is associational, but districts are starting to write emotional behaviour into tool requirements.
Operating questionFor each student-facing AI tool, can the district restrict companion or emotional-support behaviour, and what happens when a student signals distress?
Supporting evidenceThe Jed Foundation (JED)NPR (byline Rhitu Chatterjee)Fayetteville Public Schools (reported by the Northwest Arkansas Democrat-Gazette)
Draft district policies require a human decision on anything consequential for students
Fayetteville's draft bars AI as the sole basis for decisions such as grades, and Sarasota's proposed rules say AI cannot make important decisions about students without human review. Both drafts also pair this with limits on student use (teacher disable controls in Fayetteville, disclosure in Sarasota). Neither is final.
Operating questionWhich AI-assisted workflows touch grades, placement or discipline, and where is the documented human decision in each?
Supporting evidenceFayetteville Public Schools (reported by the Northwest Arkansas Democrat-Gazette)Sarasota County School Board (reported by Lakewood Ranch Beacon)
Evidence ledger
Full record · every source keeps its link and limitations
JED Foundation: survey of 5,560 middle and high school students in 21 U.S. schools finds 57% use AI when emotionally struggling, with certain uses linked to worse mental health
JED's Student Survey, collected October 2025 to April 2026 from more than 5,500 middle and high school students in 21 U.S. schools, found students use AI for schoolwork (58%), when sad, stressed, lonely or overwhelmed (57%), for relationship or problem-solving advice (28%) and for companionship (12%). Using AI for companionship, advice, or while emotionally struggling was associated with greater anxiety and depression symptoms, loneliness and lifetime suicidal thoughts and attempts; homework use was associated with higher loneliness and history of suicidal thoughts. Middle schoolers, girls and several historically marginalized groups were more likely to use AI when struggling.
Why it matters, evidence and limitations
- Why it matters
- Directly K-12 and newly current: released Oct 27 and nationally reported Oct 30, it gives districts school-based data on how students use chatbots emotionally, which bears on which AI tools schools enable and how counseling and safety teams respond.
- Evidence and measured results
- Cross-sectional survey of students in schools enrolled in JED programs, weighted for non-response and analysed with mixed models adjusted for gender, race/ethnicity and grade (per the report landing page). Associations only; not nationally representative.
- Limitations and uncertainty
- Cross-sectional and not nationally representative, so causation cannot be inferred, as JED states. Schools were already JED partners. This run read the press release and report landing page, not the full report PDF.
NPR: two new surveys find nearly two-thirds of tweens and teens use AI chatbots; JED finds 90% of 5,500 school students have used one, and a UNC study links companionship use to growing loneliness
NPR reports two new surveys: Common Sense Media found nearly half of 9-to-12-year-olds have used an AI chatbot or companion and nearly three-quarters have by age 13; JED found 90% of 5,500 students aged 9 to 17 in schools had used one, and nearly two-thirds use it for schoolwork and when stressed or lonely. JED's Kayla Anderson cautions the data do not show causation. UNC psychologist Anne Maheux, not involved in the surveys, describes a soon-to-be-published study of middle schoolers in one state finding that those using chatbots for companionship grew lonelier over time. A student said AI use was introduced through school in eighth grade.
Why it matters, evidence and limitations
- Why it matters
- K-12 and dated tonight: national coverage brings the JED school-based data to a broad audience and adds an outside researcher's longitudinal finding about middle schoolers, which districts weighing student-facing chatbots will be asked about.
- Evidence and measured results
- News report quoting the JED report author, an independent developmental psychologist and students; summarizes two surveys and a forthcoming study without detailed methods.
- Limitations and uncertainty
- Reporting, not original analysis. The '90%' any-use figure attributed to JED does not appear in JED's press release or report landing page, which this run opened; it was not verified. The UNC middle-school study is unpublished, and the Common Sense survey was not opened.
Google and Salesian schools in Italy: yearlong study across 31 schools reports teachers co-designing lessons with Gemini cut lectures 38% and raised student-reported engagement, based largely on self-reports
Google reports a yearlong study with the Salesians of Don Bosco across 31 schools, involving 1,600 primary, secondary and vocational teachers representing over 24,500 students aged 13-18, comparing lessons planned with and without Gemini for Education using classroom activities and self-reports. Teachers reported feeling more creative when co-designing with Gemini (79% vs 30%); front-of-class lectures dropped 38%; student-led collaborative learning nearly tripled; student-reported engagement rose from 77% to 84%; teacher reliance on their own energy to sustain engagement fell 66%. The same post covers the UK Eedi trial, already archived, and notes those gains did not appear on Star standardized tests.
Why it matters, evidence and limitations
- Why it matters
- K-12 relevant and new on Oct 29: a large teacher-facing deployment study about lesson planning and classroom dynamics, the most common district AI use case, distinct from the student tutoring trial archived yesterday.
- Evidence and measured results
- Vendor blog summary; methods described only as classroom activity collection and self-reports with and without Gemini. No learning outcomes, comparison design details, or sample sizes per measure are given in the post; a separate implementation report exists but was not opened.
- Limitations and uncertainty
- Vendor-authored summary of a vendor-partnered study in a single Catholic school network in Italy; self-reported measures; no achievement outcomes; implementation report not inspected. Transfer to U.S. public districts is uncertain.
Fayetteville, Arkansas school board reviews draft AI policy merging a state-required DESE policy under Act 848, barring AI as sole basis for grades and requiring tools that avoid emotional relationships with students
On Oct. 29 the Fayetteville School Board first reviewed an updated AI policy that merges the district's earlier policy with a detailed policy the Arkansas Department of Elementary and Secondary Education required districts to adopt, under Act 848 of 2025. The draft requires AI programs to align with the state K-12 cybersecurity policy and the Student Data Vendor Security Act; bars confidential student or personnel data from public or unapproved tools; bars AI as the sole basis for decisions such as grades; requires tools that let teachers disable AI for assignments, moderate unsafe interactions, limit human-like behaviour and avoid emotional relationships with students. The district counsel still has questions about contract requirements; a vote is expected in November.
Why it matters, evidence and limitations
- Why it matters
- K-12 and current: shows a state-mandated district AI policy translating into concrete product requirements (teacher disable switch, no emotional relationships with students) that affect which tools Arkansas districts can buy.
- Evidence and measured results
- Local news report of a board meeting, quoting district counsel and summarizing the draft; the draft and the DESE policy were not opened.
- Limitations and uncertainty
- Secondary reporting; part of the article contextualizes with national research rather than the policy. Draft text and DESE policy not inspected; final adoption pending.
Sarasota County, Florida: students and parents urge school board to limit classroom AI as proposed rules requiring disclosure and human review await a final vote
Nearly a dozen Sarasota County students and parents urged the school board to limit AI in classrooms, after the board voted in mid-October to advertise proposed AI rules. Under the proposal, students could use AI for brainstorming and feedback but not to replace original work and must disclose use; teachers could use AI for lesson planning, but it could not make important decisions about students without human review. A Booker High student said teachers as well as students rely on AI too heavily. Board member Bridget Ziegler asked 'at what cost?' No final vote date has been announced.
Why it matters, evidence and limitations
- Why it matters
- K-12 and current: shows student and parent pushback shaping a large Florida district's AI rules before adoption, including concern about teacher use.
- Evidence and measured results
- Short local news item based on WWSB reporting; quotes a student and a board member; cites a RAND survey for context. Policy text not opened.
- Limitations and uncertainty
- Secondary reporting of another outlet's story; brief; policy text not inspected; final rules may change.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Independent research
- Research conducted outside the implementing organization.
- Vendor claim
- A supplier-provided assertion that has not been upgraded to independent evidence.