Education · Issue 01 ·
K–12
Four newly archived sources examine fall K12 implementation and tutoring evidence: district adoption monitoring, independent research on organizational constraints, a supplier-authored supervised tutoring trial, and Stanford practice guidance. Usage growth is not learning evidence; promising same-day trial results do not establish durable gains or autonomous tutoring effectiveness. Three cross-source patterns support bounded evaluation and explicit human ownership. Sources span December 2025–August 2026; none is presented as newly published September 6 news. Long-term learning, accessibility outcomes and independently verified operating costs remain gaps.
- Evidence records
- 4
- Cross-source patterns
- 3
- Evidence classes
- 1 independent research1 government evaluation1 academic research1 standards or public-body guidance
- Outcomes
- 2 mixed1 emerging1 cautionary
- Source freshness
- 2 undated1 older, newly relevant1 recent
- Research completed
- 2026-09-07
Choose a role to see its takeaway beside every record in the ledger.
Synthesis · Lighthouse Advisory interpretation
Patterns across the evidence
Use adoption as implementation evidence, then test learning separately
East Maine's user growth tracks rollout; CRPE identifies narrow evaluation practices; the LearnLM study measures a specific same-day transfer outcome. These measures answer different questions and cannot substitute for delayed independent assessment.
Operating questionWhich adoption, staff-effort and independent-learning measures must be reported separately before renewal?
Supporting evidenceEast Maine School District 63Center on Reinventing Public Education, Arizona State UniversityLearnLM Team, Google and Eedi
Human involvement is part of the tested tutoring service
LearnLM's trial includes message-level human approval; Stanford distinguishes tutoring models by human involvement. Removing review or relationship support changes the intervention and requires fresh evidence rather than borrowing the original result.
Operating questionWho reviews, intervenes and maintains student support when the AI service is unavailable or inappropriate?
Supporting evidenceLearnLM Team, Google and EediSCALE Initiative and National Student Support Accelerator, Stanford University
Policy rollout needs operational owners and review checkpoints
CRPE's organizational findings and East Maine's planned classroom rollout show why distributing rules is only one implementation step. Stanford's privacy and support guidance adds a concrete tutoring checkpoint. This is a proposed operating response, not evidence of causal benefit.
Operating questionWhich named curriculum, school, IT and privacy owners must sign off before student access expands?
Supporting evidenceCenter on Reinventing Public Education, Arizona State UniversityEast Maine School District 63SCALE Initiative and National Student Support Accelerator, Stanford University
Full record · every source keeps its link and limitations
Evidence ledger
Early-adopter district study exposes evaluation and change-management gaps
CRPE finds more coordinated AI adoption alongside weak evaluation, delayed family involvement and procurement friction. Greater technical fluency does not itself establish instructional transformation.
Why it matters, evidence and limitations
- Why it matters
- Useful context for fall district implementation reviews; newly added evidence, not September breaking news.
- Evidence and measured results
- October–November 2025 data: 45 of 119 identified early-adopter districts responded; interviews covered 14 districts and 29 participants. The researchers coded surveys, interviews and documents. There is no causal baseline for educational benefit.
- Limitations and uncertainty
- Purposive, nonrepresentative sample; technology leaders predominate. Cross-year samples differ. Reported workload benefits are not independently measured. PDF text, methods and limitations were inspected; one table screenshot failed, so no table-derived category percentages are used.
East Maine reports AI platform uptake while classroom guidance rollout remains planned
The district reports approved AI platform users rising from 391 to 450 after staff training. Its August entries list broader classroom AI guidance and additional professional learning as not started, with June 2027 targets.
Why it matters, evidence and limitations
- Why it matters
- A concrete current-school-year example of translating staff access into classroom practice. This source was absent from the checked archive.
- Evidence and measured results
- Official administrative progress table reports a 15% increase in users, a rounded relative change. This is implementation monitoring, not an independent impact evaluation; no comparison group, user-count methodology or AI-attributable learning assessment is supplied.
- Limitations and uncertainty
- Training preceded the reported increase but causation is untested. Targets and status fields must not be read as completed work. Exact publication and event days are unavailable.
Supervised LearnLM mathematics trial offers short-term promise with uncertain transfer advantage
A supplier-authored trial supports a bounded human-supervised tutoring workflow; it does not establish autonomous tutoring effectiveness or durable learning.
Why it matters, evidence and limitations
- Why it matters
- Older research newly relevant through August tutoring scrutiny; no archive match found. Transfer requires local curriculum, staffing and safeguarding validation.
- Evidence and measured results
- 165 students aged 13–15, seven weeks: student assignment to hints or tutoring, then session-level assignment to human or supervised LearnLM. Baseline-adjusted Bayesian next-topic success was 66.2% versus 60.7%; the 5.5-percentage-point difference had a 95% credible interval of -1.4 to 12.4. Transfer was same-day.
- Limitations and uncertainty
- Preprint, provider involvement, one platform and subject; every draft was reviewed. Session crossover prevents estimating cumulative effects; throughput was not rigorously measured. Selected transfers exclude students not continuing that day. Long-term retention and US applicability remain unproven.
Stanford tutoring brief distinguishes educator assistance from unsupervised student AI
The brief distinguishes human-led AI-assisted tutoring from AI-led and AI-only models, arguing that evidence becomes thinner as sustained human involvement falls.
Why it matters, evidence and limitations
- Why it matters
- Relevant to fall district purchasing and tutoring program design. August publication is new to the checked archive, not a September event.
- Evidence and measured results
- Synthesis of tutoring research plus interviews with providers, developers and researchers; no new pooled effect estimate or representative interview sampling frame is reported.
- Limitations and uncertainty
- Practice guidance, not product certification or a new causal evaluation. Provider interview contributions are disclosed. One link attached to a two-trial claim opened a different UK study; this edition does not repeat that numerical claim.
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 research
- Research conducted outside the implementing organization.
- Government evaluation
- A public body’s measured evaluation or documented pilot.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.