From the K–12 edition of November 10, 2026
Stanford SCALE AI Hub for Education: The Evidence Base on AI in K-12, a November 2026 review of 52 causal studies
Stanford University SCALE Initiative, AI Hub for Education · K-12 education · United States and international studies
- Publisher
- Stanford SCALE Initiative
- Original publication
- 2026-11-10
- Source retrieved
- 2026-11-11
What happened
An update to the March 2026 review adds 32 strong causal papers published October 2025 to June 2026, for 52 in total. Reviewers conclude AI is more likely to support learning when students use it to understand material than to produce work, but neither use consistently improves learning, and offloading tasks can reduce learning. AI can improve learning experience and confidence without improving outcomes, and teacher-facing AI can cut planning time or raise feedback quality but one new study found it reduced student motivation and confidence.
Why it matters
Gives K-12 districts a current, independent synthesis for deciding where to allow student-facing AI and where to prefer educator-facing, structured uses. It supports requiring learning-support design rather than open-ended answer generation, and measuring outcomes instead of satisfaction.
Evidence and measured results
Review of randomized controlled trials and quasi-experimental studies drawn mainly from arXiv preprints in the AI Hub Research Repository. Of 868 papers added in the review period, 201 were RCTs or QEDs, 97 advanced to human review and 32 met the strong-causal standard (28 student-facing, 4 educator-facing). Five new student-facing studies were in U.S. K-12 settings. A language model did preliminary screening before Stanford researchers reviewed prioritized papers against a codebook based on What Works Clearinghouse standards.
Limitations and uncertainty
The authors say findings should not be treated as broadly generalizable across U.S. K-12 settings; the U.S. K-12 share of studies is limited. Many sources are non-peer-reviewed preprints, the repository is arXiv-weighted so relevant journal studies may be missed, and LLM-assisted screening may have excluded relevant studies. Some included studies involve undergraduates or adults rather than school-age students. This is a secondary synthesis; effect sizes are not pooled.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-10; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- District leaders must decide which AI uses to allow for students and teachers with sparse, mixed, mostly non-U.S. causal evidence.
- Stakeholders
- Superintendent and cabinet
- Chief academic officer
- Chief technology officer
- Curriculum and instruction leads
- Teacher union or professional learning leads
- Discovery questions
- Which student uses are permitted today: explanation and practice support, or producing finished work?
- What outcome measures beyond satisfaction would you require before scaling a tool?
- Where have teachers used AI for planning or feedback, and has student motivation been checked?
- Value hypothesis
- A scoped use-case policy and measured pilot can help a district avoid uses the review links to reduced learning while keeping learning-support and teacher-feedback uses.
- Potential engagement
- Bounded AI use-case and evidence review workshop for one district, ending with a pilot measurement plan.
- Unsupported claims
- Guaranteed achievement gains
- That education-specific tools outperform general chatbots
- That results from non-U.S. or undergraduate samples will hold in a given U.S. district
Pre-sales engineering
Role takeaway
- Fit
- Relevant to any district evaluating student-facing or teacher-facing generative AI; most useful for configuration and measurement requirements.
- Architecture and integration
- Require controls to constrain answer-giving, adaptive sequencing, and exportable usage logs integrated with the learning platform and rostering.
- Prerequisites
- Rostering and SSO
- Usage logging and export
- Baseline assessment data
- Defined use cases by grade band
- Constraints
- Evidence is mostly preprints and heavily non-U.S.
- Studies vary in tool design so findings are not tool-specific
- Includes some postsecondary populations
- Security
- The review does not address privacy or security; run separate data-protection, vendor and safety review before any pilot.
- Proof of value
- Compare a coaching configuration with an answer-giving configuration on a delayed unaided assessment
- Track engagement with support features
- Collect student motivation and confidence measures
Delivery
Role takeaway
- Work
- Run a pilot built around defined learning-support uses, teacher guidance and a pre-agreed measurement plan.
- Dependencies
- Policy decision on permitted uses
- Baseline and delayed assessments
- Teacher time for training
- Ownership
- District curriculum and instruction lead with technology office support.
- Skills and adoption
- Teachers need practical guidance on when AI feedback or planning help fits and how to keep students doing the thinking.
- Governance checkpoints
- Use-case approval before launch
- Mid-pilot check of student motivation and confidence
- End-pilot review of delayed outcomes before scaling
- Acceptance criteria
- Proposed: delayed unaided assessment shows no decline versus comparison group
- Proposed: at least an agreed share of teachers complete guidance training
- Proposed: usage logs reviewed on an agreed cadence
- Risks
- Positive satisfaction mistaken for learning
- Over-generalizing from non-U.S. or adult studies
- Teacher-facing tool unintentionally lowering student motivation
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Tool behavior matters more than hosting model: coaching, adaptive sequencing and tool-use guidance show promise, so districts should prefer configurable tools that can restrict answer-giving and expose usage logs, and should not assume education-specific tools beat general chatbots.
Governance
Who approves, reviews and stays accountable for outcomes?
Supports use-case-level policy that distinguishes learning support from work production, requires outcome evidence in procurement, and sets teacher guidance on when AI feedback or planning help is appropriate.
Security and privacy
What data, permissions and controls need testing?
The review does not evaluate data privacy or security; privacy, student-data and safety controls must be assessed separately from efficacy evidence.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
The review links effects to how educators use AI and what tasks they use it for; teacher training and judgment are central, and one study found lower-performing teachers' students lost achievement with a broad support chatbot.
Procurement
What should contracts, pricing and exit terms secure?
Ask vendors for independent outcome evidence on delayed or unaided measures, configurability of coaching versus answer modes, and usage reporting, rather than relying on satisfaction data.
Operating model
Which teams own the service once it runs?
District leaders should assign ownership for use-case policy, measurement and teacher guidance before expanding access.
Publication history
- 2026-11-10K–12 · Issue 653 resources
Stable resource ID: stanford-scale-evidence-base-ai-k12-november-review-52-causal-studies-20261110