Today the evidence caught up with the announcements. Stanford published a review arguing the strongest results in AI tutoring come from tools that help human tutors rather than replace them — and the engagement numbers underneath it are brutal. Alongside it: a free K–12 AI literacy curriculum landing as seven in ten American teenagers already use AI for schoolwork, NASA banning generative AI from a flagship student competition, and fresh district data showing schools have written AI rules far faster than they have hired anyone to run them.
1. Stanford: the strongest AI tutoring evidence is for tools that help human tutors
Source: EdTech Innovation Hub, on Stanford’s SCALE Initiative · 27 August 2026
Stanford’s National Student Support Accelerator and AI Hub for Education published a review, AI Tutoring is Not a Monolith: What We Actually Know, splitting “AI tutoring” into distinct models by how much human involvement they keep. Its central finding is that live, human-led tutoring is still the model with the strongest evidence behind it, and that fully automated tutoring has not built a comparable base.
The engagement figures are the part worth reading twice. Across two randomised controlled trials in two school districts, between 40% and 47% of students never used the AI tutoring platform at all; those who did averaged roughly two to five minutes a week. A separate study of 181,000 students on supplemental maths software found only 5% reached the recommended 30 minutes of weekly use, while 41% never logged in. Where AI supported a human tutor with real-time coaching, students were four percentage points more likely to master a topic — nine points for those taught by lower-rated tutors.
Availability is not usage. That gap is where most AI tutoring money quietly disappears.
2. Common Sense Media completes its free K–12 AI literacy curriculum as 70% of teens use AI for schoolwork
Source: EdTech Innovation Hub · 27 August 2026
Common Sense Media has finished the K–12 rollout of its Digital Literacy & Well-Being Curriculum, adding a 29-lesson high school programme across six pathways covering AI literacy, media literacy, online harms, health and financial literacy. It is free to teachers, schools and districts.
The release landed alongside new research from the organisation’s Youth AI Safety Institute, based on a nationally representative NORC survey of 1,017 teenagers aged 13–17 fielded between 30 April and 14 May 2026 (margin of error ±4.3 points). Seven in ten teens said they use AI for schoolwork. Only 27% said a teacher had ever discussed what AI is and how it works. And 35% said they do not understand their school’s AI rules or are unsure what they are — while among AI-using teens who hit a school block or ban, 59% simply switched to a personal device.
As their director of school partnerships put it: AI is already in schools; AI literacy is not.
3. Common Sense Media will independently test ChatGPT for Teens
Source: EdTech Innovation Hub · 26 August 2026
The same organisation’s Youth AI Safety Institute is preparing an independent risk assessment of OpenAI’s teen experience, which activates automatically when ChatGPT knows or suspects a user is under 18. Researchers will focus on three areas: whether the Responsible Homework Reminder can be bypassed within the same conversation or by logging out, how reliably parental alerts fire for linked accounts, and what the customisable voice options do to teenagers emotionally.
Context matters here: Common Sense Media previously rated ChatGPT “High Risk” overall and “Unacceptable Risk” for mental health support. Developmental behavioural paediatrician Jenny Radesky singled out the voice and personality settings — Professional, Friendly, Candid, Quirky and Cynical — as her biggest concern, questioning whether an authoritative tone could undercut a teenager’s trust in their own answers, or a playful one deepen emotional attachment. The full assessment is due in the coming weeks.
Worth watching if you are a parent who has just linked an account and assumed the job was done.
4. NASA bans generative AI from its 2027 Student Launch challenge
Source: EdTech Innovation Hub · 26 August 2026
NASA has opened applications for the 2027 Student Launch challenge with a new rule barring students from using generative AI or other AI tools for any challenge-related work. It applies to both the college and university division and the middle and high school division, and covers browser-based generative AI, AI-assisted coding and AI-generated imagery. NASA also says challenge content must not be uploaded to or processed by AI systems at all.
The rule sits inside an existing requirement that students complete 100% of the project themselves, from design and construction through documentation and flight preparation. University teams face an autonomous post-landing imaging payload that must find a passive visual target, photograph it and calculate the distance — without GPS, within 15 minutes of landing, and before a human touches it. Applications close on 14 September, teams are announced 1 October, and Launch Day is 10 April 2027 in Huntsville, Alabama.
Expect more competitions and coursework to copy this wording almost verbatim.
5. 79% of US districts now have AI guidelines — but not the staff to run them
Source: Sogolytics and CoSN, 2026 State of EdTech District Leadership Report · 28 August 2026
The annual CoSN and Sogolytics report, drawing on 607 responses from education technology leaders across 44 states, found districts with AI use guidelines jumped from 57% in 2025 to 79% in 2026. The share with no generative AI guidelines at all fell from 43% to 21%. Training moved too: 70% are now training instructional staff on AI tools, up from 51%, and 64% report using AI in operations, up from 37%.
What did not move is capacity. Cybersecurity staffing, instructional support for teachers, off-campus connectivity and accessibility requirements in purchasing all stayed flat or fell. Two thirds of leaders name insufficient cybersecurity staffing and the lack of a dedicated budget as their top barriers, and 98% are at least moderately concerned about AI enabling new forms of cyberattack — 75% very concerned, the single highest-rated worry in the report. On oversight, districts want direction rather than orders: 67% want non-binding state guidance, only 18% want formal mandates.
Note the shift in what leaders think AI is actually for: student tutoring as a perceived positive use rose from 7% to 46% in a year.
6. Virginia Tech adopts a universitywide human-centred AI vision
Source: Virginia Tech News · 26 August 2026
Virginia Tech has adopted a universitywide vision for how it will teach, research, deploy and lead on AI, produced by an AI Futures Working Group convened by President Tim Sands and co-chaired by engineering dean Warren Dixon and Rishi Jaitly of the Center for Humanities.
The resulting report is organised around seven commitments: mission alignment, innovation for good, human-centred benefit, responsible and ethical use, fairness and transparency, human judgement and accountability, and data security and privacy. The framing leans on the university’s land-grant mission and its motto, Ut Prosim — that I may serve — arguing the value of the technology is set by how well it serves people.
The test, as ever, is whether “human judgement and accountability” survives contact with an assessment deadline.
7. US Education Department: judge edtech by outcomes, not screen time
Source: THE Journal, K–12 Dive and EdSurge, on the Department’s 20 August Dear Colleague letter · reported 25 August 2026
The US Department of Education issued a Dear Colleague letter on the responsible use of classroom technology, with Assistant Secretary Kirsten Baesler urging districts to evaluate technology investments on measurable learning outcomes rather than total minutes spent on screens.
The guidance draws a deliberate line between recreational device use and instructional technology, telling schools to demand evidence of effectiveness and to prioritise tools with clear implementation frameworks. Coverage has read it as a partial defence of classroom technology inside a wider screen-time backlash — reframing the question from whether students should use technology to whether a given tool is demonstrably improving learning.
It also lands the same week as district data showing most schools cannot yet measure that.
Stanford Just Killed the AI Tutor Hype (Here’s What Actually Works)
It is the rare AI story with hard numbers attached, it cuts against a year of marketing, and it gives parents, teachers and students something they can act on tonight rather than a trend to worry about.
Stanford just published the numbers on AI tutors — and in two trials, nearly half the students never opened the thing once. Here is what the evidence actually says works.
Get this digest every morning. A five-minute read on what is actually changing in AI and learning.
Join the newsletter →