In June 2025, Varsity Tutors' parent company Nerdy signed a four-year White House "Investing in AI Education" pledge, committing to broaden AI literacy and accelerate student academic growth while preserving the essential human connection in teaching. Three months later, RAND released a survey showing 54% of K-12 students had used AI for schoolwork, up more than 15 percentage points in two years. Nerdy's Live+AI platform, already deployed across 1,000+ district partnerships, is now being positioned as the definitional example of AI-powered "high-dosage tutoring." Charter directors and special programs coordinators are being pitched a version of intervention that runs at platform scale, claims 2X gains in targeted skill areas, and treats the tutor-student relationship as a feature layered on top of software. That reframing is happening in real time, and it changes what school leaders need to ask before signing a contract.
Nerdy's pledge is not a small announcement. The company is publicly committing four years of investment to AI-integrated instruction, and its Live+AI product is being marketed to districts as a scalable version of what education researchers have called high-dosage tutoring, an evidence-based intervention model that traditionally means small groups, consistent tutors, and 30+ sessions per year built into the school day.
The academic consensus on what high-dosage tutoring requires is well-documented. The National Student Support Accelerator at Stanford defines it as tutoring delivered "at least three times a week, or at a rate of at least 50 hours over 36 weeks, by a consistent tutor with formal training in a one-on-one setting or a very small group." A 2020 meta-analysis of randomized tutoring trials by Nickow, Oreopoulos, and Quan, published through the National Bureau of Economic Research, found average effect sizes of 0.37 standard deviations, one of the largest evidence bases in K-12 intervention research. Consistency of tutor was one of the strongest predictors of effect size.
When a platform vendor applies the "high-dosage" label to a product where the tutor rotates, the schedule flexes to student availability, and an AI layer handles a growing share of the instructional decisions, the label is doing rhetorical work that the underlying research does not support.
If you sit on the intervention side of a charter LEA or district, this shift lands on your desk in three concrete ways.
First, your authorizer and your federal programs officer are reading the same press releases. When a national vendor claims 2X gains across a thousand districts, that number enters the conversation whether or not it is peer-reviewed. You will be asked to justify why your intervention line item does not use the platform everyone else appears to be using.
Second, the ESSA evidence standards still require Tier 1 or Tier 2 evidence for interventions funded through Title I, Title III, and school improvement dollars. Vendor-reported gains from proprietary platforms are not the same as independent research. You are the one signing the assurance.
Third, if your students are Long-Term English Learners, Tier 3 math cohorts, or students with IEPs, the intervention has to hold up under a compliance audit. In our experience working with California charter schools, the audit question is rarely "did the students improve on the platform's internal metric." It is "who was the instructor, what were their credentials, how often did they meet, and did the same instructor meet with the same students across the cycle."
The research literature converges on one point. In the Nickow, Oreopoulos, and Quan meta-analysis, tutoring programs delivered by teachers and paraprofessionals produced larger effects than programs delivered by non-professional volunteers. Programs during the school day outperformed after-school programs. Small group instruction (1:1 to 1:4) outperformed larger ratios.
Susanna Loeb, who founded the National Student Support Accelerator (now at Stanford) and previously directed the Annenberg Institute at Brown, consistently frames high-impact tutoring as relationship-based, treating the tutor-student relationship as central to the intervention. In that framing, consistency of tutor is not a luxury feature but a mechanism through which the effect operates. A tutor who knows where a specific student gets stuck, and how fast that student is willing to move, can adjust in ways that a rotating tutor or an algorithm cannot.
The Annenberg Institute's design principles for high-impact tutoring lay out several core features. Every one of them assumes a human instructor with sustained access to the same students. AI can augment scheduling, progress monitoring, and formative check-ins. It cannot, in the current evidence base, substitute for the relational continuity that produces the effect.
Evidence-based intervention that clears the ESSA Tier 1 or Tier 2 bar tends to share a specific design:
This is not the same product as an AI platform serving a thousand districts. It is a slower, smaller, more expensive-per-student model. Its advantage is that it matches the design principles the research actually supports, which is what matters when the intervention has to hold up under audit and when the students have to hold onto the gains after the cycle ends.
A+ Tutoring, a California K-12 virtual intervention provider that partners directly with charter LEAs, runs this smaller-cohort model. Our tutors are credentialed California teachers, not gig-economy contractors. The same instructor meets with the same students across the full intervention cycle. Sessions are scheduled into the school day when the charter's model allows it and are aligned to each school's MAP or state assessment data.
In our 2024-25 cycle with iLEAD Exploration, a California charter, 75% of Math Tier 3 students (9 of 12) reached MAP growth benchmarks, 87.5% of ELA Tier 3 students (7 of 8) reached theirs, and the combined Tier 3 cohort landed at 80% (16 of 20 students) growing at 3-6x national MAP Growth benchmarks. These numbers are small because the cohorts are small. That is the point of the model.
If your Tier 3 cohort were named individually, would the same students appear next year, or would you recognize different names in the same seats?
We believe, and this is A+'s view rather than a citable finding, that the reason those numbers hold is that the tutors are the same people every session, they know the students by name, and the intervention plan is built with the charter's intervention team rather than delivered from a catalog. That is not a competitive claim about any specific vendor. It is a design claim about what tends to work.
Regardless of which provider you use, five steps will sharpen the conversation this fall:
A+ Tutoring is a California K-12 virtual intervention provider partnering directly with charter LEAs and district schools. Our instructors are credentialed California teachers. Our model is small-cohort, consistent-instructor, school-day-aligned intervention, co-designed with each partner school's intervention team. A+ partner schools have shown 75% of Math Tier 3 students, 87.5% of ELA Tier 3 students, and 80% of the combined Tier 3 cohort reaching growth benchmarks at 3-6x national MAP Growth rates in the 2024-25 cycle.
Walk Your Tier 3 Data With Our Team