# Study finds AI tutoring matches human effectiveness at 1/918th cost for GRE prep

Digest AI · Marketing & Small Business · published 2026-09-29T03:20:00Z

Canonical: https://digestai.news/story/study-finds-ai-tutoring-matches-human-effectiveness-at-1-918th-cost-fo

## Summary

A new study called *StudentBench*, published September 23, 2026, compared AI and human tutoring using GRE questions with 2,383 participants. It found AI instruction statistically equivalent to human-led learning, with one AI costing **$0.0052 per 1% gain in correct answers** versus **$4.81 for human tutors**—a **918x cost difference** under controlled conditions. The authors emphasize this is a *research metric*, not a real-world pricing comparison, and note limitations: the study only tested GRE material, not elementary or middle-school exams, and measured short-term gains, not long-term retention.

The findings contrast with earlier experiments where a fifth-grader found AI-generated mock exam questions too dissimilar to real test formats. The author argues AI excels as a *tutor*—explaining unclear concepts or generating alternative explanations—rather than as a *predictor* of exam content. However, they caution against blind reliance, citing personal failures where AI misinterpreted geometry problems when given images. The study’s authors stress it’s a preprint and not yet peer-reviewed.

## Key points

- Study with 2,383 participants found AI tutoring matched human effectiveness for GRE prep, per *StudentBench* (September 2026)
- Cost per 1% learning gain: **$0.0052** for AI vs. **$4.81** for human tutors—918x cheaper in research conditions
- Researchers warn figure doesn’t reflect real-world tutoring costs and study only tested short-term gains on GRE material

## Why it matters

The study suggests AI could democratize high-quality tutoring by reducing costs, but its narrow focus on GRE prep and short-term metrics limits immediate applicability. For educators, it highlights AI’s strengths as a teaching tool—explaining concepts, not predicting exams—while underscoring the need for human oversight in problem interpretation.

## Sources

1. [Is ChatGPT Better Suited as a 'Tutor' Than a 'Predictor'? Reflecting on Mock Exam Results and Recent Research](https://note.com/ainomakasedokoro/n/n105431f71336?hl=en) (note.com, 2026-09-29)

## Cite

Digest AI, "Study finds AI tutoring matches human effectiveness at 1/918th cost for GRE prep", 29 September 2026, https://digestai.news/story/study-finds-ai-tutoring-matches-human-effectiveness-at-1-918th-cost-fo

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/study-finds-ai-tutoring-matches-human-effectiveness-at-1-918th-cost-fo.json