Home/Adoption Trends/Back-to-School Effect on AI Adoption
Adoption Trends

Back-to-School Effect on AI Adoption

Unlike most claimed “seasonal” patterns in AI — which mostly don’t hold up — the back-to-school effect is real, repeatable, and well-documented. ChatGPT usage tracks the academic calendar closely: it peaks during finals, more than halves when schools let out for summer, stays low through July, and rebounds each September as students return. In 2025, daily token volume fell from a May 27 finals-season peak of 97.4 billion to roughly half that in June. The academic fingerprint is unmistakable — a Rutgers study of 10,000 prompts concluded most usage was academic, and a weekly sawtooth (dips every weekend during term) shows the same pattern at higher frequency. First noticed in 2023, the effect has recurred every year since. That said, it’s frequently overstated: viral headlines claimed a “70% collapse,” but the real summer dip is more like 20–50%, and it’s driven by students specifically — professional and international usage is far less seasonal. This page maps the genuine seasonal signal, the evidence it’s student-driven, the September rebound, and why the scariest headlines overstated it.

4 visualizations 4 sources Last updated June 2026 Free to embed
Loading chart...
Loading chart...
CHART 1 · A REAL SEASONAL SIGNAL

ChatGPT token usage, summer 2025

ChatGPT usage halves in summer, rebounds in fall

Loading chart...
Source: OpenRouter data (via Futurism) 2025: this is one of the few genuine seasonal effects in AI. ChatGPT usage peaked at 97.4B tokens/day on May 27 (finals season), then more than halved — from ~79.6B tokens/day in May to ~36.7B in June — when schools let out. Usage stayed low through July, then rebounded in September.
CHART 2 · THE ACADEMIC FINGERPRINT

Evidence usage is student-driven

The academic fingerprint: usage tracks the calendar

Loading chart...
Source: Rutgers study / OpenRouter 2025: a Rutgers analysis of 10,000 ChatGPT prompts found student interactions strongly tracked the school calendar — spring break and summer were “dead zones” — concluding “most usage was academic.” A weekly sawtooth (weekend dips) during term is the same fingerprint at higher frequency.
CHART 3 · THE REBOUND

September brings usage back

The September rebound (repeats yearly)

Loading chart...
Source: Bloomberg / Sherwood 2023–25: the effect is repeatable — first noted in 2023, confirmed each year since. As the new school year begins, ChatGPT usage climbs back from its summer trough. The back-to-school rebound is now a predictable annual pattern, not a one-off.
CHART 4 · DON'T OVERSTATE IT

The “70% collapse” was a myth

Real effect, exaggerated headline: ~20–50%, not 70%

Loading chart...
Source: SimilarWeb / fact-check 2025: the honest caveat — viral headlines claimed a “70% traffic collapse,” but the real summer dip was more like 20–50% depending on the metric and source. And it’s student-driven, not all users — professional and international use is far less seasonal. A real effect, often exaggerated.

About this data

This page documents the back-to-school effect on AI usage, drawing on OpenRouter’s usage data (a ~2.5M-user panel, popularized by Futurism), a Rutgers study of 10,000 ChatGPT prompts, and Bloomberg/Sherwood reporting on the annual rebound.

The AI Behavior Index is the research arm of OneChat AI, an integrated multi-model AI platform. We compile and analyze data from primary research sources to make AI adoption and market trends more accessible to journalists, researchers, and decision-makers.

Why we present this as a real effect (unlike general “seasonality”): most claimed AI seasonality doesn’t survive scrutiny, but the school-calendar effect does — it’s visible in token data, corroborated by an academic prompt study, shows the expected weekly sawtooth, and repeats annually. We present it as genuine while flagging the two honest caveats: the magnitude was widely exaggerated (the “70%” figure), and it’s specific to student usage, not the whole user base.

Methodology notes: the token-volume data is from OpenRouter’s panel (~2.5M users), a large but partial sample of total ChatGPT usage, so absolute magnitudes are directional. Different trackers reported summer dips ranging ~20–50%; we cite the range and flag the exaggerated “70%” claim. The Rutgers study (10,000 prompts) supports the academic-usage interpretation but is a sample. The effect is strongest for consumer/student use and weaker for professional and non-U.S. usage, so it doesn’t generalize to “all AI adoption.”

Sources used on this page:

Corrections or suggestions: research@aibehaviorindex.org

For Journalists & Researchers

Use this data in your work.

Every statistic, chart, and graphic in this index is free to use and cite, with full source attribution. Can’t easily find what you need? Use our search bar to search by keyword, topic, or category.

✉️
Talk to our research team
Need a specific cut of data, an interview, or a quote? Email us — we typically respond within one business day.
research@aibehaviorindex.org →

How the data works

Every statistic shown is sourced from a publicly available study, survey, or report. We aggregate, organize, and contextualize this data — but the underlying research is conducted by the cited sources. Click any source link to access the original methodology. If you run into any issues or have a study to suggest, contact us at research@aibehaviorindex.org.