No detectable change in mental-health information-seeking during Italy's 2023 ChatGPT ban: a natural experiment with a verified first stage

Marta Antonetti
Additional contact information
Marta Antonetti: London School of Economics and Political Science

No 4c5wf_v1, SocArXiv from Center for Open Science

Abstract: Regulatory bans on AI platforms are accelerating globally, yet their behavioural consequences for affected populations remain unmeasured. Italy's temporary ChatGPT ban (31 March – 28 April 2023) provides a rare natural experiment: an exogenous, country-specific shock to AI access. I verify the first stage using Google Trends data on VPN and ChatGPT searches — Italian VPN searches peaked at nearly five times their pre-ban level on the first day of the ban, averaged twice their baseline throughout, and returned to pre-ban levels within a week of reinstatement; no equivalent change occurred in France or Spain. Against this confirmed exposure, I find no detectable change in loneliness-related information-seeking. The 2023 ban-window deviation (+3.5 index points above the preceding eight weeks) was smaller than in 2019 and 2022 (+19.3 and +19.4 points respectively) but larger than 2024 (−12.0 points); year-permutation tests return p ≥ 0.25 under both calendar and Easter alignment. An apparent "rise then fall" pattern in an earlier, single-draw version of this analysis reflected seasonal variation and single-draw sampling noise: three independent draws correlate at r ≈ 0.74 and show no consistent post-reinstatement decline. No construct shows a ban effect once seasonality is accounted for: the only secondary signal, a rise in sadness-related search that survives multiplicity correction and exceeds all placebo windows, is matched or exceeded by the same spring weeks in other years. I identify three methodological lessons for Google Trends research: the consequences of retaining sparse terms, relying on a single data pull, and using grammatically gendered phrases whose baseline asymmetry may reflect lexical rather than demographic differences.

Date: 2026-10-20
References: Add references at CitEc
Citations:

Downloads: (external link)
https://osf.io/download/6ad70e787a8004a832b5a730/

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:osf:socarx:4c5wf_v1

DOI: 10.31235/osf.io/4c5wf_v1

Access Statistics for this paper

More papers in SocArXiv from Center for Open Science
Bibliographic data for series maintained by OSF ().