Italy asks EU to review Google AI Search under the DSA

Italy’s communications regulator AGCOM has asked the European Commission to assess Google’s AI Overviews and AI Mode.

Fact: AGCOM published on 30 April 2026 that, in its role as Italy’s Digital Services Coordinator, it had sent a request to the Commission regarding Google Ireland Ltd. The referral concerns the public interfaces AIO, or AI Overviews, and AI Mode, assessed against Articles 27, 34 and 35 of the Digital Services Act.

The case started with a complaint from the Italian newspaper publishers’ federation FIEG. It argues that AI-generated answers in Google Search reduce the visibility and discoverability of editorial sources. AGCOM also points to risks for the economic sustainability of publishers, especially smaller and independent ones, plus risks to source pluralism and AI answers that may contain errors or fabricated information without easy source verification for users.

AGCOM says it first collected information from Google and held hearings with Google, FIEG and FISC. It then decided to refer the matter to the Commission under Article 65 of the DSA. The question is whether Google may have breached obligations for very large platforms and search engines on systemic risk mitigation, media freedom, pluralism and transparency in recommender systems. Reuters reported the same day that Google had no immediate comment.

This is not a finding that Google broke the law. It is a formal regulatory escalation. The leadership consequence is still clear: AI search shifts power from link lists to answer surfaces. For businesses, visibility, attribution and customer traffic can no longer be managed as traditional SEO alone.

CIOs, CMOs and legal teams should treat AI search as platform risk. Map how much traffic, lead generation and customer-service demand comes through Google Search. Measure the gap between impressions, clicks and citation in AI answers. Fix the technical basics: structured data, clear source and author signals, updated robots policy, and content that can be cited accurately.

Assessment: this points to a new European norm where AI interfaces may need to be explained as distribution and risk systems, not just product features. Leaders should not wait for a final EU decision before adjusting content, data and channel strategy. If Google, Microsoft or other assistants become the main entry point for customers and employees, governance must cover who receives the answer, who owns the source, and how the organisation detects mistakes.