Brussels moved to punish Google for favoring its own results, new data showed AI Mode increasingly favoring Google’s own properties, and a child-safety audit gave its AI search a failing grade. Three different kinds of pressure, one target. Here is what moved.
Self-Preferencing Under Fire, From Regulators and the Data Alike
Two developments hit the same nerve from opposite ends: a regulator preparing to force Google to stop favoring itself, and measurement suggesting Google’s AI search is favoring itself more, not less.
- EU Poised to Order Google to Stop Favoring Its Own Search Results The Financial Times reports the European Commission is finalizing a DMA finding that Google illegally ranked its own shopping, travel, and hotel services above rivals, with a possible order to share ranking and click data with competitors. For EU vertical and comparison sites this is a ranking story before a legal one. Map which of your queries show Google’s own panels now, so you have a baseline when a real deadline lands.
- Google Now Cites Itself More Than Almost Any Other Site in AI Mode Profound data shows AI Mode’s citations of google.com rose 8.4x in two months, making it the No. 2 most-cited domain, almost entirely from Business Profiles and Product Knowledge Panels surfaced inline. For local and product queries, the lever to being cited may not be a page at all, but the completeness of the Google-hosted profile or feed. Treat that data as a single vendor’s measurement, then go check your own highest-intent queries.
A Failing Grade, and What It Means for Your Clicks
Beyond the headline, this is a story about what Google’s AI will and will not answer, which is the same thing as which queries still send a click.
- Common Sense Media says Google AI search fails every child-safety test Common Sense Media’s Youth AI Safety Institute ran 2,600+ searches on minor-configured accounts and says AI Overviews and AI Mode failed all five severe-harm tests, including missing 29 percent of explicit suicide references. Google disputes the methodology and says it could not reproduce the findings. Pressure like this is what pushes Google to narrow what its AI answers directly, so watch whether health and sensitive queries start redirecting to sources again.
Today’s Quick Hits
- Google bars prediction market ads in Michigan, New York Google prohibited prediction-market product ads in Michigan and New York from July 13, and separately opened Shopping ads and free listings to more countries including Bulgaria, Croatia, and the Baltics. The free-listing expansion is the half worth acting on.
- Google Tests Hide-Sponsored Toggle Inside Shopping Results Ginny Marvin confirmed a test bringing the hide-sponsored-results control into Google Shopping, matching the one already in main Search. Any control that lets shoppers opt out of paid listings shifts a slice of demand toward organic Shopping.
- ChatGPT Ads Adds Location and Audience Exclusion Controls OpenAI added geo and audience exclusion to ChatGPT Ads Manager, letting advertisers block campaigns by state, DMA, or ZIP and suppress existing customers by list. Another sign it is speed-building toward parity with mature ad platforms.