TLDR: iPullRank testing of nearly 2,000 Google AI Mode responses found Gmail connected brands appeared up to three times more often, especially for shopping searches.
Key Takeaways:
- Google AI Mode can use Personal Intelligence, an opt in feature that draws from connected services like Gmail to tailor recommendations.
- In iPullRank tests, brands linked to Gmail activity showed up far more often in results, including hoodies, coffee machines, and running shoes.
- If personalization drives rankings, AI search could narrow discovery into a confirmation bias loop users can break by turning Personal Intelligence off.
- If you enable Personal Intelligence, the most influential signal in this research was Gmail over Photos, shaping what brands appear first.
AI search wants to be helpful, but if it keeps rewarding what you already buy, it stops acting like a discovery engine. The easy fix is hiding in Settings.
AI search wants to be helpful, but if it keeps rewarding what you already buy, it stops acting like a discovery engine. The easy fix is hiding in Settings.
Q&A
What happens when AI Mode learns what you buy, then learns what you search for next?
Personalization can compound over time, pushing you toward a smaller set of familiar brands unless you periodically reset or disable Personal Intelligence.
Why did shopping categories show the biggest lift in the study results?
Shopping queries already involve strong behavioral signals like intent and repeat purchases, so email linked brand history can steer recommendations more aggressively.
How could this change how brands bid for attention and how marketers measure success?
If AI Mode rankings reward connected user histories, marketers may focus less on generic relevance and more on audience level identity signals.
What controls matter most to users trying to preserve serendipity in AI search?
Users can toggle off Personal Intelligence in Gemini settings and disconnect services such as Gmail to reduce the behavioral inputs shaping recommendations.
If Google does not confirm the exact ranking mechanics, what should researchers and regulators still demand?
They can push for clearer disclosure of which connected data sources influence rankings, plus audit tools that test personalization effects across cohorts.
Top in AI
Nvidia Rubin CPX stalls as Groq integration tightens supply math
YouTube boosts AI labels as detection automation ramps
ElevenLabs brings Stan Lee voice, likeness back via licensing pact
PostHog trains AI models using customer data
No comments yet. Be the first to share your thoughts!