Subscribe
Conversations

EP 255 - How Google’s Personal Intelligence is Quietly Revolutionizing (Everyone's) Search Results

In this episode we are joined by Garrett Sussman (iPullRank) to discuss his provocative 12-month study on AI personalization. We dive deep into how Google’s "AI Mode" uses your Gmail, Photos, and Calendar to tailor results—and why "unopened emails" might be influencing what you see next.

Near Media, Mike Blumenthal

EP 255 - How Google’s Personal Intelligence is Quietly Revolutionizing (Everyone's) Search Results

The search landscape is shifting from keyword matching to "Personalized Intelligence," where Google leverages explicit consent to access a user's Gmail, Calendar, and Photos. Research conducted by Garrett Sussman and Profound reveals that Google’s AI doesn't just find information; it infers a user's values and disposable income to steer them toward specific brand outcomes.

The Podcast Deets

The Opt-In Ecosystem (03:11 - 10:59): A look at the "Personalized Intelligence" toggle that gives Google permission to scan your private documents.

The Persona Impact (12:22 - 21:07): How LLMs change their tone and recommendations based on whether a user makes $45k or $150k.

The Email Seeding Experiment (21:35 - 30:02): A breakdown of how sending specific emails can move a brand from the "bottom 10" to a top AI recommendation.

Key Takeaways

• Google’s Personal Intelligence is an opt-in feature that integrates Gmail, Photos, and YouTube history into AI search.
• AI search platforms (like Gemini and ChatGPT) change product recommendations based on explicit and implicit persona data, such as salary.
• Gmail content is a significantly stronger signal for AI personalization than Google Photos.
• Even unopened emails can influence the recommendations provided in Google's AI Mode.
• Google’s AI is moving toward an "agentic" model, acting as a partner that "solves" problems before a user even finishes their query.

  • 👇 Watch by topic:

00:00 – Introductions and the Personalization Shift
03:11 – Defining Google’s "Personalized Intelligence"
12:22 – The 12-Month Profound Study: How Persona Impacts Results
21:35 – Seeding Experiments: How Gmail Influences AI Recommendations
30:02 – The Privacy Paradox: Family Names and "Creepy" Data

Related Links

Implications of AI Mode and Personal Context - 2025 Google I/O Hot Takes
What did we learn about AI Mode and Personal Context at Google I/O 2025 and what does it mean for the future of SEO and Content Strategy?
iPullRankGarrett Sussman
SEOs should start obsessing over context packing and hyperpersonalization. AI assistants are context machines. Google says consent is required, but we’re going from opt-in to opt-out before people… | Garrett Sussman
SEOs should start obsessing over context packing and hyperpersonalization. AI assistants are context machines. Google says consent is required, but we’re going from opt-in to opt-out before people are ready. That’s going to shock the system. The optimist in me says people will love it. For the uninitiated, Google’s Personal Intelligence doc describes context packing as finding the right pieces of your personal data and putting them into the model’s working memory. Doc in the comments. Gmail, Photos, Search history, YouTube, saved places, past conversations, receipts, confirmations, and the random paper trail of your life can all shape the AI output. We already know personalization from local search. I’m talking about this today with Mike Blumenthal and Greg Sterling on the Near Media podcast. Google can find relevant content in a massive corpus of documents. Your digital life history can probably predict intent most of the time, like an optometrist spotting cholesterol levels. Yeah, that’s real. Personalization has to thread the needle between helpful and creepy. Google calls out the limitations: Overpersonalization: One signal can become the whole personality of the answer. Mistaken preferences: Buying something for your kid, spouse, friend, or brother can make the model think it’s your thing. Incomplete information: The model may act like it has the whole story from a few scraps in the kitchen junk drawer. Timeline confusion: AI already struggles with time. Personal history gives it more ways to mess up past, present, and future. Relationship confusion: Names, emails, photos, and messages can still lead to the wrong human read. Missing life changes: Your life can change faster than the model’s understanding of you. ~Ferris Bueller Bad assumptions: A receipt can become proof that you only want Aviator Nation hoodies and Starbucks forever. Ignored corrections: You can correct the model, and it may still wander back like my daughter insisting pink and golden are rainbow colors. Speed vs. depth: More personal answers require more work, so speed becomes a tradeoff. Multimodal risk: Once photos, videos, and visual summaries enter the mix, the risk becomes wrong context with a screenshot attached. And this isn’t just Google. ChatGPT is now surfacing Gmail-connected personalization too. Brand new FAQ in the comments. Brands have to earn their way into the context layer before the search happens. Andrea Volpini wrote about this yesterday. Emails. Newsletters. Reviews. Communities. Videos. Receipts. Support interactions. Third-party mentions. Saved lists. Past purchases. This is the year of hyperpersonalization. Google I/O is in a couple of weeks. I’m betting we’ll see more Personal Intelligence news. The psych of this gets me excited. It’s messy now. It’ll get better. The ugly part is that we’re still at the whims of corporations we can’t trust to have our best interests at heart. Y’all notice that it references receipts? What??
LinkedInGarrett Sussman

Building Personal Intelligence: A step towards truly personal AI

Interested in sponsoring this podcast or our newsletters please reach out to mblumenthal@nearmedia.co

E-mail Mike

Full Transcript --> on-line