How to Get AI to Recommend Your Product

This is a summary of a recent systematic talk by the CEO of Graphite on GEO (Generative Engine Optimization).

  • Core principle of GEO: LLM + RAG — the LLM decides whether something is possible, while RAG determines what to retrieve.
  • The essence of GEO: Making your product or brand an intrinsic part of the “answer” — not just a result, but a trusted component of the response.
  • SEO carryover: Many proven SEO tactics still work for GEO — especially foundational ones like clean site architecture, semantic clarity, and authoritative linking.
  • GEO ranking logic differs sharply from SEO: Instead of backlinks or keyword density, GEO prioritizes citation frequency — how often and where your brand appears across trusted sources. The more widely referenced you are, the higher your chance of being surfaced by AI.
  • User queries are longer and more specific in AI search versus traditional search — generating countless previously unsearched long-tail questions. A core GEO tactic is therefore systematic long-tail question mapping: anticipating and answering highly contextual, scenario-based user intents.
  • Each AI platform has its own source preferences. Managing which sources it trusts — and actively cultivating those relationships — is now mission-critical.
  • A compelling case study: Webflow’s GEO-driven traffic is six times that from Google; ~8% of its new signups now come directly from ChatGPT and similar AI platforms. Their three key tactics:
    • High-intent landing pages covering dozens of sub-questions (not just one topic),
    • YouTube videos (easily cited, rich in context and demonstration),
    • Authentic Reddit engagement (community-vetted, credibility-anchored).
  • This aligns with broader patterns: GEO-sourced traffic consistently shows higher conversion rates — likely because users have already had multiple conversational exchanges with AI before arriving at your site, indicating strong intent and elevated trust in AI recommendations.
  • Speed-to-impact is dramatically faster: While SEO may take months to move the needle, GEO benefits from recency weighting — visible results often appear within hours or days.
  • Your official website is underutilized — especially in China. For AI, the most authoritative source about your brand is your own site. Leverage that authority: publish deeply detailed, question-driven content — e.g., “How do I migrate from X to Y?”, “What happens if I exceed my quota?”, “Can this integrate with Z?” — sourced from sales calls, support logs, or AI-suggested follow-up questions.
  • GEO-friendly content should deliver high information gain: Prioritize original research, expert commentary, and coverage of underserved subtopics — not rehashing what’s already abundant.
  • AI crawlers are everywhere — and many sites now block them outright due to bandwidth strain. But the speaker urges intentional openness: allow indexing and citation, and treat AI not as a threat but as a partner — optimizing with it, not against it.
  • AI-generated content alone? Not viable. As the speaker puts it: If everything is AI-written, why would anyone need a search engine at all? The sustainable model is AI-assisted + human-reviewed + grounded in original insight.

Video link:
youtube.com

The Meaning of Running

This morning I completed a 25 km slow-distance run.

Running isn’t just physical training — it’s daily self-discipline, a moving meditation.

Physiologically, consistent running improves metabolism, cardiovascular resilience, focus, sleep quality, and overall vitality — holistically.

These bodily shifts ripple into mental states: greater emotional stability, calmness, and grit.

Lately, I’ve noticed another subtle effect: long-term running strengthens desire regulation. Entrepreneurs, for instance, often chase outcomes with intense urgency — a mindset that breeds anxiety and distorted behavior.

True freedom isn’t indulgence — it’s the capacity to hold desire lightly, to pause, to choose restraint. Running cultivates that. You learn your rhythm: not too fast, not too slow — steady, intentional, sustainable.

Upgrading and Downgrading

Learn to upgrade. Compete by downgrading.

Here, “upgrading” means elevating your thinking — beyond immediate tasks, local views, fragmented details, or single-skill execution.
“Downgrading” means translating high-level insight into simple, decisive action — so you outmaneuver others within their own low-dimensional arena.

  1. Learning = Upgrading
    • In practice: Shift from “What am I doing?” → “Why am I doing this?” → “How might we do it better?” → “Should we do this at all?” → “What system or platform could make this obsolete — or indispensable?”
    • In study: Seek higher vantage points — consult experts, read timeless books, invest in rigorous learning. These aren’t extras. They’re upgrades to your operating system.
  2. Competition = Downgrading
    If your product team stalls competing on UI polish or feature checklists, ask: Where can we break through at a higher level? — business model, user psychology, ecosystem design, data infrastructure. Once you lead there, even small tactical moves — a pricing tweak, a workflow change — can overwhelm rivals still grinding in the low-dimension race for “more features, prettier buttons.”

For companies:
If you’re in B2B, study B2C. B2B leans on logic, expertise, and complex sales cycles. B2C thrives on emotion, speed, and frictionless experience — and its marketing tactics, product rhythms, and feedback loops are sharper, faster, and more varied. Importing that energy into B2B works.

AI’s Real Impact on Work

A recent study (2015–2025) analyzed 62 million U.S. workers, 285,000 companies, and 245 million job postings — tracking actual hiring behavior, not speculation. Key findings:

  • Since Q1 2023, AI-adopting firms cut junior-role hiring by ~7.7% over six quarters — driven almost entirely by reduced hiring, not increased attrition.
  • Promotions for existing staff rose — suggesting AI isn’t replacing people, but reassigning them upward: automating routine tasks, freeing humans for higher-complexity work.
  • Senior roles grew steadily. Why? Two forces: (1) AI handles more entry-level work, and (2) organizations increasingly demand strategic, judgment-heavy work — precisely what AI can’t yet do.
  • The hardest-hit sector? Traditional wholesale & retail — junior hiring dropped ~40% quarterly. Roles like frontline customer service and clerical support are among the easiest to automate.
  • For new grads: Top-tier university graduates saw little impact; lowest-tier grads retained cost-driven demand; but mid-tier graduates — from strong-but-not-elite schools — faced the steepest pressure. Their expectations outpace their current leverage — and AI narrowed that gap fast.
  • Overall corporate strategy? Not mass layoffs — but hiring restraint at the junior level, paired with stronger internal emphasis on AI fluency.

Personal takeaways:

  1. Embrace AI — and deliberately offload your repetitive work to it. That’s step zero of mastery.
  2. Within your organization, seek out complexity. High-complexity work = high value + high defensibility.
  3. Aim for expertise. “Expert + AI” creates powerful leverage. “Generalist + AI” rarely does.

Principles for Lending Money

A friend recently struggled collecting a debt. He’d asked repeatedly — no reply. No resolution. Just quiet erosion of trust.

I shared a borrowing SOP a mutual friend uses — pragmatic, respectful, and protective:

  1. First, assess whether the person is worth lending to — not just financially, but ethically and relationally.
  2. If promising, call them — confirm the purpose and risk profile of the loan. Reasonable use + manageable risk = green light to proceed.
  3. Then, impose a cooling-off period. Also disclose upfront: interest will apply, and a formal promissory note is required. Let them sit with that.
  4. If they agree, draft the note together — clear repayment date, grace period (e.g., 3 months interest-free), then automatic interest + late fees.

Why it works:

  • The cooling-off period filters urgency from necessity.
  • It reduces emotional decision-making — for both sides.
  • It builds mutual accountability before money changes hands.

I told my friend: People who truly see you as a friend won’t borrow unless absolutely necessary. Those who borrow casually? Often don’t see you as a friend at all.

Another friend, facing multiple unpaid loans, sued each debtor individually. It worked — cleanly and effectively.

Li Xiang once put it well: When lending, don’t be a saint. Don’t be a villain. Be clear.

Why Check?

When you lack experience, you make “basic” mistakes — repeatedly. What helps? Checking. Relentlessly.

Early in my career, I had no track record — so I compensated with layered review: draft → self-review → peer review → final sweep. Each pass caught something new. Tedious? Yes. But over time, it trained my eye for detail, my sense of flow, and my instinct for quality.

Why do inexperienced people make so many basic errors? Surface reasons include:

  • Poor grasp of end-to-end process,
  • Low sensitivity to nuance,
  • Insufficient exposure to failure,
  • Overconfidence or excessive caution,
  • Missing self-audit habits.

But deeper down, two things matter more:

  1. Responsibility: My friend observed — if it were your own project, you’d rarely skip that critical step or misread that spec. Most “basic” errors stem from diluted ownership.
  2. Energy: Focus is finite. Doing small things well demands deep, uninterrupted attention — often 60–120 minutes of full presence. Exhaustion kills precision. So protect your energy — it’s the silent foundation of reliability.