SUMMARY

a16z partners Alex Rampell and Justine Moore explore AI agents' transformation of e-commerce, challenging Google's search monopoly, affiliate marketing flaws, and future shopping behaviors from impulse buys to considered purchases.

STATEMENTS

• The worldwide web is currently unhealthy due to SEO-optimized low-quality content polluting search results.
• AI agents can direct consumers to products, potentially shifting economic value away from traditional search engines like Google.
• Google acts as a tax on GDP by capturing a percentage of consumer spending through per-click charges.
• Affiliate marketing, originating from pornography tracking, relies on cookies and pixels for commissions.
• Impulse buys target emotions and are unlikely to involve AI assistance, as they occur spontaneously.
• High-value purchases benefit from extensive AI research, but lack straightforward affiliate monetization.
• Personal search volume for non-commerce queries is declining in favor of AI tools like ChatGPT.
• Online shopping represents a massive consumer market, yet few AI startups have disrupted it despite LLM potential.
• Observing consumer behavior, like using CamelCamelCamel for price alerts, reveals demand for automated purchasing.
• Viral trends among teenagers using ChatGPT to identify and buy celebrity-inspired items predict broader AI adoption in shopping.
• Dynamic pricing based on consumer surplus could face regulatory and popularity challenges.
• E-commerce constitutes only 16% of total retail sales due to demand for immediacy and experiential shopping.
• Research often occurs online, but purchases happen in-person for big-ticket items like laptops or clothes.
• Last-click attribution in marketing is corrosive, rewarding the final touchpoint inaccurately.
• Tools like Honey steal attribution by inserting affiliate links at the point of purchase.
• Aggregators like Amazon and Shopify thrive, while single-brand resellers fall into commodity traps without recurring revenue.
• Trends in consumer products like shoes are fleeting, favoring aggregators over niche brands.
• AI struggles to generate demand through trends, excelling instead in fulfilling known needs.
• Google's freemium model succeeds by improving search relevance through paid ads alongside organic results.
• Search volume is declining, but Google's revenue grows by retaining premium, monetizable queries.
• AI hallucinations in product recommendations have driven users back to traditional search.
• The internet's commercialization has polluted content with affiliate-driven junk, eroding trust.
• Video reviews on YouTube offer honest feedback due to creator incentives, unlike text-based SEO spam.
• Amazon's marketplace is flooded with low-quality imports from AliExpress, marked up for profit.
• Costco's membership model ensures trust by refusing high margins and poor products, prioritizing value.
• Impulse and highly considered purchases resist full AI disruption, while mid-tier items like handbags or bikes are ripe for agents.
• Products with UPCs enable automated price optimization, while unique items require more research.
• Opportunities exist for specialized AI agents in price hunting and personalized recommendations.
• Merchant infrastructure must adapt to AI agents browsing and purchasing autonomously.

IDEAS

• AI agents could automate the "consumer as agent" model seen in price alert tools, completing purchase loops efficiently.
• Teenage girls using ChatGPT to replicate celebrity outfits highlights AI's potential for visual product discovery and affordable alternatives.
• Dynamic pricing via device type (iPhone vs. Android) infers elasticity but risks backlash and regulation.
• Demand curves differ for immediate needs like toothpaste versus delayed e-commerce deliveries, explaining slow e-commerce growth.
• Online research followed by in-person buys blurs attribution, complicating marketing ROI.
• Honey's coupon interception exemplifies attribution theft, benefiting affiliates over original influencers.
• Commodity resellers like Casper fail without proprietary products, as factories supply identical goods to competitors.
• Trend-driven products like running shoes commoditize quickly, empowering aggregators to capture shifting demands.
• AI cannot easily replicate social proof from viral videos, limiting its role to utility-based fulfillment.
• Google's AdWords innovation blended paid and organic results, enhancing overall search quality.
• Informational queries shift to AI like ChatGPT, but transactional ones remain with Google for monetization.
• Hallucinations in AI recommendations erode early trust, stalling commerce integration.
• Walled gardens fracture search, with platforms like X for real-time and Facebook for social queries.
• Pre-internet publications like Consumer Reports thrived on ad-free subscriptions for unbiased reviews.
• Craigslist's classifieds disrupted newspapers, killing local ad revenue and objective product journalism.
• Summarizing SEO-optimized affiliate junk yields unreliable AI outputs, necessitating "decrapification."
• YouTube videos provide authentic reviews via creator accountability, untapped by traditional search.
• Amazon sellers arbitrage AliExpress imports, flooding markets with fake reviews via SKU swaps.
• Costco's low-margin insistence preserves membership value, building unbreakable trust.
• Bezos contrasted high-margin premium models (Apple) with low-margin scale (Amazon).
• Mid-consideration purchases like laundry detergent suit AI for price scanning and auto-buying.
• Conversational AI could simulate expert consultations for durable goods like bikes.
• UPC-coded products streamline AI automation, while non-standard items demand hybrid research.
• Niche AI agents for cashback optimization could mainstream among non-technical users.
• AI as the "last click" in attribution shifts power to intelligent intermediaries.

INSIGHTS

• AI's strength lies in automating routine price hunting for known products, freeing consumers from manual optimization.
• Trust in commerce stems from curation and transparency, as exemplified by Costco's membership-driven quality control.
• Attribution's complexity will intensify with AI, rewarding holistic multi-touch models over simplistic last-click theft.
• E-commerce's stagnation reflects human preferences for immediacy and tactile experiences, beyond algorithmic efficiency.
• Aggregators dominate by riding trends without inventory risk, while brands falter in commoditized markets.
• Informational search migrates to AI, preserving transactional revenue for incumbents like Google.
• Content pollution from affiliates undermines AI's analytical potential, demanding new verification ecosystems.
• Video as a medium fosters authenticity through visible creator incentives, a model for future review aggregation.
• Hallucinations highlight AI's current limits in dynamic product data, necessitating real-time integrations.
• Specialized AI agents could democratize expert advice, enhancing decisions for complex mid-tier purchases.
• Merchant adaptations to agentic browsing will redefine site design around machine-readable intent.
• UPC standardization enables seamless automation, bridging research and transaction in AI workflows.

QUOTES

• "The worldwide web is unhealthy right now. Most of the things on the internet are crap and they're crap and we know that they're crap, but they SEO optimize crap."
• "Google, they kind of are a tax on GDP. Consumer spending is a huge part of GDP. They get a percentage of all that spend because they're charging per click."
• "Impulse buys are huge and with impulse buys like almost tautologically you're not going to use AI to tell you to buy something like you shouldn't buy anything that's an impulse buy."
• "Last-click attribution... is the most kind of pervasively corrosive business model I think on the internet."
• "You can't turn shill junk into honest analysis."
• "Costco is the greatest company in the world because Costco refuses to sell bad things. They refuse to take a high gross margin."
• "AI agents can direct people to things if people start their purchase activities there."
• "The internet's commercialization has polluted content with affiliate-driven junk, eroding trust."
• "Summarizing that crap is not helpful. So, how do you decrapify that?"
• "There's demand curve for like real-time toothpaste. That's part."

HABITS

• Observe consumer behavior objectively before predicting trends, using tools like price alert sites.
• Rely on personal experience to gauge search volume shifts, favoring AI for non-commerce queries.
• Use ChatGPT extensively for research, reducing Google usage by orders of magnitude.
• Scan daily for price drops on routine items like laundry detergent via alerts or agents.
• Research online extensively for big purchases, then verify in-person for tactile assessment.
• Prioritize trusted brands like Apple for premium items to avoid deep forum dives.
• Watch unsponsored YouTube videos for honest product reviews on specific needs.
• Value money over time by manually hunting coupons and cashback, soon to be automated.
• Shop at Costco for everything from glasses to travel deals due to inherent trust in value.

FACTS

• E-commerce accounts for 16% of total U.S. retail sales, lower than expected due to immediacy preferences.
• Affiliate marketing predates AdWords, originating from tracking in online pornography.
• CamelCamelCamel serves as Amazon's largest affiliate through price tracking alerts.
• Google's AdWords model, inspired by Overture, propelled it to a $2 trillion valuation.
• ChatGPT boasts 800 million weekly active users, primarily for informational queries.
• Craigslist disrupted newspaper classifieds, contributing to the decline of traditional media.
• Costco generates net income primarily from 50 million+ memberships at around $100 each annually.
• Amazon derives significant revenue from 100% gross margin advertising clicks.
• Wayfair succeeded in non-UPC furniture by matching dimensions without scannable codes.
• Delta Airlines has experimented with dynamic pricing based on inferred consumer wealth.

REFERENCES

• Trial Pay: Alex's early company, a major affiliate in online payments.
• CamelCamelCamel: Price tracking site for Amazon, likened to Google News alerts for deals.
• ChatGPT: AI tool used for product research and recommendations, with hallucination issues.
• Honey: Browser extension that intercepts purchases for affiliate commissions.
• RetailMeNot: Original coupon site that went public via attribution theft.
• Consumer Reports: Ad-free subscription publication for unbiased product reviews.
• Craigslist: Classifieds platform that killed newspaper local ad revenue.
• Wirecutter: New York Times-acquired site for product recommendations, often affiliate-linked.
• AliExpress: Source for cheap imports resold on Amazon with markups.
• DropCam: E-commerce camera product acquired by Nest, attached to subscriptions.
• Ebates: Cashback site bought by Rakuten for deal optimization.
• Quidco: UK-based cashback platform similar to Ebates.
• Overture: Bill Gross's search ad model that influenced Google's AdWords.
• H-index: Academic metric analogy for Google's PageRank in linking quality.
• TikTok Shop: Platform for impulse buys via video-integrated shopping.
• Reddit: Forum for deep research on buy-it-for-life products.
• YouTube: Source for unsponsored video reviews of items like shoes or blow dryers.
• Bama Rush sorority TikToks: Viral content influencing shoe trends.
• Wise: Money transfer company showing real-time vs. delayed demand differences.
• Costco Kirkland products: Generic lines like wine, beer, shirts rivaling premium brands.
• Lululemon lawsuit: Against Costco for superior, cheaper yoga pants.

HOW TO APPLY

• Observe current consumer tools like price alerts to identify automation opportunities for AI agents.
• Segment purchases by type: avoid AI for pure impulses, leverage for mid-tier research like handbags.
• Integrate visual uploads for AI to match celebrity styles, suggesting affordable duplicates.
• Implement dynamic pricing cautiously, using device signals but monitoring regulatory risks.
• Conduct online research for specs, then transition to in-person trials for high-value items.
• Dismantle last-click attribution by adopting multi-touch models for fairer marketing rewards.
• Curate content sources, prioritizing ad-free or video reviews to feed clean data to AI.
• Standardize products with UPCs for seamless price scanning and auto-purchase workflows.
• Fine-tune AI models on expert conversations for personalized consultations on durables like bikes.
• Adapt merchant sites for agentic browsing with machine-readable structured data.
• Build trust via low-margin memberships, refusing subpar products to sustain loyalty.
• Automate routine buys like detergent by scanning prices daily and stocking extras when deals hit.
• Specialize AI agents in niches like cashback optimization, simplifying for non-technical users.
• Verify AI recommendations against hallucinations by cross-referencing real-time product data.
• Expand infrastructure for AI payments, securing credit card integrations for autonomous transactions.

ONE-SENTENCE TAKEAWAY

AI agents will automate efficient shopping for mid-tier needs, disrupting search while preserving trust-based models like Costco.

RECOMMENDATIONS

• Prioritize AI development for price-optimized auto-purchases on UPC-coded staples to capture routine spending.
• Invest in specialized agents for visual discovery, targeting young demographics with trend-matching capabilities.
• Reform attribution systems to value multi-channel influences, reducing theft from tools like Honey.
• Curate high-quality data sources, transcribing YouTube videos for reliable AI review synthesis.
• Build merchant tools for AI compatibility, emphasizing structured intent signals over human UX.
• Emulate Costco's trust model in digital spaces, using subscriptions to enforce quality curation.
• Focus AI on utility fulfillment rather than demand generation, avoiding trend prediction pitfalls.
• Integrate real-time APIs to combat hallucinations, ensuring accurate product availability and pricing.
• Target mid-consideration purchases like electronics with conversational agents simulating expert advice.
• Mainstream cashback agents for all users, simplifying interfaces to appeal beyond tech-savvy savers.
• Monitor e-commerce latency, accelerating delivery to compete with in-person immediacy demands.
• Diversify beyond aggregators by creating proprietary products or recurring services for durability.
• Leverage video authenticity in AI training, weighting unsponsored content for unbiased outputs.
• Prepare for Google's premium retention by innovating in non-monetizable informational queries.
• Expand AI into financial services like low-margin loans, building on membership trust ecosystems.

MEMO

In an era where the internet feels increasingly cluttered with SEO-manipulated drivel, a16z partners Alex Rampell and Justine Moore argue that AI agents could revitalize online shopping by cutting through the noise. Drawing from Rampell's experience founding affiliate giant Trial Pay, they dissect how these intelligent intermediaries might redirect consumer journeys, potentially eroding Google's vaunted search dominance. "Google kind of is a tax on GDP," Rampell notes, highlighting how the tech behemoth siphons commissions from vast consumer spending. Yet, as search volumes dip—Rampell admits using ChatGPT "three orders of magnitude more" than Google—the duo envisions a shift where AI handles not just queries but transactions, from price hunts to personalized picks.

The conversation pivots to purchase psychology, distinguishing impulse buys—like TikTok-fueled T-shirt grabs—from deliberated splurges on homes or cars. Impulse decisions, emotionally driven and instantaneous, resist AI's rational nudge, much like Coca-Cola's checkout line premium. High-stakes buys, meanwhile, demand tactile verification, blending online research with in-store touches. The sweet spot lies in mid-tier items: a travel handbag fitting laptops and water bottles, or a durable bike. Here, AI excels, sifting Reddit threads and TikToks to recommend options, perhaps even auto-buying when prices dip. Viral anecdotes, like teens reverse-engineering Taylor Swift's street style via ChatGPT, underscore this potential, though early hallucinations—recommending nonexistent leggings—have tempered enthusiasm.

Affiliate marketing's shadowy underbelly emerges as a core villain, born from porn-site tracking and reliant on cookies that enable "attribution theft." Tools like Honey swoop in at checkout, claiming undue credit for sales influenced by Reddit posts or Super Bowl ads. This last-click fallacy, Rampell laments, corrodes the web, much as commoditized resellers like Casper flop without proprietary goods. Factories in Shenzhen churn identical mattresses for countless brands, dooming middlemen to price wars. Aggregators like Amazon and Shopify prevail by riding trends—yesterday's Allbirds, today's On Running shoes—while single-SKU players chase fleeting fads.

E-commerce's modest 16% retail share surprises, explained by humanity's craving for immediacy: real-time toothpaste trumps overnight delivery, and mall browsing offers irreplaceable serendipity. Online research often precedes in-person buys, blurring lines and baffling attribution. Rampell praises sites like CamelCamelCamel for chronicling price-conscious behavior, predicting AI will close the loop—alerts evolve into autonomous purchases. Yet, the web's "unhealthiness," as former Firefox CEO John Lilly termed it, stems from commercialization: top-10 lists masquerade as advice but chase affiliate bucks, polluting data AI must navigate.

Google's freemium fortress, built on AdWords' elegant fusion of paid and organic results, endures for transactional intent. Informational queries—like Oscar winners—flee to ChatGPT's 800 million users, but revenue holds as premium searches persist. Moore flags AI's product pitfalls: fabricated recommendations drove experimenters back to Amazon. The internet's walled gardens—X for real-time, Facebook for social—further fracture unified search, while junk content defies summarization. "You can't turn shill junk into honest analysis," Rampell warns, advocating "decrapification" through trusted sources like ad-free Consumer Reports or YouTube's accountable creators.

Costco stands as commerce's gold standard, its membership model—netting billions from 50 million dues-paying loyalists—forcing low margins and rigorous curation. Refusing 50% shirt profits preserves trust; Kirkland generics rival Lululemon (hence the lawsuit) at fractionally the price. Unlike Amazon's ad-fueled, review-rigged bazaar—where AliExpress knockoffs swap fake stars—Costco treats shoppers as insiders, immune to digital disruptions. Bezos once contrasted Apple's margin-maximizing ethos with Amazon's scale-at-all-costs approach; Costco threads a unique needle, proving trust's enduring value.

Looking ahead, Rampell and Moore foresee new winners in AI-native niches: hyper-optimized price agents mainstreaming cashback for retirees, or fine-tuned models for bike consultations mimicking Reddit's buy-it-for-life wisdom. UPC-coded goods invite automation—feed a SKU, snag the best deal—while bespoke items like Wayfair stools demand hybrid smarts. Merchants must evolve, rendering sites agent-friendly with financial plumbing for cardless buys. OpenAI's commerce push signals horizontals' ambition, but vertical specialists could thrive, wresting control from Amazon's ad skew.

Ultimately, AI promises a more intuitive commerce layer, where agents act as tireless concierges for the value-driven. Yet success hinges on trust and accuracy, lest hallucinations or pollution perpetuate the web's malaise. As Rampell reflects on Trial Pay's cookie era, the affiliate ghost may linger, but intelligent automation could finally empower consumers over middlemen, fostering a cleaner, more efficient marketplace.