Skip to content
a16za16z

The Death of Search: How Shopping Will Work In The Age of AI

The web is unhealthy, and AI agents are about to rewrite how we shop. In this episode, a16z General Partner Alex Rampell and Partner Justine Moore explore how AI agents will change commerce and the implications for Google’s business model, affiliate marketing, online shopping, and more. Timecodes: 0:00 Introduction 0:36 Affiliate Marketing and the Evolution of Online Purchases 1:59 Impulse Buys vs. Considered Purchases 2:51 Observing Consumer Behavior and Viral Trends 6:09 Dynamic Pricing and Consumer Surplus 7:14 E-commerce vs. In-Person Shopping 10:20 The Challenge of Attribution in Online Sales 12:28 Aggregators, Brands, and the Commodity Trap 15:36 Trends, Trust, and the Power of Aggregators 17:46 Google’s Freemium Model and the Future of Search 21:17 The Problem of Hallucination in AI Product Recommendations 22:52 The Commercialization and Pollution of the Internet 29:30 The Costco Model: Trust and Value 33:20 The Spectrum of Purchases and AI’s Role 37:57 UPCs, SKUs, and Automation in Commerce 39:52 Opportunities for New Winners in AI-Driven Commerce Resources: Read the article: http://a16z.com/ai-x-commerce/ Find Alex on X: https://x.com/arampell Find Justine on X: https://x.com/venturetwins Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://x.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Podcast on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Podcast on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.

Alex RampellhostJustine Moorehost
Sep 17, 202545mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 0:44

    The web is “crap”: SEO spam, polluted marketplaces, and why search feels broken

    Alex argues the open web has degraded: low-quality, SEO-optimized pages dominate, and even major marketplaces are flooded with dubious products and reviews. This sets up the core problem for AI and search: summarizing bad inputs won’t create trustworthy outputs.

    • SEO incentives reward low-quality affiliate content
    • “De-crapifying” the web is harder than improving retrieval
    • Amazon itself behaves like a polluted search engine in many categories
    • Trust and information quality become the bottleneck for AI shopping
  2. 0:44 – 1:44

    Why this conversation now: Google’s weakening hold and affiliate marketing’s role in commerce

    Alex explains his background in affiliate marketing (TrialPay) and why AI-driven shifts threaten the existing click-based commerce pipeline. The discussion frames Google as a dominant starting point for purchases and asks where that value capture goes if behavior changes.

    • Affiliate marketing mechanics: cookies, pixels, confirmation-page tracking
    • The looming question: does affiliate attribution still “power” commerce in an agent world?
    • Personal behavior shift: using ChatGPT far more than Google for non-commerce queries
    • Google’s commerce role is massive because it sits at the start of purchase journeys
  3. 1:44 – 3:53

    Impulse buys vs. researched purchases: where AI helps (and where it doesn’t)

    The hosts separate shopping into impulse purchases versus considered, research-heavy purchases. Alex argues impulse buying is emotionally engineered and inherently anti-agent, while high-stakes items invite heavy research but complicate transaction and monetization models.

    • Impulse purchases are designed to bypass deliberation (e.g., checkout-line pricing)
    • AI is unlikely to be the driver of true impulse spending
    • Considered purchases invite extensive research—ideal for AI assistance
    • Hard question: how to connect AI research to actual transaction and economics
  4. 3:53 – 6:09

    Observed “agent behavior” today: price tracking and viral product-finding use cases

    They focus on present-day signals that predict agentic shopping: tools like CamelCamelCamel show consumers already automate attention around price drops, while teens use ChatGPT for “what is she wearing?” queries. These behaviors suggest demand for AI that can both identify items and take action when conditions are met.

    • CamelCamelCamel proves consumers will act automatically when price thresholds hit
    • “The consumer is the agent” today—just inefficiently and manually
    • Viral examples: image-based fashion identification, plus cheaper alternatives
    • Youth behavior is treated as a leading indicator of mainstream adoption
  5. 6:09 – 7:09

    Dynamic pricing and consumer surplus: the temptation—and backlash—of personalization

    They explore the economic logic of charging each customer a different price to capture consumer surplus, and why it’s difficult in practice. Regulatory risk, customer outrage, and failed historical attempts make “perfect” dynamic pricing hard to sustain.

    • Personalized pricing is attractive economically but unpopular socially
    • Signals like device type (iPhone vs Android) can imply willingness to pay
    • Airlines (e.g., Delta) experimenting highlights the direction of travel
    • Practical constraints: regulation, PR blowback, and trust erosion
  6. 7:09 – 10:20

    Why e-commerce isn’t as dominant as predicted: immediacy and shopping-as-entertainment

    Alex explains why in-person retail persists despite online convenience: some demand is time-critical (buy toothpaste now), and some shopping is experiential (browsing malls, aspirational purchases). Justine adds that many “offline” purchases still involve heavy online research, blurring measurement.

    • Two key forces: immediacy needs and experiential/aspirational shopping
    • Shipping improvements expanded online demand but didn’t eliminate offline retail
    • Research-online-buy-offline behavior complicates “e-commerce share” stats
    • Returns and local store density influence online apparel purchasing patterns
  7. 10:20 – 12:28

    Attribution is broken: last-click economics, coupon ‘theft,’ and AI making it harder

    They argue attribution is the most corrosive issue in online commerce, with last-click models distorting incentives. Alex criticizes coupon extensions (e.g., Honey/RetailMeNot) for capturing credit at checkout, and predicts AI will further complicate multi-touch journeys and who gets paid.

    • Last-click attribution confuses correlation with causation
    • Coupon tools can ‘steal’ attribution right before purchase
    • AI adds more touchpoints: Reddit, ads, LLM research, then purchase
    • Determining fair value split across influences remains unsolved
  8. 12:28 – 15:35

    Why aggregators win and brands struggle: the commodity trap and lack of defensibility

    Alex explains that many DTC brands are effectively marketers, not manufacturers, buying traffic from Google/Facebook and selling commoditized goods. Competition quickly copies the product and undercuts price; without subscription or repeat purchase, unit economics degrade and the brand becomes fragile.

    • Many DTC ‘brands’ don’t manufacture; OEM supply makes copying easy
    • Google/Facebook captured much of the value via paid acquisition
    • One-time purchases (e.g., mattresses) force constant reacquisition
    • Subscriptions can help, but even then categories commoditize over time
  9. 15:35 – 18:00

    Trends, demand creation, and the limits of AI: trust still flows through culture

    Justine emphasizes trend-driven consumer categories (shoes, makeup) where tastes shift rapidly and aggregators benefit by carrying everything. Alex adds AI struggles to “inculcate demand” because demand often starts with cultural exposure (e.g., TikTok), not a rational query—meaning discovery engines remain powerful.

    • Trends rotate quickly; single-brand focus is a structural disadvantage
    • Aggregators can ride any SKU trend, while brands miss shifts
    • AI can optimize fulfillment of known intent but is weak at generating desire
    • Cultural feeds (TikTok/Instagram) remain central for demand creation
  10. 18:00 – 21:13

    Google’s freemium model and the ‘tax on GDP’: what AI is (and isn’t) taking away

    Alex frames Google as the canonical freemium model: free informational queries plus monetized commercial intent via ads that improved search relevance. He argues AI is mostly siphoning off the “free” informational queries for now, while the monetized “imium” remains—though the commerce tax could shift to new intermediaries.

    • AdWords made search both monetizable and, in some cases, better (relevance via clicks)
    • AI is pulling informational queries (non-monetizing) away from Google
    • Google revenue can rise even if search volume falls, implying mix shift
    • If commerce journeys start elsewhere, Google’s ‘tax’ may migrate
  11. 21:13 – 22:52

    LLM hallucinations block AI shopping today—and why up-to-date product truth matters

    Justine highlights a core blocker: LLMs often hallucinate products, prices, or availability, pushing users back to Google/Amazon for purchase decisions. Until product data is grounded, current, and transactional, AI’s role in shopping remains limited.

    • Natural-language shopping queries are ideal—but accuracy is the constraint
    • Common failure modes: nonexistent items, outdated products, wrong prices
    • User trust is fragile; early experiments often revert to old platforms
    • OpenAI and others are working to integrate real-time commerce information
  12. 22:52 – 29:22

    Commercialization, walled gardens, and ‘de-crapifying’ content: the missing Consumer Reports

    Alex argues the web’s incentives degraded content quality: affiliate monetization and SEO spam crowd out honest reviews, while walled gardens fragment what’s searchable. They discuss how the old Consumer Reports model (no ads, subscription trust) largely disappeared, leaving AI to summarize a compromised corpus.

    • Search is already fractured across X, Facebook, and other walled gardens
    • Affiliate economics incentivize “top 10” shill content and keyword spam
    • Summarization can’t convert biased junk into trustworthy guidance
    • Video reviews may be higher quality but are less skimmable/indexed
  13. 29:22 – 33:19

    Trust as a business model: Costco vs. Amazon vs. Apple

    Alex presents Costco as an “AI-proof” model built on extreme trust and membership economics: low margins protect the membership’s value, and curation prevents low-quality goods. They contrast this with Amazon’s scale/low-price ethos (and ad monetization) and Apple’s premium pricing strategy.

    • Costco’s profits map closely to membership fees, incentivizing customer value
    • Strict margin discipline and curation preserve trust over time
    • Amazon tolerates ‘sea of crap’ dynamics; advertising becomes its best ‘SKU’
    • Premium brands (Apple) optimize for maximum willingness-to-pay vs minimum price
  14. 33:19 – 37:37

    Which purchases will be ‘AI’d’: a spectrum from TikTok impulse to high-stakes decisions

    Justine outlines where AI agents can meaningfully change shopping: not at pure impulse or extremely high-consideration ends, but across the broad middle. She describes AI helping with research synthesis (Reddit/TikTok aggregation), personalized recommendations, and eventually agentic purchasing when integrations mature.

    • Impulse buys are immediate and feed-driven, not research-driven
    • Very high-stakes purchases still require human/in-person confirmation
    • The middle is ripe: travel bags, laptops, bikes, couches—complex but tractable
    • Agents can optimize price, timing, and preferences once trust and tooling exist
  15. 37:37 – 39:52

    UPCs, SKUs, and automation: when commerce becomes an ‘execute’ problem

    Alex adds a practical lens: whether a product has a UPC/SKU determines how automatable buying can be. Once the item is uniquely identified, AI can optimize price, coupons, cashback, credit card choice, and shipping—automating what “money-over-time” shoppers do manually today.

    • UPC/SKU identification enables straightforward price comparison and execution
    • Pre-AI, consumers manually optimized deals; AI can automate end-to-end
    • Products without UPCs (e.g., many furniture items) require different discovery flows
    • Agent automation expands bargain-hunting beyond technical power users
  16. 39:52 – 45:04

    New winners in AI commerce: specialized shopping agents and merchant infrastructure

    They conclude with where durable new companies could emerge: consumer-facing specialized agents that own the ‘last click’ in an agentic world, and merchant-side infrastructure to serve AI browsers and buyers. The shift may threaten incumbent presentation layers (Google/Amazon ads) and create new standards for agent-compatible checkout and site design.

    • Specialized agents could beat horizontals by optimizing narrow purchase workflows
    • Opportunity: automate coupons, cashback, and payment optimization at scale
    • Merchant-facing shift: sites must become agent-browsable and agent-transactable
    • Infrastructure needs: delegated payments, authorization, and new web interaction patterns

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.