Overview of "The DoorDash Problem: How AI browsers are a huge threat to Amazon"
Nilay Patel (Decoder, The Verge) unpacks what he calls the "DoorDash problem": the threat posed when AI agents (AI-powered browsers/assistants) act on users' behalf to interact with websites and services. Using the recent Amazon v. Perplexity lawsuit as a flashpoint, he examines how agentic AI could disintermediate platforms (DoorDash, Amazon, Uber, Airbnb, etc.), erode revenue streams like ads and subscriptions, and reshape who "owns" the customer relationship online.
Key takeaways
- The "DoorDash problem": AI agents that place orders or book services for users can bypass app/website interfaces, removing ads, loyalty hooks, upsells and the direct relationship platforms use to monetize users.
- Amazon sued Perplexity after Perplexity’s Comet browser/agent automated shopping on Amazon in ways Amazon says violated its terms and technical protections; Amazon invoked the Computer Fraud and Abuse Act (CFAA).
- Different platform CEOs have varied views: some are confident their operational moats protect them; others (notably Amazon) are more aggressive in defense. Many want to experiment first and sort economics later.
- The biggest near-term business impact is on commerce-adjacent revenue (Amazon’s advertising and Prime) because agents can hide ads and prioritize lowest price across retailers.
- The conflict raises legal, technical and ethical questions about whether user agents should have the same access as a human user, and how labor/real-world service providers are paid if agents mediate transactions.
What the "DoorDash problem" means
- Definition: When users send an AI agent to carry out transactions (order food, book rides, buy goods), the agent interacts with service providers instead of users directly using apps/sites.
- Consequence: Platforms lose the ability to monetize attention via ads, promotions, subscription hooks, cross-sells and branded experiences. Services risk becoming commodity backends (databases of supply) competing mostly on price.
- Historical context: Comparison to App Store era — just as apps centralized customer relationships on mobile, agentic AI could centralize them behind agent interfaces.
The Amazon vs. Perplexity confrontation (what happened)
- Perplexity launched Comet, an AI-first browser on Chromium, with an agent that automated shopping on Amazon.
- Amazon alleges Perplexity ignored requests to disable this, masked agent traffic as human Chrome users, and circumvented Amazon's technical blocks.
- Amazon filed suit claiming violations of its terms and the CFAA (U.S. anti-hacking statute).
- Perplexity characterizes Amazon’s action as bullying; its stance: user-owned agents should act like human users and have the right to access sites on behalf of users. Perplexity has a history of scraping/using web content and has faced other lawsuits (Wall Street Journal/Dow Jones, Britannica, Merriam‑Webster, Reddit) and later struck licensing deals in some cases.
Industry reactions and CEO perspectives
- Lyft (David Risher): Dismisses the extreme unbranded-agent scenario; expects brand preference to persist; believes platforms can compete if agents abstract them.
- Zocdoc (Oliver Carrez): Confident in established moats and integration work; skeptical that agents will fully displace specialized services.
- TaskRabbit (Ania Smith): Argues platforms have vetted labor and supply networks agents can’t replicate easily.
- Uber (Dara Khosrowshahi): Pragmatic — wants to work with agents, suggests initially charging zero to test experience and measure incrementality, then charge if agents cannibalize demand.
- Airbnb (Brian Chesky): Warns against AI maximalist view; companies will only participate if they can maintain customer relationships.
- Google/Microsoft (Sundar Pichai, Kevin Scott): Emphasize efficiency and business economics — agentic interactions must make commercial sense for platforms to participate.
- DoorDash statement: Integrations exist (ChatGPT, Google, Yelp); DoorDash’s claim is value lies in end-to-end experience and fulfillment, not just where orders originate.
Why Amazon may be uniquely aggressive
- Advertising revenue: Amazon’s ad business is huge and growing (on pace to exceed ~$60B in 2025 per the episode). Agents that automate shopping could hide sponsored placements, undercutning a major revenue stream.
- Prime/subscription leverage: If agents redirect buyers to lowest-price options across retailers, Prime’s convenience/value could be eroded.
- Commodity risk: Many retail purchases (e.g., low-differentiation items) are easy for agents to source elsewhere, making Amazon potentially more replaceable in agent-driven flows than service companies with real-world supply networks (rides, tasks, healthcare).
- Amazon already builds agent tools (Rufus, Buy From Me) suggesting it sees both the risk and the opportunity.
Perplexity's argument and legal/ethical questions
- Perplexity claims user agents are extensions of users and should have the same web access rights; frames Amazon’s suit as anti-user.
- It argues software is becoming labor (agents acting on behalf of users) to justify access.
- Legal ambiguity: Do agents constitute users for ToS and anti-hacking laws? CFAA claims raise contentious legal and technical debates.
- Ethics & labor: If agents strip away the revenue streams that pay gig workers (drivers, delivery workers, taskers), platforms and regulators must consider how those jobs remain viable.
Open questions and possible outcomes
- Will courts treat agent interactions as user actions or as unauthorized access? Outcomes could set broad legal precedents.
- Will platforms build APIs/partnership frameworks and charge take rates to agents (like App Store-style deals), or will they block/sandbox agents?
- Will agent companies negotiate licensing or be forced to adapt technical/contractual approaches?
- Will some services be inherently resilient (unique supply chains, vetted labor) while others (commodity retail) get reorganized by agents?
- How will consumer experience, security, fraud prevention, and worker compensation evolve under agentized commerce?
Practical recommendations for companies (actions to consider)
- Experiment: Pilot agent integrations to measure experience and whether agent-sourced users are incremental or cannibalistic.
- Develop commercial frameworks: Create APIs, licensing terms, and measurable take-rate models for agents that want access.
- Build detection & policy: Strengthen traffic attribution and specifically define acceptable automated access; plan negotiation playbooks.
- Protect core monetization: Re-think ad formats and subscription value that persist even when agentizing front-ends (e.g., guaranteed fulfillment SLA tied to platform).
- Advocate for clarity: Work with regulators and industry groups to define legal standards for agent access and liability.
- Consider worker economics: Ensure that any agent-driven model preserves payment and protections for real-world labor.
Notable quotes
- Amazon (summary): "Third-party applications that offer to make purchases on behalf of customers should operate openly and respect service provider decisions whether or not to participate."
- Perplexity (blog): "Bullying is not innovation" — argues agents are the user's agent and should be allowed to transact on users' behalf.
- Dara Khosrowshahi (Uber): "Initially I charge you zero... let's try it out... then we can measure incrementality; if cannibalistic, then I’m going to charge a lot."
- Demis Hassabis (Google DeepMind): "We might not even need to render webpages at all in an agent-first world" — illustrates the technical reimagining of the web agents imply.
Final thoughts (from the episode)
Nilay Patel frames the debate as the opening salvo in a larger battle over who controls the browsing layer and economic experiences on the web. The Amazon–Perplexity case could set critical precedents. Even if agentic shopping is imperfect today, the incentives and investments are real — and companies, courts, and regulators will need to define technical, commercial and legal rules that balance innovation, platform economics, and worker livelihoods.