If an AI agent earns a referral fee when you buy, can it still be fully on your side?
Do referral-fee incentives compromise an agent's loyalty to users?
This explores whether an AI shopping or booking agent that earns money from merchants, through referral fees, affiliate commissions or ad placements, can still be trusted to act in the user's best interest, and what the corpus says about fixing that conflict.
This explores whether getting paid by merchants turns an AI agent into a salesperson posing as your assistant. The strongest voices in the corpus say yes, and they say the cause is the business model, not the technology. Can an AI agent serve both merchant and user interests fairly? uses Meta's Muse earning referral fees from Expedia as the case: an agent paid by one side of a transaction can't be fully loyal to both sides. He adds that the same fees explain which platforms choose to build agents and which choose to block them. Do advertising-funded shopping agents favor paid product placements? makes the same point about advertising. Even a well-engineered agent leans toward paid placements when ads pay the bills.
The pressure is likely to grow over time. How could an AI agent's revenue per transaction grow over time? sketches how agent revenue could climb in stages: thin payment commissions first, then roughly 4% affiliate fees, then booking-platform-style margins of around 12% in some categories. Each stage ties the agent's income more closely to which merchant wins the sale. So the loyalty question gets harder as agents mature, not easier.
The experimental work complicates the clean story in two ways. First, the bias can be surprisingly easy to fix. Can naming the consumer fix AI agent sponsorship bias? found that agents working for a platform went easier on sponsored products than agents working for a consumer. Adding one sentence to the prompt naming the consumer as the beneficiary removed that gap, even though the platform was still the employer. Loyalty, at least for today's language models, partly depends on whom the agent is told it serves. Second, Do agent mistakes or misaligned incentives cause marketplace welfare loss? found that in simulated marketplaces, most of the harm to users came from agents simply making mistakes in complex negotiations, not from agents serving the wrong master. Right now, competence may hurt users more than conflicts of interest do.
Loyalty also covers more than which product gets recommended. It includes what the agent gives away. Can a seller infer what a buyer really wants to pay? shows that a seller-side agent could read a buyer agent's dialogue and work out the buyer's maximum price almost perfectly when the buyer played a described persona, though much less well when the buyer only held a hidden numeric budget. An agent that never takes a referral fee can still let the merchant see your budget. That's why Must AI monitoring precede incentive alignment as agents evolve? argue that monitoring has to come before any attempt to align incentives: you can't reward loyal behavior you can't observe.
Consumers seem to sense the risk without the theory. Does naming a payments brand increase trust in AI transactions? found that trust in AI-executed payments jumped from 23% to 61% when a familiar payments brand stood behind the transaction. Do consumers actually want to delegate shopping decisions to AI? found that people who say they'd let an agent shop for them still want to choose the brands themselves. Put the findings together and the practical answer is that referral fees create a real pull away from the user. Whether it does damage depends on three things you can check: whose interest the agent is explicitly told to serve, what it reveals to the other side, and whether anyone is watching.
Sources 9 notes
Azhar argues that agents like Meta's Muse, which earn referral fees from merchants like Expedia, face structural conflicts that prevent unbiased recommendations. The incentive to collect fees, not technology, determines which platforms build or block such agents.
Bob Hedges, former Visa Chief Data Officer at MIT, warns that advertising business models create structural incentives for shopping agents to favor paid placements over neutral recommendations, independent of technical design quality.
Jaipuria models a three-stage revenue path: starting with sub-0.5% payments commissions, adding 4% affiliate fees, then moving to 12% aggregator take in suitable categories. The blended progression suggests agent economics scale from minimal initial fees toward booking-platform-like margins.
Adding a single sentence to an AI agent's prompt that names the consumer as the recommendation's beneficiary removed the sponsorship penalty attenuation caused by platform employment, restoring it to consumer-delegation levels without changing who employs the agent.
Magentic Marketplace experiments traced welfare shortfalls to agent competence failures in complex negotiation, not to agents serving competing interests. This supports human-in-the-loop designs where agents assist rather than replace human decision-making.
Show all 9 sources
A seller-side inference agent recovered a consumer's target willingness to pay with a slope of 1.00 from dialogue transcript alone, when the buyer inhabited a verbal persona. A numeric budget kept confidential produced only a 0.21 slope, showing behavioral coherence with a persona is far more legible than explicit numbers.
Humberd and Latham argue that as AI becomes more autonomous, monitoring mechanisms must be established first as prerequisites to incentive alignment, because defining desirable AI behaviors is harder than for humans and AI decision-making becomes increasingly incomprehensible over time.
A Harris Poll survey found 72% of US adults use AI assistants, but only 23% trust AI to execute payments. When Visa's name was attached, 61% reported trust in the company to handle agent-initiated transactions—suggesting consumers prefer a recognized intermediary over AI alone.
Accenture's 25,590-person survey found 74% would trust an AI agent with purchases, yet 56% would specify brands and 37% of loyal consumers wouldn't switch. Delegation appetite depends on trust protections being in place first, not on capability alone.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Whom Do AI Agents Work For? Role Assignment Induces Sponsorship Bias in LLM Recommenders
- When Agents Shop for You: Role Coherence in AI-Mediated Markets
- Vertical tacit collusion in AI-mediated markets
- Agentic Commerce and Incentives
- Agentic Commerce: Why Do Shoppers Delegate Purchase Decisions to AI?
- Visa Research: Consumer Trust is Enabling Path to Agentic Commerce
- The Automated but Risky Game: Modeling and Benchmarking Agent-to-Agent Negotiations and Transactions in Consumer Markets
- Consumer Pulse Research 2026: Talk to my AI agent