OpenAI and Meta are building revenue models where they earn a fee every time their AI recommends a product and a user buys it. That sentence sounds like a tech story. For anyone who manages sales commissions, runs an ISO, or operates a channel sales team, it’s also a structural warning about where platform economics are heading.
Here’s the short version: both companies are experimenting with taking a percentage cut from merchant transactions that flow through their AI assistants. OpenAI tested a 4% transaction fee through a feature called Instant Checkout. Meta partnered with Shopify in September 2026 to let its Muse AI agent complete purchases through Shop Pay. Neither company has finalized its fee structure, and both have already changed course at least once.
What neither has figured out yet is how to handle attribution, transparency, or the conflict between earning commissions and giving honest recommendations. Those are problems that commission-based businesses have been solving, imperfectly, for decades.
What “commission-based AI” actually means (and what it doesn’t)
The word “commission” is doing a lot of work in coverage of this topic, and it’s worth being precise about which model we’re talking about.
There are three structures in play:
- Referral or affiliate fees: The AI platform directs a user to a merchant. If the user buys, the platform earns a percentage of the sale. No purchase, no fee.
- Transaction fees: The AI hosts the checkout itself. Every completed transaction inside the AI surface triggers a fee, similar to how a payment processor charges per transaction.
- Lead-generation fees: The platform charges merchants for sending qualified traffic, regardless of whether a sale closes. This is closer to advertising, but tied to intent signals rather than impressions.
OpenAI has experimented with options two and three. Meta appears to be running a version of option one through its Muse agent, with transaction fees layered in through the Shopify partnership.
This is meaningfully different from subscriptions (you pay monthly whether or not you get value) and from advertising (you pay for exposure, not outcomes). Commission ties the platform’s revenue to a completed transaction. That’s the same logic that drives sales agent compensation, ISO residuals, and insurance distributor fees. The AI platforms are borrowing a model that commission-driven businesses have used for a long time.
OpenAI’s path from chatbot to checkout counter
Product recommendations without paid placement (the stated goal)
OpenAI states that product recommendations inside ChatGPT are ranked by relevance, not by merchant payments. The framing is that ChatGPT functions like an honest advisor, surfacing the best match for what a user is asking, without the sponsored-placement dynamics of Google Shopping or Amazon search results.
Sam Altman floated an affiliate-style fee of around 2% in earlier discussions, according to an analysis published by the Center for Democracy and Technology. The CDT’s reading of the situation is direct: when a platform earns money from the transactions its recommendations generate, there’s a structural pull toward recommending things that convert, not things that are best.
OpenAI hasn’t answered this tension publicly. “We don’t take payments for rankings” and “we earn a fee when users buy” can coexist in the short term. Over time, they create competing incentives that will be hard to ignore.
Instant Checkout, the 4% fee, and the pivot away
In September 2026, OpenAI launched Instant Checkout, allowing users to complete purchases without leaving ChatGPT. By January 2026, Shopify merchants were reportedly paying a 4% transaction fee on purchases completed through the feature, according to reporting covered by Productsup’s agentic commerce tracker and MakeInfluence.
Then, in March 2026, CNBC reported that OpenAI pulled back from native checkout in favor of letting merchants build their own apps inside ChatGPT. The platform would facilitate the transaction environment; merchants would own the checkout flow.
That pivot tells you something. Owning checkout means owning liability, customer data, fraud risk, and dispute resolution. OpenAI apparently decided that wasn’t a business it wanted to be in, at least not yet. The shift to merchant-built apps reduces OpenAI’s operational burden while keeping the revenue relationship intact through app store-style fees or referral arrangements.
The revenue math at scale
The arithmetic is straightforward. At 2%, $10 billion in AI-attributed purchases generates $200 million. At 4%, $400 million. Those numbers sound large until you look at OpenAI’s overall revenue trajectory.
Sacra estimated OpenAI’s annualized revenue at roughly $40 billion as of mid-2026, with approximately $1 billion coming from advertising. Commerce commission revenue is still small relative to that base. But the direction matters more than the current number.
Reuters Breakingviews reported that OpenAI’s internal revenue targets approach $50 billion by 2030, with advertising and commerce as diversification levers beyond subscriptions and API fees. If ChatGPT reaches the purchase-intent density that would make those projections real, commission revenue stops being a rounding error.
Meta’s version: commerce grafted onto a social empire
Shopping inside Meta AI
Meta tested a product-recommendation carousel inside Meta AI for U.S. users, showing product images, prices, brand names, and merchant links. The feature lets users move from a conversation about what to buy directly to a merchant page.
What Meta has not confirmed is whether those recommendations generate referral commissions or whether advertisers get preferential placement. That silence is not reassuring. Meta’s entire business model is built on selling attention to the highest bidder. If the same dynamic applies inside Meta AI’s shopping recommendations, users have no way of knowing whether they’re getting honest product guidance or paid placement dressed up as AI advice.
Muse, Shopify, and the September 2026 deal
The more significant development is the Shopify-Meta partnership that lets Meta’s Muse agent complete purchases through Shop Pay across Shopify’s merchant network. A transaction fee is reportedly part of the arrangement.
This is structurally different from Meta’s ad model. Advertising charges for exposure. This charges for completed sales. Meta is now running both simultaneously, which creates an interesting internal tension: does a merchant pay for an ad to get visibility, buy into the AI recommendation ecosystem to get conversions, or both?
For merchants operating on thin margins, those fees layer on top of existing payment processing costs, Shopify subscription fees, and whatever ad spend they’re already running.
Why Meta’s position is structurally different from OpenAI’s
Meta has more than 3.3 billion monthly active users across its apps. OpenAI has around 400 million. That difference in scale means Meta’s commission model doesn’t need to be efficient to generate revenue; it just needs a small conversion rate on a massive user base.
Meta also earns approximately $160 billion annually from advertising. Commission revenue doesn’t need to replace that. It layers on top, which takes pressure off any individual transaction fee and makes the model easier to experiment with.
The risk is cannibalization. If Meta’s AI recommendations pull merchant attention away from the ad auction, Meta could undercut its own highest-margin product. That’s a genuine tension that Meta’s finance team is certainly modeling, even if it hasn’t surfaced publicly.
Side-by-side comparison: two companies, two very different bets
| Factor | OpenAI | Meta |
|---|---|---|
| — | — | — |
| Primary AI surface | ChatGPT | Meta AI (across Facebook, Instagram, WhatsApp) |
| Monthly user base | ~400 million | 3.3 billion+ |
| Commission trigger | Transaction completion (Instant Checkout / app fees) | Referral + transaction (Muse / Shop Pay) |
| Merchant relationship | Developer-style app integration | Social advertising + commerce layered together |
| Existing revenue base | ~$40B annualized (subscriptions, API) | ~$160B/year (advertising dominant) |
| Main strategic risk | Can it reach purchase-intent volume to matter? | Will commerce fees cannibalize the ad auction? |
OpenAI is betting on depth: a smaller, more intentional user base with high purchase intent, where agentic task completion drives conversion. Meta is betting on scale: billions of users already in platform, lower friction per transaction, commerce as an add-on to social discovery.
Both bets could pay off. They’re also not mutually exclusive from a market perspective. The more interesting question is what happens when both platforms are competing for the same merchant relationships.
The trust problem no one has solved yet
If an AI recommends a product and earns a fee from that recommendation, how does a user know the recommendation is honest? OpenAI says fees don’t influence rankings. There is no independent audit confirming this.
The CDT’s analysis flags the structural conflict directly: a platform that earns revenue from completed transactions has a financial incentive to surface products that convert, not products that are genuinely best-suited to the user’s needs. Those two things overlap often enough to be plausible, but they diverge in edge cases, and the user has no way to see when that divergence is happening.
This pattern has regulatory precedent. The EU fined Google for favoring its own shopping results in the Google Shopping antitrust case. Apple’s App Store commission structure has faced sustained legal challenge in multiple jurisdictions. “Platform as recommender and as beneficiary” is a combination that regulators have already examined, and AI-mediated commerce is the next version of that problem.
For anyone managing a sales team on commission, this conflict is not abstract. Incentive design determines behavior. If a sales rep’s commission is tied to closing, they’re going to close. If an AI’s commission is tied to transaction volume, it will optimize for transaction volume. These are the same dynamic. The difference is that your sales team’s incentive structure is something you can see, negotiate, and audit. An AI platform’s incentive structure is opaque by default.
The attribution mess
Traditional affiliate marketing tracks sales through cookies, referral codes, and trackable links. AI agents break most of those mechanisms. A user asks ChatGPT for a product recommendation, clicks through to a merchant site without a proper affiliate tag, and completes a purchase. Who gets credit?
MakeInfluence’s analysis of AI shopping agents identifies three attribution models currently in use: discovery-and-redirect (AI suggests, user goes to merchant site directly), agent-hosted checkout (AI captures the transaction), and merchant-owned checkout embedded in the AI surface (developer-built apps). Each model creates different attribution gaps.
As of mid-2026, there is no universal standard for passing affiliate or influencer attribution data through AI agents. Creators and affiliates risk losing commission credit when an AI intermediates the transaction. That’s a problem for individual affiliates, but the same structural gap affects internal sales teams.
Consider this scenario: an ISO agent sources a merchant relationship, introduces the business owner to a payment processing solution, and spends three weeks on follow-up. The merchant then completes onboarding through a ChatGPT-integrated checkout process. Under most current commission management setups, that transaction may not connect back to the ISO agent who originated it. The platform earns its fee. The agent earns nothing, or gets into a dispute about attribution.
This isn’t hypothetical. As AI-mediated purchasing becomes more common in B2B contexts, this kind of attribution gap will create real compensation disputes. Compensation expectations that actually retain sales agents matter precisely because gaps like this erode trust in ways that are hard to recover from.
What this means for commission-based businesses
If you manage sales commissions, your attribution model needs stress-testing
AI-mediated purchases may not map cleanly to your existing CRM or commission tracking workflows. A transaction that originated with an agent conversation but completed through an AI assistant may show up in your data as a direct sale, a platform referral, or simply unattributed.
Commission management software needs to distinguish between AI-originated leads, agent-originated leads, and hybrid paths where both contributed. If your agents are using ChatGPT or Meta AI to research prospects or draft outreach, and those same platforms are also earning fees on the resulting transactions, your commission calculations need to account for a new variable. Accurate commission calculation becomes harder when the transaction path involves third-party AI intermediaries.
If you sell through channels, your partners may face new platform fees
ISOs and payment processors already operate on margins where a single percentage point matters. A 2% to 4% platform fee on AI-mediated transactions isn’t a rounding error in that context. It’s the difference between a profitable channel and a break-even one.
Merchants and channel partners should be asking, before they integrate with any AI commerce platform, how attribution is handled, what the fee structure is, whether it’s guaranteed to stay fixed, and whether the platform’s commission on a transaction reduces or replaces any other fees in the stack.
Commission transparency becomes a competitive advantage
When AI platforms obscure the line between honest recommendation and paid placement, businesses that offer clear and auditable commission structures to their agents have something valuable to offer. Your agents know exactly how their commissions are calculated. They can see the logic. They can dispute errors with data.
That transparency is exactly what AI platforms are currently struggling to provide to users and merchants. Automated commission management doesn’t just reduce errors; it builds the kind of trust that becomes a differentiator when the broader environment is murky.
What to watch over the next 12 months
A few specific developments will determine how significant this shift becomes:
- Whether OpenAI standardizes its merchant fee structure or continues experimenting with different models per integration partner.
- Whether Meta discloses commission rates and ranking criteria for AI shopping recommendations, either voluntarily or because regulators require it.
- Whether the FTC or EU regulators require AI platforms to disclose commission relationships at the point of recommendation, similar to existing influencer marketing disclosure rules.
- Whether Shopify, Stripe, or other commerce infrastructure providers establish themselves as the default checkout layer for agentic commerce, which would reduce platform lock-in and potentially standardize attribution.
- Whether attribution standards emerge that let businesses track AI-influenced transactions alongside traditional sales channels in a single commission management system.
The companies building these models are still figuring out what they want to be: marketplaces, referral engines, or app platforms. Each option carries different fee structures, different merchant relationships, and different implications for the businesses that sell through or alongside these platforms.
What’s already clear is that the revenue model is commission-based in structure, regardless of what terminology the platforms use. And for businesses that have been managing commission structures for years, the patterns here are recognizable: outcome-based pay, attribution disputes, incentive misalignment, and the persistent tension between what the commission structure rewards and what the customer actually needs.
Frequently asked questions
What is commission-based AI monetization? Commission-based AI monetization means an AI platform earns a percentage of a sale when its recommendation leads to a completed purchase. The fee is paid by the merchant, not the user, and is triggered by a transaction rather than an impression or subscription.
What fees is OpenAI charging merchants? OpenAI tested a 4% transaction fee through its Instant Checkout feature starting in early 2026. It subsequently shifted toward a model where merchants build apps inside ChatGPT, which may carry different fee structures. Rates have not been publicly standardized.
How does Meta’s commission model differ from its advertising model? Meta’s advertising model charges for exposure: a merchant pays when someone sees an ad. A commission model charges for outcomes: a merchant pays when a user completes a purchase through Meta’s AI. Meta is now testing both simultaneously through its Muse agent and the Shopify Shop Pay integration.
How does AI-mediated commerce affect sales commission tracking? When an AI agent intermediates a transaction, traditional attribution signals (referral codes, cookies, CRM activity) may not carry through correctly. This creates a risk that agent-originated deals get credited to the platform rather than the sales rep or ISO who sourced them. Commission management systems need to account for AI-mediated transactions as a distinct channel.
Are there regulatory concerns about AI platforms earning commissions from their own recommendations? Yes. The CDT has flagged the structural conflict between earning transaction fees and providing unbiased recommendations. The pattern mirrors antitrust concerns raised about Google Shopping and Apple’s App Store, both of which faced regulatory scrutiny for similar platform-as-recommender-and-beneficiary dynamics.