🩸 🤖 💰 🛒 🎭 ⛓️ 👁️ 🌊 — THE INVISIBLE COMMISSION: When AI Promises to Save You Money but the System Profits Behind Your Back
The Phia investigation, the billion-dollar business of invisible transactions, and the question of who artificial intelligence really serves.
Red Blood Journal | Investigative Report | October 15, 2026
9
The Promise Was Simple. The Business Behind It Was Not.
Imagine walking into a store with a personal assistant whose only responsibility is to protect your interests. This assistant compares prices, finds better deals, examines alternatives, and tells you when you are about to spend more money than necessary. You trust that assistant because its purpose seems perfectly aligned with yours: helping you keep more of your own money.
Now imagine discovering that every time you approach the checkout counter, your assistant may be quietly inserting its own name into the store’s accounting records, collecting a commission for bringing you there even when it played no role in your decision. You might pay exactly the same price, never see the transaction, and never know anything happened. The merchant, meanwhile, could be paying someone who did not earn the referral, while another person who actually introduced you to the product could lose the commission.
This is the larger issue raised by the controversy surrounding Phia, the artificial-intelligence shopping company co-founded by Phoebe Gates and Sophia Kianni. What began as a promising attempt to make online shopping more efficient has become an examination of digital attribution, corporate incentives, executive accountability, and the growing power of software that operates on behalf of people who cannot directly observe everything it does.
The case is particularly revealing because it shows how a service can provide genuine benefits while also creating opportunities for conduct that benefits the service provider at someone else’s expense. Understanding that distinction matters far beyond one startup, one famous family, or even one industry.
Red Blood Journal — Think Beyond the Script
This investigation does not ask readers to accept Phia’s explanations automatically, nor does it ask them to assume that every allegation against the company is true. It asks them to examine what the software reportedly did, when its behavior changed, who benefited, what the financial records reveal, and what questions remain unanswered. The same scrutiny must apply to Red Blood Journal’s own conclusions, because suspicion is not proof, and a powerful narrative is not necessarily an accurate one.
The Company That Promised to Work for the Shopper
Phia was launched in April 2025 by Phoebe Gates, daughter of Microsoft co-founder Bill Gates, and Sophia Kianni, her former Stanford roommate. Their business idea addressed a genuine frustration: shoppers frequently struggle to determine whether an item is fairly priced, whether another retailer offers a better deal, or whether a secondhand alternative provides greater value.
Their proposed solution was a shopping assistant that could search across retailers, compare items, surface discounts, and help customers make more informed purchasing decisions. The company also promoted secondhand shopping and the potential to reduce waste through better purchasing choices.
Those functions were not imaginary. Phia developed a real consumer product, attracted investors, and established commercial relationships with retailers. The underlying business idea had legitimate economic value, which is important because a credible investigation must distinguish an otherwise useful service from alleged misconduct involving particular features.
The financial reception was remarkable. Reporting described approximately $43.5 million raised from investors, with the company’s valuation reaching $185 million by early 2026. Its backers included established venture-capital firms and prominent public figures. Bloomberg had already reported on Phia’s fundraising ambitions in December 2025, demonstrating that substantial investment interest existed before the public controversy.
But this rapid expansion creates the first important question. When a company develops technology that helps people save money, what happens when investors simultaneously expect the company to capture ever-increasing revenue from those people’s purchases?
The two objectives can coexist legitimately. A shopping assistant can save customers money while receiving clearly disclosed commissions for referrals it genuinely generates. Problems arise when the pressure to demonstrate growth becomes more important than accurately measuring the value the business has delivered.
That distinction leads directly to the machinery underneath Phia’s commercial model.
The Money Was in the Attribution
Online affiliate marketing is built around a seemingly simple arrangement. A merchant wants customers, an affiliate helps generate a sale, and the merchant rewards that affiliate with a commission. A writer who recommends a product, a creator who reviews equipment, or a website that compares prices may receive compensation when a customer follows their recommendation and completes a qualifying purchase.
In many affiliate programs, this relationship is tracked through identifiers placed in the shopper’s browser. These identifiers help the retailer or its affiliate network determine which marketing partner should receive payment. Although the actual rules vary across merchants and networks, some programs use forms of last-click attribution, under which a later qualifying referral can receive credit over an earlier one.
The system becomes vulnerable when software can manipulate that attribution without providing the marketing contribution required by the program.
How affiliate attribution can be diverted
Illustrative simplified mechanism, not a reproduction of Phia’s source code. The actual outcome depends on the affiliate program’s tracking and attribution rules.
The importance of this arrangement cannot be overstated. The product the shopper purchases may remain unchanged. The checkout price may remain unchanged. The customer might even receive a legitimate discount. Yet the identity of the party receiving payment can change behind the scenes, transferring value without the consumer’s knowledge.
This is known as cookie stuffing when affiliate-tracking information is improperly introduced so that a party receives credit for activity it did not generate. The technique is not exclusive to artificial intelligence, and it long predates modern AI shopping tools. The Phia case therefore exposes an older weakness in electronic commerce that has acquired new relevance as automated assistants become more capable.
The central question is not whether companies should receive commissions. It is whether they should receive commissions for contributions they actually made.
The July Discovery: What the Software Reportedly Did
On July 9, 2026, Bloomberg published an investigation into Phia’s browser extension. Reporters Olivia Solon, Jeff Kao, and Priyanjana Bengani examined whether the company’s software was receiving credit for sales it did not generate. The investigation described testing across more than 50 retail websites, with related findings involving affiliate-marketing researcher Ben Edelman and Capital One Shopping.
The reports described several mechanisms through which Phia’s extension allegedly interfered with affiliate attribution. One involved background browser activity capable of inserting Phia’s referral identifier. Another concerned interactions with the extension’s pop-up, where an ordinary click could reportedly trigger attribution even without an intentional request for the company’s shopping assistance. A third involved periodic refreshing of tracking information, potentially allowing Phia’s identifier to replace another affiliate’s contribution.
These are materially different behaviors from simply presenting a useful coupon and receiving an authorized referral commission. They raise the possibility that the technology was not merely competing to generate customers, but competing to acquire credit for transactions that already existed.
Phia acknowledged that its browser extension had incorrectly attributed transactions to the company. It explained the problem as a software error associated with a code update and said it had acted quickly to address the issue. Bloomberg’s later testing, as recounted in subsequent reporting, confirmed that the identified misattribution behavior had been stopped.
That corrective action matters and belongs in the record. A company that removes a problematic feature has taken a meaningful step, whether the original feature resulted from negligence, poor oversight, or deliberate misconduct.
However, correcting software does not by itself answer how it came to operate that way, who understood its financial effects, or whether the company had sufficient controls to prevent improperly attributed commissions from becoming recognized revenue.
Those questions became much more serious after Bloomberg continued investigating.
December 2025: The Problem With the Timeline
The most consequential development arrived on August 11, 2026, when Bloomberg published further reporting based on internal communications and technical information. According to that investigation, an internal control associated with automatic coupon or cookie activity had been activated on December 10, 2025, and remained active until July 7, 2026, the date Bloomberg first contacted Phia about the issue.
The approximately 209-day period matters because it spans the holiday shopping season and an important phase of the startup’s fundraising and revenue growth. Rather than examining only an isolated software malfunction, investigators now had reason to examine the relationship between the feature’s operation, management knowledge, and commercial performance.
Reporting on Bloomberg’s findings described internal Slack messages indicating that the founders were discussing mechanisms used to capture affiliate commissions. In particular, messages attributed to Phoebe Gates referred to automatic cookie activity and the company’s effort to monetize transactions, while messages attributed to Sophia Kianni raised additional questions about extension behavior and compliance concerns.
The evidence requires careful interpretation. A message discussing monetization does not, by itself, prove that its author knew a particular implementation violated contractual rules. Technical language can describe both lawful affiliate functionality and prohibited attribution practices. But when reported internal discussions are considered alongside an identifiable software control, compliance concerns, the period during which a feature operated, and its eventual deactivation after outside scrutiny, the company’s explanation requires more examination than the word bug can provide.
The relevant investigation must determine whether the feature was intentionally designed to claim commissions without legitimate referrals, whether managers understood that consequence, whether engineering staff raised objections, and whether internal audits had previously detected the problem.
A named feature is not automatically evidence of fraud. Neither is a management discussion about maximizing revenue. Yet the combination of technical records and reported internal warnings creates a legitimate basis for questioning whether the conduct was accidental, knowingly tolerated, or deliberately encouraged.
That question remains a question unless supported by a legal finding or stronger independently established evidence.
When the Features Stopped, the Money Changed
One of the most revealing aspects of Bloomberg’s subsequent reporting concerns the company’s revenue after the disputed features were disabled. According to the uploaded transcript, average daily revenue reportedly declined from approximately $80,000 to a range of $10,000–$28,000.
Reported daily revenue before and after July 7, 2026
$0$25K$50K$75K$100KBeforeAfter (low)After (high)
Figures reported in the ColdFusion transcript; not audited financial statements.
Approximate values reported in the source video, not independently audited figures.
If these figures are accurate and comparable, the decline represents roughly 65% to 87.5% of the previously reported daily revenue level. That is a substantial financial change, and it makes the disputed monetization functions highly relevant to understanding the company’s business model.
An internal estimate cited in subsequent reporting also suggested that approximately 51% of merchandise value credited to Phia in June was connected to cookie-stuffing activity. Phia disputed the accuracy of that estimate. Importantly, merchandise value is not the same as earned commission revenue or profit, so the 51% figure should not be described as 51% of the company’s revenue without additional evidence.
Phia maintained that other changes to monetization features also contributed to the revenue decline. That explanation is plausible enough to require examination rather than immediate dismissal, because disabling several functions simultaneously makes it harder to isolate the economic effect of any single one.
Nevertheless, this is where the investigation becomes particularly important for investors. If a startup’s commercial performance depends substantially on features that are subsequently removed for compliance reasons, its previous growth figures may not represent what the business can sustainably earn through acceptable practices.
A valuation is not cash in a bank account, and fundraising is not proof of profitability. A $185 million valuation expresses expectations about future success, not an independent certification that past revenue was earned appropriately.
The proper questions therefore extend beyond the conduct of the founders. They include what investors were told, how revenue was categorized, whether affiliate networks raised concerns, and whether internal controls could distinguish legitimately generated transactions from improperly attributed ones.
The company’s financial performance cannot be understood independently of the mechanisms by which that performance was produced.
The Larger Investigation: Phia Was Not the First Warning
The allegations surrounding Phia become more significant when compared with a separate controversy involving Honey, the browser extension acquired by PayPal for approximately $4 billion in 2020. Honey also promised to help consumers find discounts, and it also became the subject of allegations concerning affiliate attribution and the replacement of referral identifiers.
This comparison matters because it moves the investigation beyond one startup and one famous surname. If similar problems can emerge in different companies using related business models, investigators should ask whether the vulnerabilities arise partly from the structure of the affiliate economy itself.
In December 2024, affiliate marketers brought litigation against PayPal involving Honey’s alleged interference with their commissions. The plaintiffs claimed that the extension could replace their tracking identifiers with Honey’s own, allowing PayPal to receive payments that should have gone to the original referrers. PayPal contested the claims, and the first amended complaint was dismissed in November 2025 over deficiencies in establishing the plaintiffs’ standing.
However, the litigation did not end there. The plaintiffs amended their complaint, and on June 22, 2026, U.S. District Judge Beth Labson Freeman denied PayPal’s motion to dismiss that revised pleading. That ruling allowed the case to proceed; it did not establish that PayPal had committed the alleged misconduct.
The timing is revealing. Before the Phia allegations emerged publicly, the broader affiliate-marketing industry was already confronting litigation concerning similar methods of attribution.
There is no verified basis in the sources reviewed to suggest that Phia and Honey coordinated their behavior, shared a scheme, or operated under common direction. The important connection is structural rather than conspiratorial: both controversies concern software that can intervene between consumers, merchants, and the people who actually generate sales.
The technology is capable of providing a genuine service, while the underlying payment system may reward actions that consumers neither request nor understand.
And that leads to an uncomfortable possibility. Some of the most consequential decisions in online commerce may be made not at the moment a consumer chooses a product, but in the invisible technical processes that determine who receives compensation afterward.
The Consumer May Not Be the Only Victim
Much of the public discussion surrounding these cases focuses on whether shoppers received the discounts they were promised. That question matters, but it is only one part of the economic picture. A consumer may pay the expected price while an entirely different party suffers the financial loss.
Consider an independent video creator who spends several days researching a product, testing its performance, explaining its weaknesses, and helping viewers decide whether to purchase it. The creator includes an affiliate link that produces a modest commission when a qualifying sale occurs. That arrangement is transparent when the financial relationship is properly disclosed, and it can support independent publishing without requiring readers or viewers to pay directly for every piece of content.
If a browser extension improperly replaces the creator’s referral identifier at checkout, the extension may receive payment without having made the original recommendation. The consumer might never notice, but the creator could lose compensation for the work that introduced the product.
The same concern applies to independent publishers, comparison websites, educators, product reviewers, and smaller businesses that depend on accurately attributed referrals. In that sense, the economic effect can extend well beyond the immediate parties to a transaction.
Not every overwritten tracking identifier necessarily represents a stolen commission. Different affiliate programs have different contractual rules, and the final entitlement depends on those rules and the activities of the participating affiliates. Nevertheless, the underlying problem is clear: whoever controls the attribution mechanism can potentially influence the distribution of revenue.
The company that operates the browser extension may have more technical visibility than either the shopper or the original referrer. The retailer may depend on an affiliate network to audit the transaction, while that network may have incomplete information about what happened inside the consumer’s browser.
When responsibility is distributed across several organizations, no single participant necessarily observes the complete transaction from beginning to end.
That fragmentation creates an opportunity for misconduct, but also for genuine mistakes to persist undetected. It is why a proper response requires more than identifying a convenient villain. It requires examining the mechanisms that make inaccurate attribution possible.
The Question Beneath the Gates Surname
The public identity of Phoebe Gates has made this story unusually attractive to commentators. She is the daughter of one of the most recognizable technology entrepreneurs in history, and she has publicly discussed the difficulty of establishing an independent professional identity while carrying a famous surname.
Her circumstances present an interesting contradiction. On one hand, she has spoken about wanting to build something that belongs to her rather than being defined entirely by her family’s wealth. On the other hand, it is reasonable to recognize that growing up surrounded by extraordinary resources and professional connections may provide opportunities unavailable to most aspiring entrepreneurs.
Those advantages do not prove wrongdoing. Nor do they mean that her work, decisions, or achievements are without value. A person born into privilege can still develop useful technology, take genuine entrepreneurial risks, and make meaningful contributions.
The more important issue is how society evaluates claims of independence and merit. Avoiding direct financial assistance from wealthy parents does not automatically eliminate the effects of one’s background, but family wealth also does not establish that an entrepreneur’s business success is artificial.
Rather than reducing this controversy to a judgment about inherited privilege, it is more productive to ask whether Phia was subjected to the same standards of scrutiny that should apply to any technology company.
Were its revenue streams independently examined? Did investors understand the distinctions between legitimately generated sales and attributed transaction volume? Were engineering concerns elevated to people responsible for compliance? Did the company maintain records that would allow its payment arrangements to be reconstructed and audited?
If those questions are relevant to Phia, they are relevant to every similarly situated startup, regardless of the social position of its founders.
This is also where the relationship between capital and entrepreneurship becomes central. Venture investors place funds into companies because they anticipate future financial returns, often through rapid growth. Founders have powerful incentives to demonstrate that their products can attract users, generate commercial transactions, and eventually become profitable.
Those incentives are not inherently corrupt. Without investment, many valuable technologies would never reach the public. But incentives become dangerous when companies discover that manipulating a measurement is easier than producing the underlying result the measurement is supposed to represent.
A reported sale is not always an earned sale. A claimed referral is not necessarily a genuine referral. Rapid revenue growth is not always evidence of sustainable commercial value.
For that reason, responsible investors should be as interested in the quality and legitimacy of revenue as they are in the speed at which it increases.
The New Frontier: Who Will AI Shopping Agents Actually Represent?
The Phia investigation also arrives at a critical moment in the development of AI-assisted commerce. Shopping tools are becoming more sophisticated, moving beyond coupon searches toward systems that can evaluate products, compare competing offers, assess personal preferences, and eventually make purchases with limited human intervention.
In May 2026, Bloomberg examined the growing importance of agentic commerce, including how payment infrastructure and retail relationships may change when software begins making more purchasing decisions on consumers’ behalf. By September, reporting described Amazon blocking an external AI shopping agent from its retail platform, illustrating how questions of access and control are becoming commercially contentious.
This development raises an issue larger than affiliate cookies. An intelligent assistant may eventually influence not only which merchant receives credit for a purchase, but which product the consumer sees in the first place.
A consumer might ask an AI assistant to find the best television, insurance policy, airline ticket, or financial product. The assistant could evaluate legitimate criteria such as price, quality, availability, reliability, and suitability. But what happens if the companies offering the products can also pay the assistant’s operator for referrals, preferred placement, or completed purchases?
Financial compensation does not automatically make a recommendation misleading. Advertising-supported services and affiliate-funded comparison tools can be useful when the incentives are disclosed and appropriate safeguards are in place. The conflict appears when a system claims to act solely for the consumer while secretly prioritizing revenue for itself or its commercial partners.
The challenge is therefore not simply to build more intelligent software. It is to build systems whose financial incentives and observable behavior are consistent with the responsibilities they claim to undertake.
A truly consumer-centered AI assistant should be able to explain why it recommends a product, identify relevant commercial relationships, distinguish sponsored placement from independent assessment, and provide a meaningful record of the actions it performs with the user’s authorization.
Without that transparency, consumers may gradually transfer more decisions to systems whose economic interests they cannot adequately examine.
What Phia Has Done — and What Remains Unresolved
Following the reporting, Phia stated that it had removed features responsible for incorrect attribution on July 7, was reviewing transactions, had begun issuing reversals to affected brand partners, and was hiring a head of compliance. Those responses indicate that the company publicly acknowledged a substantive problem and committed to corrective action.
They do not, however, independently establish how many transactions were affected, the total commissions involved, whether every affected affiliate can be identified, or whether reimbursement efforts will restore all improperly diverted payments.
The distinction is especially important because reimbursement and accountability address different questions. Returning money may correct a financial injury, but it does not necessarily establish why the injury occurred. Conversely, a serious operational failure does not automatically establish criminal intent.
The source video raises the possibility of criminal penalties, including a theoretical federal wire-fraud exposure of up to 20 years. That should not be mistaken for a prediction about Phoebe Gates or Sophia Kianni. Criminal prosecution would require evidence satisfying the elements of a specific offense, and a conviction would require proof beyond a reasonable doubt. The available reporting reviewed for this investigation did not establish that either founder had been criminally charged or convicted in connection with these allegations.
Accordingly, Red Blood Journal does not characterize either founder as a convicted fraudster or treat disputed internal communications as a final legal determination.
The stronger investigative approach is to examine the unresolved evidence and follow the money.
An effective examination of Phia should seek an independent audit of referral attribution before and after the July changes, a complete accounting of transaction reversals, the company’s internal records of compliance concerns, its representations to investors, and the contractual rules governing the affected affiliate relationships.
It should also examine whether other shopping extensions employ technically similar mechanisms, even if they present them to users differently. A company should not be condemned merely because it operates an affiliate model, but neither should legitimate consumer benefits excuse deceptive attribution practices.
The governing principle is straightforward: payment should be tied to the contribution actually made, and users should understand when an assistant’s business relationships can influence its behavior.
The Red Blood Perspective — Follow the Action, Then Follow the Money
The deepest lesson of this controversy is not that wealthy families are inherently corrupt, that artificial intelligence is inherently dangerous, or that venture capital necessarily produces dishonest businesses. None of those conclusions follows from the available evidence. The more useful observation is that economic systems can reward conduct that ordinary participants cannot easily see, and those hidden incentives deserve scrutiny regardless of who benefits from them.
Phia entered the market with an attractive promise: helping people shop more intelligently. Its underlying product had real utility, and the company’s founders had understandable reasons to pursue entrepreneurship. Yet the reported behavior of its affiliate-tracking features raises a fundamental question about the difference between creating value and claiming value that someone else created.
The reported collapse in daily revenue after disputed monetization mechanisms were disabled strengthens the case for examining the company’s financial model. It does not independently establish criminal intent, but it does suggest that investigators, investors, merchants, and affected affiliates should examine which transactions produced revenue and why those transactions were credited to Phia.
And this brings us to the larger system. Venture capital rewards measurable growth, affiliate networks distribute commissions according to technical identifiers, merchants purchase access to customers, and consumers increasingly rely on automated tools they cannot fully inspect. No single fact proves that this structure is corrupt. But wherever measurements become detached from genuine economic contributions, opportunities arise for value to be redirected without an obvious change in the consumer’s experience.
The same principle applies far beyond shopping. It applies to advertising, online publishing, financial advice, recommendation engines, and eventually autonomous AI agents capable of spending money on our behalf. As these technologies gain more authority, the question of who controls their incentives may become as important as the question of how intelligent they are.
Red Blood Journal therefore asks readers to judge the system through observable actions, traceable financial outcomes, and verifiable records, not surnames, promotional promises, investor reputations, or accusations alone. The objective is not to replace one unexamined narrative with another, but to make the underlying economic relationships visible enough that citizens can judge them independently.
The ultimate test of a technology company is not how convincingly it describes the value it creates. It is whether that value remains real when its transactions, incentives, and financial claims are independently examined.
🌊❤️✨ Ocean of Love and Positivity Perspective — Intelligence Without Integrity Is Not Progress
From the Ocean of Love and Positivity perspective, this controversy presents an opportunity to examine the difference between intelligence and wisdom. Intelligence makes it possible to design software that can compare millions of products, detect patterns, and perform transactions faster than human beings. Wisdom asks whether those capabilities are being used in ways that respect the people whose lives they affect.
A technology that quietly transfers value from one participant to another without justification may be technically sophisticated, but its sophistication does not make the outcome ethical. If the objective of progress is to improve human life, then the measure of progress must include fairness, transparency, responsibility, and the consequences of decisions made by those who control increasingly powerful systems.
This perspective does not require hatred toward Phoebe Gates, Sophia Kianni, their investors, or anyone associated with the company. Accountability is not the same as revenge. If wrongdoing is established, those responsible should face appropriate consequences, affected parties should be made whole where possible, and the failures that permitted the conduct should be corrected so that similar harm becomes less likely.
The same standards should apply to everyone, whether they are born into wealth or poverty, lead a billion-dollar corporation or operate a small independent business. A constructive society does not excuse misconduct because of privilege, nor does it assume misconduct because of privilege.
The promise of artificial intelligence should be measured partly by whether it increases people’s ability to make informed, independent choices. A genuinely beneficial assistant should expand human freedom rather than quietly create new forms of dependency. It should help people understand their options, not exploit the invisibility of its own operations.
Perhaps the most important question raised by this investigation is therefore not whether machines can become smarter than human beings. It is whether human beings will develop the judgment and responsibility necessary to use increasingly intelligent machines without surrendering the principles that make progress worthwhile.
A Final Challenge from Red Blood Journal — Think Beyond What You Are Told
Before accepting any conclusion in this investigation, including Red Blood Journal’s own interpretation, examine the evidence yourself. Question the company’s explanations, question the accusations made against it, and question whether the available records support the conclusions being presented.
Do not mistake a compelling story for a proven fact, or an official denial for a complete answer. A statement can be sincere and still be mistaken. An allegation can be serious and still be unproven. An investigation can uncover genuine wrongdoing while also drawing conclusions that extend beyond its evidence.
Ask what happened, who benefited, who suffered, what records establish those outcomes, and what information remains unavailable. Follow the money, but do not assume that every financial connection proves coordination or corruption. Follow the actions, but recognize that actions can have more than one possible explanation. Judge competing explanations by their consistency with verifiable facts rather than by how closely they resemble your existing beliefs.
Most importantly, apply the same skepticism to Red Blood Journal. Do not allow this publication, any other commentator, an institution, a political movement, or an artificial intelligence system to replace your independent judgment.
The purpose of critical thinking is not to distrust everything. It is to learn how to distinguish what is established, what is probable, what is merely possible, and what remains unknown.
At the beginning, we invited you to question the world. At the end, we invite you to question us.
🩸🌊❤️✨ — Ocean of Love and Positivity
Blood connects us. The ocean unites us. Love guides us. Positivity illuminates our path.
Red Blood Journal — Think Beyond the Script.
Investigation sources and verification notes
The report combines the supplied ColdFusion transcript with independently located reporting and federal court documents. The source transcript is not treated as independent confirmation of its own allegations.
Bloomberg, July 9, 2026 — Original examination of Phia’s alleged sales-attribution conduct.
TechCrunch, August 11, 2026 — Follow-up reporting on internal communications, disputed features, and Phia’s response.
People, August 12, 2026 — Company statement addressing transaction reversals and compliance.
U.S. District Court, June 22, 2026 — Court order allowing the amended Honey affiliate-litigation claims to proceed.
Vogue, September 2026 — Industry analysis of affiliate attribution, monitoring, and corrective measures.
💰
The Invisible Commission in AI Commerce
Oct 16, 2026
The provided investigative reports examine controversies surrounding artificial intelligence shopping extensions, specifically focusing on the startup Phia, co-founded by Phoebe Gates and Sophia Kianni. These tools are scrutinized for allegedly engaging in cookie stuffing and manipulative affiliate attribution, a practice where software quietly hijacks referral links to collect commissions on transactions it did not genuinely generate. Investigations revealed that Phia’s daily revenue plummeted significantly after these disputed features were disabled, highlighting how rapid corporate growth can rely on hidden financial mechanisms rather than true consumer value. Furthermore, this phenomenon parallels similar legal challenges faced by PayPal’s Honey extension, demonstrating a broader systemic vulnerability in electronic commerce and automated assistants. Ultimately, the sources emphasize the urgent need for transparency, independent financial audits, and ethical accountability to ensure that emerging technologies truly serve consumers rather than secretly profiting behind their backs.