E-commerce · Deep dive

Phia

The 'Should I buy this?' button — a Gen-Z AI shopping agent from Phoebe Gates and Sophia Kianni that price-checks any item against 40,000 new and resale sites, monetized on affiliate commissions and dogged by cookie-stuffing and data-harvesting scandals.

emerging

The question that decides it: Does an affiliate-fee shopping agent retain users and margin once Google, Amazon and Chrome ship native price comparison for free — and can a business built to help people spend less monetize price-conscious shoppers without hijacking the very affiliate commissions (cookie-stuffing) and user data (undisclosed HTML capture) that got it suspended and exposed in its first year?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Phia is a free browser extension and iOS app that answers one question — “should I buy this?” — by comparing whatever product a shopper is looking at against real-time prices across a claimed 40,000 retail and resale sites, then surfacing cheaper new or secondhand matches. Co-founded by Phoebe Gates (Bill and Melinda Gates’ daughter) and Sophia Kianni, it launched in April 2025, raised an $8M Kleiner Perkins seed in September 2025 and a $35.5M Series A in January 2026 at a $185.5M valuation, and claims 1.5M users and ~10,000 brand partners in roughly a year. It is also the rare year-one startup to have weathered two damaging scandals — an undisclosed data-harvesting exposé (Nov 2025) and a cookie-stuffing affiliate-fraud finding that got it suspended from impact.com (Jul 2026). The valuation rests on celebrity distribution; the durability rests on whether an affiliate-funded agent can survive native price comparison shipping into every browser for free.

Founding story

The two founders met as Stanford roommates. Gates arrived pre-med, following her sister Jennifer, and graduated in three years with a human-biology degree in June 2024 — but her real pull was fashion: she had trained as a ballerina at Juilliard-adjacent schools, then interned at British Vogue at 19 and worked fashion weeks in New York, Paris and Copenhagen. Kianni came in wanting to be an environmental lawyer, having already founded Climate Cardinals (a ~9,000-volunteer youth climate nonprofit) and become, in 2020, the youngest UN advisor in US history.

The idea came out of a shared, specific frustration: as fashion-obsessed young women who liked secondhand, they found it maddening to know whether a price was good and where the same item sold cheaper across the fragmented resale landscape. Earlier brainstorms went nowhere — they have openly joked about an abandoned “Bluetooth tampon” concept — before landing on price-comparison shopping. They launched Phia in April 2025 and leaned hard on a Gen-Z growth playbook: a podcast (“The Burnouts,” on Alex Cooper’s Unwell network), heavy personal-brand content, and a rolodex of famous friends and family that would later become the cap table. That last asset is the whole story of Phia’s early rise — and the source of most of the skepticism about it.

How it works

The core interaction is a button. On mobile Safari or in the Chrome extension, a shopper viewing a product taps “Should I Buy This?” Phia reads the page, identifies the item, and checks it against its price database — the company cites 40,000+ retail and resale sites and 350M+ products, including ~250M secondhand listings. It returns a verdict on whether the current price is high, typical or fair, and if the item looks overpriced it surfaces cheaper exact matches and similar alternatives, pulling resale inventory from 150+ secondhand platforms including The RealReal, Vestiaire Collective, ThredUp, Poshmark and eBay. It also tracks price drops.

The uncomfortable detail the marketing omits is what happens in the background. To monetize, the extension needs to be the last affiliate “touch” before purchase. Bloomberg and security researchers found Phia’s extension opening a background tab at checkout and, in code added around December 2025, overwriting other affiliates’ referral cookies with its own to claim commission even on sales it did not drive — cookie stuffing. Separately, researchers found the extension transmitting the full HTML of every page visited — including Gmail and banking pages — to Phia’s servers via a function named logCompleteHTMLtoGCS. Both were removed after exposure. They are not bugs; they are the economics of an affiliate agent pushed to their logical, legally fraught edge.

Product and business overview

Phia sells nothing to consumers. The product is three surfaces over one price-and-resale graph. The extension (Chrome desktop) and the iOS app deliver the “Should I buy this?” check, cheaper-alternative surfacing, resale aggregation and price-drop tracking. Phia Intelligence / partnerships is the nascent B2B side: Phia pitches brands on data and placement, claiming partners see ~13% higher conversion, ~30% stronger new-customer acquisition, ~15% higher AOV and >50% lower return rates (all self-reported, Jan 2026). The unifying pitch the company now uses is that Phia is the “alignment layer” or “AI agent for shopping” — a neutral advisor that represents the shopper rather than the seller.

That positioning is in direct tension with how the money is made: a neutral advisor and an affiliate engine want opposite things at the margin — the advisor wants the cheapest option, the affiliate wants the option that pays Phia. The scandals are what that tension looks like when it breaks.

Business model and pricing

Free to the user. Revenue is affiliate commission: when a shopper buys through a Phia link or after a Phia referral is credited, the retailer or marketplace pays Phia a cut (typically single-digit to low-double-digit percent of order value, varying by merchant). There is no subscription, no consumer price point, and no published take-rate disclosure.

This is the same model as Honey, Karma and Capital One Shopping — and it carries the same two structural problems, both of which Phia has now demonstrated in the wild. First, the incentive to claim commissions aggressively (last-click attribution rewards whoever sets the final cookie), which is what produced the cookie-stuffing finding. Second, weak unit economics on the target user: Phia’s entire value proposition is helping people spend less, yet affiliate revenue scales with spending more. The company’s growth claims — “11x revenue growth” since launch, a “nine-figure sales” pace for 2026 — almost certainly describe gross sales or GMV influenced, not Phia’s net revenue, which it has never disclosed. On affiliate rates, nine figures of influenced sales could still be a single-digit-millions revenue business.

Traction over time

The trajectory: launch April 2025; TechCrunch Disrupt stage appearance October 2025 (one month after the seed); $8M seed led by Kleiner Perkins closed September 2025; $35.5M Series A announced 27 January 2026 at $185.5M, oversubscribed and led by Notable Capital with Kleiner Perkins and Khosla returning. The 1.5M-user and 10,000-partner figures are company-stated (Jan 2026) and unaudited. Notably, the November 2025 data-collection exposé did not stop the Series A from closing two months later — a sign the celebrity cap table and momentum narrative outweighed the governance red flag for these investors.

Market analysis

Phia sits at the intersection of two markets. Agentic commerce — AI that researches, compares and transacts on a shopper’s behalf — is the hot one: Bain estimates the US agentic-commerce market could reach $300–500B by 2030 (roughly 15–25% of e-commerce), and Morgan Stanley pegs the impact near $385B by 2030; category revenue estimates for 2025 range wildly from ~$550M to ~$5.7B depending on scope. Online resale is the other: a structurally growing, sustainability-driven segment (ThredUp’s annual report has pegged US secondhand growth at multiples of overall retail), fragmented across dozens of marketplaces — exactly the fragmentation an aggregator can exploit.

The problem is that the TAM is enormous precisely because the largest platforms are racing into it. Agentic commerce is a market defined by Google, Amazon, OpenAI, Perplexity and the payment networks. A free extension is a thin slice of a market whose infrastructure is owned by the companies most able to make the slice unnecessary. The favorable read is that resale coverage and Gen-Z taste are a defensible niche the giants under-serve; the skeptical read is that “cheapest price across sites” is the single feature a browser or search engine is most obviously going to commoditize.

Competitive intel

The direct analog is PayPal Honey — same free-extension, affiliate-funded model, vastly more distribution — and, tellingly, the subject of its own 2024 cookie-stuffing/creator-hijacking scandal (the MegaLag exposé). Honey shows both the ceiling (mass adoption) and the reputational trap Phia walked straight into. Karma and Capital One Shopping deliver the price-check with more retailer depth and, in Capital One’s case, a trusted bank brand and no affiliate-conflict optics; Capital One Shopping’s own research helped surface Phia’s cookie-stuffing. Klarna (which owns PriceRunner) can bundle comparison into a 100M-user checkout app for free.

Above all of them sit the platforms. Google is shipping native AI comparison and agentic checkout into search and Chrome — if the browser compares prices for free, the extension’s core job disappears. Perplexity, ChatGPT and Amazon’s Alexa for Shopping (renamed from Rufus in May 2026) do conversational cross-retailer comparison with the models, traffic and checkout Phia lacks. And the resale marketplaces Phia calls partners — The RealReal, ThredUp, Poshmark, eBay — are also rivals for the transaction, each building its own discovery and each able to restrict a third-party aggregator. Phia’s defensibility claim reduces to neutrality plus resale breadth plus brand — and the cookie-stuffing episode directly undermined the neutrality claim.

History and evolution

• 2023–24 — Gates and Kianni, Stanford roommates, brainstorm ideas (including the abandoned “Bluetooth tampon”) before settling on price-comparison shopping. Company formed while at Stanford.
• Apr 2025 — Phia launches publicly as a free iOS app and browser extension.
• Sep 2025 — $8M seed led by Kleiner Perkins; angels include Sheryl Sandberg, Sara Blakely, Kris Jenner, Michael Rubin, Hailey Bieber.
• Oct 2025 — Founders appear on the TechCrunch Disrupt stage; “The Burnouts” podcast running as a growth engine.
• Nov 2025 — Fortune reports (via ex-Meta engineer Maahir Sharma, replicated by independent researchers) that the extension had been uploading full HTML of every visited page — Gmail, bank statements — to Phia servers via logCompleteHTMLtoGCS. Phia quietly removes the feature without disclosing the breach to users.
• Jan 2026 — $35.5M Series A at $185.5M, led by Notable Capital with Kleiner Perkins and Khosla; 30+ celebrity investors revealed (Sydney Sweeney, Paris Hilton, Khloé Kardashian, Priyanka Chopra Jonas, Vlad Tenev, and more). Claims 1.5M users, ~10,000 partners.
• Jul 2026 — Bloomberg finds the extension overwriting rival affiliates’ cookies at checkout (cookie-stuffing, code added ~Dec 2025); within 48 hours of the 9 July report, impact.com suspends Phia’s account. Phia says it had already fixed the code.

What people say

The case for. Supporters — led by a marquee investor syndicate — argue Phia has done the hard part twice: built a genuinely useful consumer wedge (a one-tap “is this a good price?” answer with real resale coverage) and, harder, distribution, reaching 1.5M users in a year mostly through owned content rather than paid acquisition. Kleiner Perkins framed it as the shopping agent for a generation that already shops secondhand and expects AI to do the legwork; the brand-side metrics (higher conversion, lower returns) suggest real merchant value if they hold up. The App Store rating hovers near 4.8, and the founders are unusually effective media operators. For believers, the celebrity cap table isn’t a gimmick — it’s a permanent, low-cost distribution channel most consumer startups would kill for.

The complaints. They are serious and they are structural, not cosmetic. (1) Affiliate fraud: Bloomberg, researcher Ben Edelman and Capital One Shopping all found the extension cookie-stuffing — claiming commissions on sales it didn’t drive and diverting them from the creators who did — leading to the impact.com suspension. (2) Privacy: the undisclosed capture of full page HTML including Gmail and banking pages, quietly patched without notifying users, per Fortune and independent researchers. (3) Trust in the numbers: Trustpilot and app-review analysts allege the App Store five-star reviews are LLM-generated; JustUseApp’s automated safety score sits near zero. (4) Product accuracy: users report the price matcher surfacing wrong items — different shoes, wrong sizes — undercutting the “never overpay” promise. (5) The privilege critique: a recurring press theme (Bloomberg, multiple Substacks) that Phia is what VC now rewards — proximity, manufactured hype and enough FOMO to suspend diligence — with the $185.5M mark reflecting the founders’ names more than the fundamentals. Even sympathetic commentators note the core conflict: an AI built on affiliate commissions cannot be a neutral advisor, and Phia’s first year is the proof.

Outlook: the open question

This resolves on two things, and both currently point the wrong way. First, whether an affiliate-funded agent can hold users and margin once native price comparison ships free into Chrome, Google search, Amazon and the chat assistants — “find me the cheapest price” is the single feature the platforms are most certain to commoditize, and Phia’s answer (resale breadth + Gen-Z taste + brand) is a thin moat against companies that own the browser, the model and the checkout. Second, whether a business whose entire promise is helping people spend less can monetize price-conscious shoppers at all without the two things that already blew up in its face: hijacking other affiliates’ commissions to inflate revenue, and harvesting user data to power the matching. Strip both out and you have not yet seen whether the clean version has real economics.

For the bull case: Phia converts the 1.5M-user base and celebrity distribution into a durable habit that survives Google/Amazon shipping the same feature free; the brand/partnerships side becomes a real, disclosed revenue line not dependent on last-click affiliate games; and it rebuilds trust after two governance failures in eighteen months — publishing net revenue, retention and a credible privacy posture rather than GMV-flavored vanity metrics. For the bear case: flat or declining active users after the incumbents ship native comparison; continued reliance on affiliate mechanics that networks keep policing (impact.com already suspended it); and a next round marked flat or down once the celebrity halo fades. The $185.5M valuation is a bet on the founders’ distribution; the business underneath still has to prove it is more than a beautifully marketed skin on the most commoditizable feature in commerce.

How a challenger would attack it

Weaponise the trust deficit. Phia’s entire positioning is “the agent that represents the shopper,” and its first eighteen months produced the counter-evidence a rival needs: undisclosed full-page HTML harvesting (Gmail and banking pages included, quietly patched without user notification), cookie-stuffing code that got it suspended from impact.com, alleged LLM-generated App Store reviews, and users reporting wrong-item price matches. A challenger launches as the provably clean agent — open-source extension code, no background tabs at checkout, a published attribution policy, subscription or flat-fee pricing that removes the affiliate conflict entirely — and markets directly against Phia’s scandal record to the same Gen-Z audience, which is precisely the demographic most fluent in the exposés. The second vector is accuracy: the matching engine surfaces different shoes and wrong sizes, so a rival that nails exact-match resale search on a narrower category (designer resale, sneakers) beats the 40,000-site breadth claim with depth. Third, the distribution asset is also a ceiling: Phia’s growth runs through two founders’ celebrity, which does not transfer, scale internationally, or survive a down round narrative — while a challenger can buy performance distribution against a brand whose press coverage now leads with “cookie stuffing.” Phia cannot answer with transparency without disclosing the net revenue and retention numbers it has conspicuously withheld.

Same playbook, new buyer

Sell the “should I buy this?” verdict to whoever isn’t conflicted about the answer. Phia’s structural bind is monetizing thrift through affiliate fees that reward spending. Three buyers escape it. First, the resale marketplaces themselves: The RealReal, Vestiaire and Poshmark need cross-market pricing intelligence to price consignments — a B2B pricing-data product built on the same 250M-listing graph monetizes the asset without touching a consumer cookie, and Phia’s consumer-facing neutrality claims make it awkward for the incumbent to sell sell-side data. Second, banks and fintechs: Capital One Shopping already shows the model — a price-check agent as a retention feature inside a card or budgeting app, paid for by interchange and loyalty economics rather than last-click commissions, aimed at the exact price-conscious user Phia can’t profitably serve. Third, geography: European resale (Vinted’s home turf) has deeper secondhand penetration, stricter privacy law that rewards a clean-by-design architecture, and no incumbent “should I buy this?” layer. Phia is unlikely to follow any of these: its cap table and podcast flywheel are wired to US consumer virality, a B2B pivot would concede the consumer-agent story its $185.5M mark was priced on, and its data-handling record is disqualifying in privacy-regulated channels — bank partners and GDPR regulators read Fortune too.

Sources and further reading

• TechCrunch — Phoebe Gates and Sophia Kianni’s Phia raises $35M to ‘make shopping fun again’ (27 Jan 2026)
• Forbes — How Phoebe Gates and Sophia Kianni raised $35 million in hopes of building an AI shopping unicorn (30 Jan 2026)
• PR Newswire — Phia raises $8M seed round led by Kleiner Perkins (Sep 2025)
• TechCrunch — Phia accused of ‘cookie stuffing,’ taking affiliate credit on purchases it didn’t earn (10 Jul 2026)
• Business of Fashion — The Phia scandal and ‘cookie stuffing,’ explained (Jul 2026)
• Fortune — Phia, the buzzy AI shopping tool, was pulling far more user data than disclosed, security researchers say (15 Nov 2025)
• Stanford Daily — Q&A: Sophia Kianni and Phoebe Gates on Stanford and co-founding Phia (23 Sep 2025)
• Kleiner Perkins — Phia: the AI shopping agent for the next generation (2025)
• Digital Commerce 360 — Phia enters price-comparison space for ecommerce and fashion (28 Apr 2025)
• Bain & Company — 2030 forecast: how agentic AI will reshape US retail (2025)