Analyst Sells Airbnb Stock After Testing Meta’s AI App Muse

In Silicon Valley and Wall Street, discussions about how AI is transforming the business world never cease. But when theory becomes reality, the first to feel the chill may be the internet platform giants we know best.

Recently, on the popular finance podcast "The Synopsis," host Drew engaged in an in-depth conversation with Mostly Borrowed Ideas (hereinafter referred to as MBI), a seasoned independent stock analyst with over 150,000 followers.

MBI said that about 10 days after Meta released Muse, he downloaded the app and began testing it. This experience ultimately led him to make an investment decision: selling his heavily weighted shares in Airbnb and increasing his position in Meta.

In this conversation, Drew also delved with MBI into how Meta’s AI agent product, Muse, could disrupt the business models of internet aggregation platforms such as Airbnb, Booking, Uber, DoorDash, and Amazon. A new commercial era—“Proactive Commerce”—may be just beginning.

Left: Drew, Right: Mostly Borrowed Ideas

As a heavy Airbnb user (having spent over $7,000 this year) and a major investor, MBI initially dismissed, like many others, the idea that AI could disrupt online travel agencies (OTAs). After all, booking a stay requires viewing photos, reading reviews, and even communicating with hosts—traditional text-based AI, such as early versions of ChatGPT, simply couldn’t meet these needs.

But Muse changed his perspective.

MBI shared his experience: “I asked Muse to recommend five accommodations within a two-hour drive from my home, based on my past stays on Airbnb. Amazingly, Muse automatically reviewed my history and even analyzed travel short videos (Reels) I’d saved on Instagram to accurately understand my preferences. It acted like a real person, browsing and filtering in the background, then presented me with options rich in images and detailed reasoning.”

When MBI spotted a cottage and asked if he could book it directly, Muse told him within a minute: Yes—and by booking directly, bypassing Airbnb, he could save 60%. Muse had already linked a credit card; all MBI needed to do was click once to confirm and complete the payment.

"At that moment, I realized Airbnb might be on the wrong side of future trends," MBI admitted.

Host Drew raised a skeptical defense from a traditional business perspective: “Airbnb’s biggest moat is ‘trust’ and ‘exclusive supply.’ As a consumer, I wouldn’t feel comfortable transacting directly with strangers on the internet—I need Airbnb as an intermediary to guarantee refunds and after-sales service. At the same time, many hosts don’t have their own independent websites.”

MBI countered this by stating that AI agents can fully establish "trust" assessments by scraping real reviews from across the web. More importantly, in the AI era, a host need only tell an AI, “Sync this listing to Booking and VRBO and manage my calendar,” and the barrier of “exclusive inventory” will instantly dissolve. Hosts have long been frustrated by the dominance of single platforms, but previously, the friction costs of managing multiple platforms were too high—AI eliminates this friction.

After sharing their experience selling on Airbnb, the two then turned their attention to another major online travel platform: Booking.

The curse of direct traffic: Airbnb currently derives up to 90% of its traffic from “organic/direct” sources, which is also why its profit margins are high. However, in the era of AI agents, consumers may no longer open the Airbnb app directly, but instead rely on AI to coordinate their travel plans. MBI predicts that within the next 5 to 10 years, Airbnb’s direct traffic will decline significantly.

Shift in customer acquisition costs: Currently, Booking needs to pay Google for one-third of its traffic, with annual marketing expenses exceeding $5 billion. For Booking, shifting the traffic source from Google to AI (such as Muse or ChatGPT) may simply mean paying the “toll” elsewhere—and could even reduce customer acquisition costs (CAC).

Valuation protection: Currently, Booking’s valuation is only at 10–11 times EBITDA, as the market essentially treats it as a “traditional pipeline”; in contrast, Airbnb is valued at 30 times. If both eventually become underlying infrastructure for AI, Airbnb will face a far more severe valuation reset.

Host Drew added another defensive advantage of Booking—the loyalty program: “More than half of Booking’s room nights come from its loyalty members. The loyalty program offers points and free upgrades, encouraging consumers to stick with Booking even when using AI to compare prices.”

But both agreed on one point: AI will not make these OTAs "better." At best, they will barely maintain the status quo; more likely, their long-term compounded profit growth rates will be compressed by AI.

If the tourism industry, characterized by low frequency and high average spending, is hit first, what about e-commerce and local life services, which have high frequency?

Take Amazon as an example. MBI points out that Amazon’s greatest crisis is its $76 billion advertising business. “In the past, when we opened Amazon, we would often buy products ranked in the top three search results without thinking—that’s the logic behind how bid-based advertising makes money. But AI agents won’t click on ads. If you instruct your AI to ‘prioritize quality, then price among equal-quality options, and finally consider shipping speed,’ the AI will rationally seek the optimal solution across the entire web.”

For example, MBI saw a T-shirt in a Meta ad and asked Muse to evaluate whether he should buy it. Instead of favoring the advertised product, Muse advised him that the brand offered poor value for money and recommended a better alternative.

Host Drew agreed, offering Amazon’s path forward: “Amazon’s current app experience is actually poor—only the ‘search-buy’ function works smoothly. They must immediately deeply integrate native AI capabilities into their app to protect the ‘user interface’ (UI). As long as users still习惯 opening Amazon first when they have shopping needs, and Amazon’s AI is sufficiently useful, combined with its powerful logistics fulfillment network, it can hold onto its core base.”

Compared to OTA and e-commerce, Drew and MBI both believe that DoorDash (food delivery) and Uber (ride-hailing) will experience delayed impacts.

MBI analyzed: "Both fall into the category of high-frequency, low-average-order-value impulse purchases. When ordering food delivery, users often aren’t sure what they want to eat and need to 'browse.' More importantly, food delivery is a three-sided network (platform, merchants, delivery drivers); AI can replace the platform in placing orders with restaurants, but it cannot replace drivers in delivering food to customers’ doors. Currently, DoorDash earns only 50 cents in profit per order, indicating they are not extracting excessive profits—this, in fact, becomes a moat."

Host Drew believes that ride-hailing (Uber) carries greater risk than food delivery: “If there’s an error in food delivery, users need customer service for refunds, and the platform’s value in after-sales service is significant. But with ride-hailing, as autonomous vehicles (AVs) like Waymo and Tesla enter the market, combined with AI agents instantly comparing prices across all platforms (Uber, Lyft, Waymo), user loyalty to Uber will be nearly zero, and price competition in the ride-hailing market will be extremely fierce.”

At the end of the show, the two experts introduced a new concept that will define the future of commerce: Proactive Commerce.

Traditional commerce falls into two categories: “intent-driven” (such as searching for and buying books on Amazon), and “discovery-driven” (such as seeing an ad on Instagram and developing a desire to purchase).

AI agents will introduce a third model. MBI paints a chilling scenario: “Airbnb has now launched a ‘book now, pay later’ feature (book in advance, charge seven days before check-in). In the future, AI agents like Muse will scan your email, notice that you’ve booked a hotel but haven’t paid yet, and proactively intervene: ‘Hey, I just noticed your hotel booking—if you book directly through the hotel’s website right now, you’ll save 20%. Should I help you cancel your Airbnb reservation and rebook?’”

Without requiring user input, the AI tirelessly works in the background 24/7 to find better solutions and proactively suggests replacements.

In this new era, the business model of platforms that rely solely on information asymmetry and traffic monopolies to charge high “tolls” is being fundamentally uprooted. For investors, it’s also time to reevaluate your “platform-type” stocks: in a future where AI possesses autonomous agency, will they be indispensable infrastructure—or costly intermediaries destined to be bypassed?