Nine months ago, the points community spent a lot of energy worrying about AI travel agents. The headlines said booking sites were about to be replaced overnight, loyalty programs would be sidelined, and anyone holding a stash of transferable points should think hard about burning them. As of May 2026, none of those worst-case scenarios actually played out. Some real things shifted, but the story is more nuanced than the early hype suggested.

This is a retrospective on what happened. We'll cover how OpenAI Operator, Google Gemini's travel features, and the major OTAs' responses reshaped the booking landscape, where points strategy genuinely changed, what hasn't budged, and what to do about it as a points-and-miles person right now. The dust hasn't fully settled. Verify program-by-program before you act.

What Actually Shipped in the AI-Agent Era

Going back to September 2025, the big launches were OpenAI's Operator (a browser-using agent that could book flights, hotels, and rentals on your behalf), Google Gemini's deeper integration with Google Flights and Hotels, and Anthropic's expansion of Claude's tool-use capabilities. Booking.com, Expedia, and Airbnb each shipped their own AI booking flows in the same quarter, mostly to defend territory rather than push the frontier.

What that meant in practice: a user could say "find me a five-night trip to Lisbon under $2,000 with a hotel near Bairro Alto" and the agent would propose an itinerary, fill out the booking forms, and request approval before clicking buy. The tech worked, especially for cash bookings on major OTAs. By the end of 2025, roughly 18% of new Booking.com users in the US had used an AI-assisted booking flow at least once, per the company's Q4 earnings call.

Award bookings are a different story. AI agents largely didn't touch them. Loyalty program interfaces are messy, login flows require two-factor authentication, and award space moves too fast for an agent to handle reliably. Nine months in, no AI agent has shown it can consistently book a complex award itinerary, like a transferable-points-to-Aeroplan partner award, without a human in the loop.

Why OTAs Were Nervous (and Why Some Stayed Nervous)

The fear in late 2025 was that AI agents would commoditize the booking layer. If an agent can compare prices across 10 sites in three seconds, why start your search on Expedia? The middleman gets cut out, ad revenue collapses, and loyalty programs that fund award redemptions through cash bookings lose their funding.

Some of that played out, just slower and at smaller magnitudes than the early commentary suggested. Booking.com's direct traffic share dropped about four percentage points between September 2025 and April 2026, with the gap going partly to agent-driven traffic that bypasses the homepage entirely. Expedia held more ground because of its Vrbo and corporate travel businesses. Airbnb actually grew, mostly because hosts can't be aggregated cleanly by third-party agents.

Legacy metasearch players are still in trouble. Kayak, Skyscanner, and Trivago all reported double-digit declines in unique visitors over the past nine months. Their value proposition (comparing prices across booking sites) is exactly what AI agents do natively. If you used Kayak for award research, you've probably noticed the interface getting worse as the company cuts costs.

Where Points Strategy Actually Changed

Three real shifts happened in points strategy over the past nine months. Two are headwinds and one is a tailwind.

Headwind 1: Dynamic pricing accelerated. Programs like Delta SkyMiles, United MileagePlus, and JetBlue TrueBlue leaned further into demand-based award pricing partly because AI agents made it easier for airlines to model what travelers would actually pay. Delta's AI-driven pricing rollout, which we covered in our Delta AI pricing analysis, accelerated in early 2026. The result: the predictable sweet spots that defined points strategy for years are getting eroded. A flight that cost 25,000 SkyMiles in 2024 might now price at 45,000 or higher on the same date a year later.

Headwind 2: Manufactured spending got harder. AI fraud detection at issuers improved as the same underlying model capabilities got deployed by banks. Several gift card resellers shut down or tightened limits between November 2025 and March 2026. If your strategy depended on heavy gift card cycling, you've already felt this.

Tailwind: Transferable points became more valuable, not less. This is the part most early commentary got wrong. Because dynamic pricing made airline-specific currencies less reliable, points that can move across multiple programs (Chase Ultimate Rewards, Amex Membership Rewards, Capital One, Citi ThankYou) gained relative value. When SkyMiles spikes a route to 80,000, sending points to Virgin Atlantic or Air France instead is what keeps the redemption affordable. The Chase points-boost feature we covered in our Chase points-boost guide became more useful, not less.

How Travel Companies Have Actually Responded

The big OTAs each made distinct choices. Booking.com leaned into AI as a partner, opening up an agent-friendly API and effectively saying: if you must use an agent, use ours, or at least one that pays us a referral fee. Expedia followed the same path more cautiously. Airbnb did the opposite. It tightened its terms of service to block third-party agents from booking on behalf of users without explicit Airbnb authentication, and it launched its own conversational booking flow inside the app.

Airlines were a mixed bag. Delta and United, both of which have strong direct-channel preferences, made it harder for AI agents to log into accounts and search award space. American Airlines went the other way and started exposing more award data through partner APIs, partly to make its program more agent-friendly. The practical result for a points enthusiast: Delta and United awards are now harder to research at scale, while American awards are slightly easier.

Hotel chains have been quieter. Marriott, Hilton, and Hyatt all rolled out AI booking assistants on their own apps, but none of them have moved to block third-party agents in a serious way. World of Hyatt, in particular, has remained the most points-friendly major program through this shift, partly because it doesn't depend on dynamic award pricing and partly because Hyatt's distribution strategy still favors direct booking. If you're building a long-term points strategy in 2026, Hyatt looks more attractive now than it did a year ago, and a card like the Chase Sapphire Reserve still feeds Hyatt cleanly through the 1:1 transfer.

Winners and Losers Nine Months In

Winners. Flexible transferable-points holders. Travelers who learned how to use AI agents for the cash booking pieces (research, comparison, alert monitoring) while keeping award booking manual. Hyatt loyalists. Airbnb hosts. Travel advisors who position themselves as humans for the parts agents can't handle, like complex multi-stop itineraries and award booking.

Losers. Single-program loyalists who locked in heavily with one airline. Manufactured spenders who relied on gift card arbitrage. Kayak-style metasearch. Travel bloggers who built entire businesses around point valuations that assumed static award charts. Anyone who burned points in late 2025 expecting the system to collapse.

Mixed. Big OTAs (Booking, Expedia) lost homepage traffic but kept most booking volume. Airline programs are less valuable as fixed currencies but still useful when redeemed strategically.

Action Plan for Points Enthusiasts

Short-term (next 90 days):

  1. Audit your points balances. Identify any positions in single-program currencies (especially Delta SkyMiles and JetBlue TrueBlue) that you're holding for "someday" trips. If a trip is more than 12 months out, your points are probably losing relative value while you wait.
  2. Diversify into transferable points if you aren't already. The card lineup that gives you the most flexibility in 2026 is built around Chase Ultimate Rewards plus Amex Membership Rewards, with Capital One or Citi as a secondary stack.
  3. Test one AI booking flow for a cash trip you would have booked anyway. Booking.com's AI assistant or Google Gemini's flight tool are reasonable starting points. The goal isn't to replace your award booking workflow, it's to understand what these tools can and can't do.

Long-term (next 12 months):

  1. Treat single-program awards as use-it-or-lose-it. Burn down SkyMiles balances on near-term trips. Park new earning into transferable currencies.
  2. Lean into Hyatt for hotels. The math has gotten better, not worse, since AI agents arrived.
  3. Watch the AA story. If American keeps opening up its award data, it could become the most agent-friendly major airline program, which would change the calculus on AAdvantage versus other US carriers.
  4. Don't panic-redeem. The single biggest mistake people made in late 2025 was burning points at bad rates because they feared the system was collapsing. It didn't collapse, and people who held flexible balances came out ahead.

What's Still Uncertain

Some of this will keep moving. Airline-AI integrations could deepen, and a few more programs may go fully dynamic on awards. Issuers might tighten welcome bonuses if AI-assisted churning gets out of hand. The next round of OTA consolidation could reshape who owns the booking layer. None of these are crises, but they're worth watching. Verify the specifics on any program-change you're betting on, because the half-life of a points-strategy take in 2026 is shorter than it used to be.

The bottom line: AI agents are now part of how trips get planned and booked, but they haven't displaced points strategy. The fundamentals (earn flexible points, redeem strategically, diversify across programs) still work. The mistakes were made by people who acted on the hype, not by people who kept their heads down and held transferable currencies. Keep playing the long game.

This article contains affiliate links. The Points Party may earn a commission if you apply for a card through one of our links, at no extra cost to you.

Some of the links in this article are affiliate links. We may receive a small commission at no extra cost to you if you apply through these links. This helps us keep the site running and continue creating free content.