American Set an End Date. The Award Space Went First.

I had this issue written on Friday. It ran five sweeps instead of four because I held it one more day, and that extra day killed three of the conclusions I was about to publish and turned a fourth into the story. Here is what eight days of the same search actually found, including one thing I did not expect to watch happen in real time.

What I did

American and United put twenty new international routes for 2027 on sale in late August, United on the twenty fifth and American on the thirty first. I searched all of them, plus American's added fourth daily flight to London, from each route's first operating day out to July 29, 2027, where the booking window currently ends. Every result is resolved down to the airline actually operating the nonstop segment, so a connection through somewhere else does not count.

I ran that search on August 31, September 1, September 3, September 4, and September 8. Same script, same window, same filter, five times over eight days.

One definition matters before any number. A day below means a calendar date on which at least one qualifying seat showed up in that cabin. It is not a seat count and it is not a record count. Award searches return the same date many times over, and reporting those as availability would inflate everything by a factor of three or four.

The cabin that emptied while I was watching

American's fourth daily New York to London flight is the one mature route in the set. It is an added frequency on a route American has flown for decades, and it was the only route in my whole sweep showing first class award space.

On September 2, between my second sweep and my third, American published end dates for Flagship First. It stops being sold on the Boeing 777-300ER for travel from November 19, 2026, and on transcontinental flights for travel from March 28, 2027. The cabin is being replaced with a larger business class during a retrofit of all twenty of those aircraft, and the first reconfigured one entered service on September 2.

Every single first class date my search had been returning falls between March 31 and May 27 of 2027. All of them sit after that November cutoff. That inventory was for flights on which the cabin will not be sold, and over the six days following the announcement it went to zero.

I want to be careful here. I can show you the sequence, and I can show you that every affected date sits past the published cutoff. I cannot show you a decision inside American, and some of those dates may simply have been booked. What makes the timing hard to dismiss is the second row of that table. Business class on the identical flights, in the identical search, rose across the same eight days and never left a saver price of 57,500. One cabin left and the other did not.

Three things I had wrong on Friday

Now the uncomfortable part, which is also why this issue is worth more than the one I nearly sent.

I was going to tell you the front cabin had frozen. Across American's seven genuinely new routes, business class award days went zero, then two, then two, then two. The same two dates, Philadelphia to Porto on April 5 and July 12, at 146,000 and 143,500 miles, unmoved through three consecutive sweeps. That is a tidy finding and I had a whole section built on it. On September 8 both dates were gone. The series is zero, two, two, two, zero. Nothing froze. A pair of expensive dates appeared for a week and then left.

I was going to tell you economy had plateaued. Economy days on those seven routes ran 208, 348, 415, and then 425, which reads like a curve flattening out. The fifth sweep returned 648. Philadelphia to Porto went from 42 days to 122, Vienna from 26 to 85, and Reykjavik from 26 to 64. American loaded more award space over that long weekend than it had in the whole first week the routes were on sale.

I was going to tell you United was draining. Its new cities ran 102, 101, 95, 95, which looked like inventory being bought up. The fifth sweep returned 112, and resolving those numbers down to United's own program rather than every program that can sell a United seat made the story disappear. MileagePlus by itself ran 83, 83, 83, 87, and 88. Flat for eight days. The swing I was ready to call a trend was partner programs moving in and out of the data on top of a base that never moved.

What survived, and got stronger

One claim held up, and five sweeps make it firmer than four did. American is mostly not repricing as it loads. Taking AAdvantage's own cheapest economy price on each of the seven new routes, and starting from September 1, which is the first sweep on which all seven were visible, five of the seven have not moved by a single mile since. Tokyo has sat at 35,000, Vienna at 30,500, Reykjavik at 23,500, Porto at 21,000, and Nice at 29,000.

Two moved, and both moved down. Amsterdam went from 21,500 to 21,000 on September 3, which is small enough to be rounding in whatever American uses to set these. Charlotte to Barcelona went from 40,500 to 34,000 on the last sweep, which is not small.

So American released a great deal more inventory and, on most routes, left the price alone. If you looked at Vienna on September 1 and decided to wait, waiting cost you nothing and gained you fifty nine more dates.

The trap that nearly caught me twice

When I first pulled the fifth sweep, two more routes looked like price cuts. Vienna appeared to drop from 30,500 to 27,500, and Reykjavik from 23,500 to 22,500.

Neither of those is a price cut. Both are Alaska Mileage Plan showing up for the first time on routes where AAdvantage had been the only program in the data. American's own price on both routes never moved. I had been reading the cheapest number across all programs, and the cheapest number across all programs will tell you a price fell when what actually happened is that a different seller arrived.

This is the same mistake in a new costume. An aggregated minimum is not a price history. It is a mix of sellers, and the mix changes underneath it.

The useful thing, which arrived a week late

For the first week these routes were on sale, most of them showed only AAdvantage. On the fifth sweep, Alaska and Qantas appeared on Vienna and Reykjavik. That matters because partner programs price American metal on their own charts, and they do not always price it higher.

Alaska wins on three of the six and loses on three, which is exactly the point. The lesson is not that Alaska is cheaper. It is that on a brand new American route the partner charts arrive roughly a week behind the airline's own, and once they do you have more than one price to compare.

The control, which is why any of this is publishable

A number that moves oddly is also what a broken search looks like, so every sweep runs the same filter against five American routes the airline has flown for years. Those returned 169 days of business award space on September 1, then 178, then 178 again, then 168 on September 8, with the same route shape and the same 57,500 floor every time.

The search sees American business class where American business class exists. That is what lets me tell you the Porto pair really left and the London first class cabin really emptied, rather than telling you my tooling had a bad Tuesday.

What to do with this

On a route that just went on sale, one look is worth almost nothing. Four consecutive looks over five days had me ready to publish three findings that an eighth day reversed.

Check the partner charts about a week after a route opens, not on day one. On day one they are usually not there yet.

When a price appears to fall, ask who is selling it before you conclude anything about the airline. A minimum across programs is a mix, not a trend.

And when an airline announces the end of a cabin, treat the award chart as the first thing to go, not the last. The sale cutoff is in November. The award space was gone in six days.

Limits, stated plainly

This is a third party award search, so everything above is what a search surfaced rather than what exists. Eight days is still a short series, and two of the five points are eighteen hours apart, so the direction is the claim and not the slope. The link between the September 2 announcement and the first class availability going to zero is a documented sequence, not a proven cause. I have said nothing about United business class on purpose: a probe of twelve established United long-haul routes returned zero United nonstop business records while returning 1,671 for other airlines on the identical city pairs, so that gap is in the data rather than in United's cabins. Premium economy is excluded throughout because one program prices it above 500,000 miles on two routes, which is dynamic noise rather than availability.

Four looks told me a story. The fifth one took it back.