How we picked our sources, and why a third of the candidates failed

When you build a news aggregator, you assume the hard part is writing the code. In practice, the hard part was finding sources that actually work.

We started with 36 candidates: the travel desks of major newspapers, trade publications covering aviation and hotels, cruise news, and blogs about points and miles. Twenty-five made it into the finished site. Here is what happened to the others — and why it matters to a reader.

Dead feeds at living publications

The most common problem by far: the publication is alive, its RSS feed is not.

National Geographic, Afar and Rough Guides all return 404 at their feed addresses. Travel Market Report does the same. Lonely Planet answers as though nothing were wrong, then delivers a megabyte and a half of ordinary web page: the news section in its old form no longer exists.

Travel + Leisure responds with status 402 — “payment required”. For an RSS feed that is an exotic answer.

Feeds closed to robots

PhocusWire, TravelPulse and Travel Weekly all answer with 403: access denied. These publications plainly do not want their feeds read automatically.

That is their call, and there is nothing to argue about. Getting around such a block is technically trivial — you pretend to be an ordinary browser. We do not do that: if a publisher closes the door, we do not climb in through the window. Three sources simply did not make the list.

A feed stuck in last year

Aviation Week was a case of its own. The feed opens, returns no errors and looks perfectly healthy. But its newest entry was dated March 2025 — roughly eighteen months old at the time of testing.

A source like that is worse than a missing one: it looks functional while quietly seeding the feed with stale material. We replaced it with FlightGlobal and AeroTime.

German news in an English feed

Business Traveller produced a surprise of a different kind. The feed is live, it updates, the material is relevant — and roughly every fifth item is in German. The publication mixes its German edition into the same stream, and the article addresses give nothing away: the only way to tell them apart is the text itself.

So we wrote a language filter that compares common function words. On testing it removed four German items and left all sixty-two English ones untouched.

Hotels without photographs

The last discovery concerns images. Hospitality Net, Hotel Dive and eTurboNews put no pictures in their feeds at all. Normally you would then take the image from the article page, where it sits in the page metadata.

But these publications answer robots with 403 on article pages even though their feeds are open. The message is clear enough: read the feed, stay off the articles. So cards in the Hotels section appear with a coloured tile carrying the source's initials instead of a photograph. That is a deliberate choice, not an oversight.

Where that leaves us

Twenty-five feeds, six sections, around 380 items in the feed at any moment. Refreshed every ten minutes, duplicates collapsed, anything older than a month dropped.

The state of every source is visible at the bottom of the front page: how many items it delivered, and which error it failed with if it failed. We think a reader is entitled to know what the feed they are reading is made of.