The announcer on the platform and the voice that was assembled from parts

These are listed in the order worth acting on, which with station announcements is not the order they are usually presented in.

The announcer was in the building

A station announcer stood or sat where they could see the platforms, and their information came from the same reality the passengers were in. They could say that the train was standing at the signal, that the driver was walking down the platform, or that nobody knew yet. That third statement is the valuable one and is almost impossible to produce automatically.

Honest uncertainty is the hardest message for a system built out of prerecorded certainties.

Automated announcements are assembled from fragments

Systems concatenate recorded words and phrases — station names, times, platform numbers, reasons — from a fixed library. That is why the prosody sounds slightly wrong at the joins and why the vocabulary of excuses is so limited.

For a while, the library is expensive to record and to extend, so the range of things the system can say is fixed for years. The strange stiltedness of station announcements is a direct artefact of how they are built.

The reasons became a joke because the list is short

Standardised delay attribution categories exist for operational and regulatory reporting, and announcements draw from them. That produces a small set of repeated phrases that passengers learn and then mistrust, because the same words cover very different situations. A human announcer could distinguish between a signal failure that would clear in five minutes and one that would not.

Consistency in reporting and usefulness to passengers turned out to be in tension.

Clarity genuinely improved

Recorded announcements are consistently intelligible, correctly pronounced and available at unstaffed stations at every hour. Accessibility improved substantially, particularly for passengers who rely on consistent phrasing or on visual displays generated by the same system. Multilingual announcement is practical in a way live announcing rarely was.

This is a real gain that arguments about lost character tend to skip.

Disruption exposes the gap

When something unusual happens, automated systems either repeat an inadequate phrase or fall silent, and passengers are left with no information at all. Operators respond by putting staff on platforms with megaphones, which is the old system reintroduced as an emergency measure.

That pattern — automation for the normal case, humans for the exception — recurs across transport, retail and support. The exception is where passengers most need judgement, and it is exactly what was removed.

Dates in this period are approximate; the change was gradual and nobody logged it.

Some voices became famous

Announcement voice artists have recorded vocabularies used by millions daily, and in several networks the voice has become locally iconic. Replacing or re-recording such a voice reliably produces public complaint, which is a strange consequence for a technical procurement decision. Synthesised speech is increasingly capable of replacing recorded libraries entirely and raises the same objections again.

People become attached to a voice that never spoke to them specifically, which is worth sitting with.

Everything above, in order of what to do first

1. The announcer was in the building. A station announcer stood or sat where they could see the platforms, and their information came from the same reality the passengers were in.
2. Automated announcements are assembled from fragments. Systems concatenate recorded words and phrases — station names, times, platform numbers, reasons — from a fixed library.
3. The reasons became a joke because the list is short. Standardised delay attribution categories exist for operational and regulatory reporting, and announcements draw from them.
4. Clarity genuinely improved. Recorded announcements are consistently intelligible, correctly pronounced and available at unstaffed stations at every hour.
5. Disruption exposes the gap. When something unusual happens, automated systems either repeat an inadequate phrase or fall silent, and passengers are left with no information at all.
6. Some voices became famous. Announcement voice artists have recorded vocabularies used by millions daily, and in several networks the voice has become locally iconic.

The takeaway

The voice got clearer, arrived everywhere and never once told you the thing you actually wanted to know.

What replaced it works. That is not the same as it being nothing.