Why do printers work in CMYK when the image starts as RGB?

Because a screen and a page produce colour by opposite mechanisms, and the ink set has to match the mechanism rather than the file.

A screen emits light. It starts black — nothing emitting — and adds light. Red plus green plus blue at full gives white. That is additive, and red, green and blue are the right emitters.

A page reflects light. It starts white — the paper reflecting everything — and each ink removes part of the spectrum. That is subtractive, and the useful inks are the ones that each subtract roughly one of the three channels the eye uses.

Cyan absorbs red, magenta absorbs green, yellow absorbs blue. So each ink is the complement of one screen primary, and combinations of them can subtract any amount of any channel.

If you printed with red, green and blue inks instead, each would absorb two channels rather than one. Overlapping two of them would leave almost nothing reflected, so you could produce dark colours and very little else. The gamut would be tiny.

On the frustration about colours shifting, which is the real question underneath: the conversion is not the only culprit, and understanding what is lost helps.

The two gamuts overlap but neither contains the other. A screen can show saturated blues and greens that no ink set can reach. Print can produce some yellows and warm tones that a screen struggles with. So the conversion is not simply a reduction — it is a mapping between two differently shaped spaces.

What actually helps in practice:

• Calibrate the screen, so what you see is a known quantity rather than whatever the panel does.
• Soft proof against the specific paper and printer profile before printing, so you see the shifts on screen first.
• Expect the saturated blues to move. That is the single most common surprise, and no amount of adjustment recovers a colour the inks cannot make.

The file staying in RGB until the last moment is correct, incidentally. Converting early throws away information that the printer's own conversion could have used.

One more reason the four-ink model persists that has nothing to do with colour theory: it is what the printing industry standardised on, decades ago, for offset presses.

Four plates, four ink stations, four separations. Everything downstream — proofing, profiles, file formats, the way designers work — grew around that. Even where a printer physically has more inks, the model people design in is usually still the four-colour one plus additions.

Standards outlive their technical justification, and in this case the technical justification also happens to still be sound.

The K is worth explaining separately since it is the part that looks like an inconsistency.

In theory, full cyan plus magenta plus yellow absorbs everything and gives black. In practice it gives a muddy dark brown, because real inks are not perfect absorbers.

So a dedicated black ink is added, and it does four jobs: proper blacks, cleaner neutral greys, sharper black text, and a lot less ink on the page — laying down three saturated inks everywhere something is dark soaks the paper and costs three times as much.

That is a good illustration of the general pattern: the theory says three inks suffice, and the materials say otherwise. It is also why many photo printers use more than four inks — extra light cyan, light magenta, greys and sometimes orange or green — each one added to fix a specific place where the four-ink model falls short.