Perhaps to the annoyance of many vintage and expert computer keyboard enthusiasts is the term "mechanical keyboard". Indeed, the many demerits have been widely discussed before in debates and high-profile YouTuber videos (such as Chyrosran22's take on the subject). Yet, the term persists, and the mere doubt of a keyboard's 'mechanicalness' is a vessel for people to form unobjective conclusions without even exploring or trying it. I don't like this, or how it's effectively become a channel for keyboard bias and even shaming. So, whether you like it or not, here's my take on the subject. This is also based on a comment I made to a post in February 2021. It's a decent TLDR of what's said here.
In my eyes, the largest problem with the term is that people assume it's some sort of universally agreed-upon technical definition that's clear-cut. It's not. The fact that people have argued over the definition of it for almost a decade is evidence enough of this, but let's tackle some of the common 'definitions' people pick and choose to assign to "mechanical" at will. If you've already watched chyros' video on it, some of these will be familiar to you since I did take inspiration from his video.
Firstly, here are the common definitions that comment specifically on the type of technology used rather than the resultant characteristics of the switch.
The most basic and common meaning assigned the term. Obviously, it's nice to have a convenient distinction between the two largest groupings of keyboards (ignoring the fact you could use "non-rubber dome keyboard" for distinction...) but this one simply doesn't make sense. A literal mechanical definition (ie, something to do with machines) does not help, however, a linear mechanical motion definition stipulates that rubber domes exhibit the necessary behaviour. This concludes that the term would only exclude solid-state devices.
Also, enforcing this definition would cause the Topre and other dome-with-slider switches gang to come after you.
Something that tries to include rather than exclude things. Problem is, it would include too much for it to be useful. Well, by definition, rubber domes are a form of spring! The things excluded would be non-contact switches and solid-state. In a number of cases, non-contact switches can provide unsurpassable reliability and smoothness and thus rendering 'mechanical' an inferior category of keyboards. Especially for those wanting the best possible linear switches.
A very specific requirement that effectively limits to just two families - Alps/Alps-clone and Cherry/Cherry-clone switches. How does the fact that a given switch is discrete or not affect how it fundamentally works? Discrete rubber dome switches are also available - Alps integrated dome comes to mind. Speaking theoretically, you could also enclose a membrane-driven rubber dome into its own housing, which wouldn't change how it fundamentally works but it would now be 'mechanical' under this definition.
Also, enforcing this definition will cause the entire IBM gang to come after you. Both capacitive and membrane flavours of buckling springs are tied down with a communal sensing mechanism that renders each barrel/spring combo non-discrete. This is definition is bad enough to even exclude beam springs, whose mechanisms are otherwise separate entities except for the communal capacitive PCB needed for the system to work.
Another basic and common one. The problem is, I believe it's too limiting since membranes are just the sensing mechanism and thus only one part of the story. This definition ignores the entire mechanism that could yet be insanely complex and/or good. And by literal definitions, membranes should be mechanical. Especially in terms of motion since membrane contacts bridging together is more of a motion than stationary PCB terminals or capacitive pads. At the very least, I would argue "membrane" and "mechanical" shouldn't be mutually exclusive.
Also, enforcing this definition will cause the Model M gang to come after use. Model Ms are often cited as the most well-known mechanical keyboard with a switch revered by many yet it's in fact a membrane-driven keyboard.
And now secondly, here are the common definitions that are formed by categorising a switch based on a property or characteristic that a large number of people expect them to have.
The act of a switch registering before it hits the bottom of possible key travel. This is cited as one of the biggest benefits of supposedly 'mechanical' keyboards by gamers and typists alike. But, it's not suitable for categorising a type of switch. Metal spring-based switches can be and have been made to actuate full-way, and rubber dome-based switches can be and have been made to actuate part-way.
In some cases, you could modify an existing switch to convert them from one behaviour to another, and you could do this without fundamentally changing the mechanism or how it senses actuation.
The ability for all keystrokes to be registered without any blocking or phantom key generation. This is cited by gamers are a must from their keyboards, although most non-gaming typists are not concerned about this. Anyway, you can approach this from two perspectives - technical and artificial limitations. A technical limitation usually occurs from the circuit design being simplistic, lacking diodes or is a membrane (which is basically 'lacking diodes' again but in a different form). An artificial limitation could be an incomplete USB HID stack implementation that several modern gaming keyboards suffer from and are thus limited to 6KRO over USB since their firmware writers only implemented the HID boot protocol.
To enforce NKRO as a factor, you have to make some technical prerequisites as a part of the definition too. Essentially, the inclusion of diodes (a solid-state component) or a capacitive-sensing assembly would become a prerequisite for a keyboard to be 'mechanical'. Diodes speak nothing of the quality or function of the mechanism itself, just what the sensing circuit's capabilities are. Whilst impractical, it's technically possible to insert diodes into a membrane, which opens a possibility for any vanilla rubber dome keyboard to become mechanical with modification.
There are three ways to overcome the boot protocol-enforced limitation on KRO - ask the developers to implement a full HID stack, use a custom controller, or enumerate the keyboard as several devices to the OS so there are numerous vessels to send the limited boot protocol packets with to the point the keyboard functionally acts NKRO. In the former and latter cases, the difference between 'mechanical' and 'non-mechanical' is software?! In the middle case, it's a combined hardware/software one that opens a slippery slope of categorising keyboards as 'mechanical' by a combination of microcontrollers and firmware. Going back to the latter case, saying a keyboard can be made 'mechanical' by what is frankly a countermeasure for a given problem is a pandora's box.
TLDR, "mechanical keyboard" as a concept is the Wild West of terms. However, people are of course unlikely to drop its usage since it's considered to be a convenient comparator to categorise keyboards. As such, I propose five 'recourses' to this problem: