Gemini seeks to be minimal, deliberately excluding markup for bold, italics, and images. A concerning trend in Geminispace is to use Unicode "look-alikes" to emulate these features, generally by abusing mathematical symbols or other special characters. While this defeats the Gemini "minimalism" point, my concern is not simplicity but accessibility.
Consider loading gmisub, a list of aggregated feeds that includes the website Techrights.
Techrights "styles" itself with some... unusual Unicode.
πΏπππππππππ
To a sighted person, this is inherently out-of-place on Gemini. But to a screen reader user, how does this single word title sound? To espeak, the word "Techrights" written in this funny way will be pronounced as:
> Letter 1d57f Letter 1d58a Letter 1d588 Letter 1d58d Letter 1d597 Letter 1d58e...
Yikes! After reading this for about 8 seconds at the default speed, espeak quits, having finished the phrase "Techri". Ostensibly this should be followed up with the rest of the word:
Letter 1d58c Letter 1d58d Letter 1d599 Letter 1d598
So trying to read Gemini not only wastes a screenreader user's time tremendously but also completely obscures the meaning. What is "letter 1d57f" supposed to be? Apparently, "Mathematical Bold Fraktur Capital T". I'm a mathematics student, and I don't believe I have ever typed this character. What's it doing in your Gemini document?
Of course, the user isn't much better off if their screen reader read
Mathematical Bold Fraktur Capital T Mathematics Bold Fraktur Small E Mathematics Bold Fraktur Small C Mathematics Bold Fraktur Small H Mathematics Bold Fraktur Small R Mathematics Bold Fraktur Small I Mathematics Bold Fraktur Small G Mathematics Bold Fraktur Small H Mathematics Bold Fraktur Small T Mathematics Bold Fraktur Small S
If the user is paying tremendous attention they might be able to parse out the letters and reconstruct the word... after having to listen to a single word in Gemini for over 20 seconds at the default speed. You've wasted your users' time, confused them, _and_ most likely they have no idea what site it is you want them to click on. Why?
By the way, it's not only screen reader users who have trouble with this sort of text. Anyone with stimulation issues, overwhelmed with the web, can be thrown off guard by unexpected emphasized text. Gemini specifically provides a safe haven for people with disabilities affecting content media, including blindness but also (for example) the autism spectrum. Gemini is designed to put the user in control of the presentation of content, instead of deferring to the publisher's wishes. This principle is powerful, and attempts to subvert it via abuses of Unicode is deeply unkind.
There is a subtle point about a Gemini limitation here. Abusing Unicode for stylization became common as a hack to allow bold and italics text on plain text-only social media. We see the same trend of Gemini, which also bars formatting. Perhaps if Gemini supported semantic bold and italics via Markdown-like syntax, Geminauts wouldn't feel inclined to reach for the dirty tricks to get there instead. Paradoxically, this could return control to the user, by allowing bold and italics to be softened or removed altogether when presenting, although bad actors might still spoil it but bolding entire web pages and such. Alternatively, the fact that Gemini allows arbitrary Unicode at all is at best an oversight. Either way, this needs attention.
To Gemini authors: Please, I beg you not to abuse Unicode like this. Sprinkling in a single "heart symbol" isn't a crisis, but chaining special characters together and abusing characters for their aesthetics rather than semantics does present a meaningful accessibility barrier. Let's keep Geminispace kind β₯
To Gemini developers: How can we fix this? Could a specification change mitigate this? What about client changes? Let's talk.
Techrights has corrected the issue here. Thank you to Roy Schestowitz. The considerations still apply to other capsules, of course.
Emojis and Accessibility: The Dos and Donβts of Including Emojis in Texts and Emails
Can screen readers interpret Unicode styles fonts such as bold and italics?