The NVDA screen reader reads this text as: "This is my email frame link email us." That is by no means equivalent to actually seeing the email address. I found that HTML entity encoding every single character of the link takes care of any spam problem already and is much more accessible.

Being accessible and behind machine-unreadable are literal opposites. A screen reader is not that different from an ad blocker or web scraper in how it accesses content.

There's a reason that many end-to-end testing experts recommend writing selectors based on accessibility labels instead of CSS classes or IDs, especially if you're using a library like Styled Components.

The email address wouldn't be in the document directly, only in the SVG. Whether the title of the SVG contains "Email us" or the email address wouldn't affect how it works.

If the scrapper is searching the DOM rather than simply downloading the webpages, then the email will found regardless.