Comment on Open source project fools AI scrapers with poisoned font

• kbobabob@lemmy.dbzer0.com ⁨2⁩ ⁨weeks⁩ ago

  Why couldn’t a screen reader still work if the words that are on the screen are the correct ones? Does a screen reader not actually read the screen?

  source

  • dfyx@lemmy.helios42.de ⁨2⁩ ⁨weeks⁩ ago

    It (usually) does not, at least not for websites. Reading the screen would require a lot of computing power and be prone to mistakes, especially with unusual fonts and noisy backgrounds.

    Instead, screen readers read a website‘s HTML source code which also gives the developer the chance to annotate how certain elements should be read. Accessibility on the web has been a huge topic for years and serving obfuscated HTML that only works for seeing users goes against everything we have been fighting for.

    source

    • kbobabob@lemmy.dbzer0.com ⁨2⁩ ⁨weeks⁩ ago

      It seems with today’s computer power we could do it, but I understand accessibility needs to work for everyone. I wish I had a better solution.

      source

      • ByteSorcerer@beehaw.org ⁨2⁩ ⁨weeks⁩ ago

        If you can use the page code instead then you have the information of which blocks of text belong together, and you know with 100% accuracy what character the text contains. And you also have additional information such as which elements are interactive, and, if semantic tags are used correctly, which part is the header, main content, or such.

        Besides, scraping can be done with OCR too. Both softwares can use the same tricks, so if screen readers can work using a workaround, then scraping can work with that same workaround too.

        source