But that is the point. CSS arrived, and people started saying you "just need to distinguish beteen the semantics of" a table "and the layout effect". Back then, the chant of the day was, "blood for the blood god, tabular data for the <table> god, layout for the CSS god".
And then it spiraled into general "separation of content and presentation" nonsense that, like in any good cult, people believed with their whole hearts, preached to make new converts, then did the opposite because reality demanded it, and couldn't see the problem. CSS Zen Garden notwithstanding, we ended up with div soups (later, semantic div-equivalent soups) inseparable from complex CSS, because form is function; your content is always designed with a particular form in mind.
Layout tables were never a serious problem for screen readers, this has always been overblown. Screen readers adopted working heuristics to distinguish between layout and data tables early on, because tables were used for layout almost from the beginning (as soon as it was possible to remove the visible borders).
See for example WebAIM https://webaim.org/techniques/tables/: “It is sometimes suggested, even by some accessibility advocates, that layout tables are bad for accessibility. In reality, layout tables do not pose inherent accessibility issues.”
Google never had a problem working with content that was blended with it's visual presentation. They literally doesn't care. They crawl anything even pdf and other non-html content.
The semantic web was born from a belief that it would enable a peer to peer web that never materialized. Accessibility was better solved by aria annotations a long time ago and the semantic web was never going to properly solve it.
FWIW, it's hard for a peer to peer web to materialize, where the web is almost entirely built and funded by entities who very much don't want a peer to peer web, and in fact actively fight it.
> How could Google even work if content could not be seperated from its visual presention?
Approximately. Like it did and does.
The simplest way is to just strip the markup (or, for purposes of indexing, just ignore any token that resembles a HTML tag). That'll easily get you 90% there. You'll lose important information that's tied to form, but that's a consequence of operating in plaintext land. If doing this to build an index of the web, you'll have problem with scoundrels stuffing irrelevant text made invisible to humans by markup/CSS/JS, but that's exactly the problem Google used to have for a long time, and the essence of what SEO is :).
And then it spiraled into general "separation of content and presentation" nonsense that, like in any good cult, people believed with their whole hearts, preached to make new converts, then did the opposite because reality demanded it, and couldn't see the problem. CSS Zen Garden notwithstanding, we ended up with div soups (later, semantic div-equivalent soups) inseparable from complex CSS, because form is function; your content is always designed with a particular form in mind.