> I have never, ever – not even once – seen a graphical abstract that usefully “summarized the contents of an article”.

The author has never worked in chemistry. They are absolutely ubiquitous in scholarly journals for chemistry and I'd say that most people trained in chemistry often only read the graphical abstract when initially deciding whether the paper is worth a deeper read. Perhaps chemistry is a naturally visual field, given that we rely on visualizations of reactions and mechanisms from the very start of our university education.

The table of contents (TOC) graphic usually shows a key chemical structure (thus bonding connectivity, angles, elements, and electronic configurations), reaction mechanisms or conditions, models, or even core results as graphs. Now I do not claim to understand what is implied in every TOC graphic, however I can immediately see which branches of chemistry the article relates to faster than I can process the title, especially in a multidisciplinary chemistry journal.

That ability to categorize to identify relevance is a bigger issue in reading a multidisciplinary chemistry journal like the Journal of the American Chemical Society:

https://pubs.acs.org/toc/jacsat/current

But graphical abstracts are still immensely useful within a subdiscipline such as organic chemistry, which I suspect were the early adopters of TOC graphics:

https://pubs.acs.org/toc/joceah/current

The better graphical abstracts tend towards simplicity, but it's a challenging task to distill a year or more of research into a single picture. And, unfortunately, graphic design is not a subject for which scientists generally receive additional training, despite it being a core element of digital communication.