Hard agree. I was recently at a talk from Jaron Lanier[0], who proposed that AI should, after every query, present on the right-side of the page a list of all clickable sources where AI gathered it's data from, so that we could verify accuracy, as well as allowing us to continue giving traffic to websites.

[0] https://www.jaronlanier.com

edit: grammar

When I do use LLMs, I explicitly ask for all claims with a footnote and the source used for citation.

I almost always get the claim(1) and footnote with URL or book, or DOI.

These will mostly match, but they arent nessecarily the sources just some links that are plausibly the sources.

I haven't used LLMs much but Perplexity always give me tons of links, I really appreciate it vs. chatgpt.

> AI should, after every query, present on the right-side of the page a list of all clickable sources

The default internet device these days is the phone; so many people don’t even use desktop any more. Space limitations on small screens mean that this is unlikely to be shown by default. Moreover, phone interfaces discourage most users from opening multiple new tabs forking off any webpage. You might show desktop users this and get some uptake, but that’s not enough to save the open web.

> Unfortunately it’s in the interests of search and AI companies to keep you inside their portals, so they may be less than willing to link to the outside even when it would improve the experience.

This is true, but aren't "AI" summaries directly opposed to this interest? The user will usually get the answer they need much more quickly than if they had to scroll down the page, hunt for the right result, and get exposed to ads. So "AI" summaries are actually the better user experience.

In time I'm sure that we'll see ads embedded in these as well, but in the current stage of the "AI" hype cycle, users actually benefit from this feature.

Might as well hope that websites optimize for sending people to bookstores.