- Downweighting commercial sources and upweighting academic, government, and nonprofit sources.
- Using some measure of quality, maybe even something simple like length + reading age?
- Building big whitelists of quality sites and blacklists of low-quality sites, as picked by human curators and users
Beyond this, perhaps users could plug in sources that are particularly useful for them, including ones they have subscriptions to. I've thought it would be handy to have search results include ebooks, papers, and notes on my computer, for example.
But maybe the userbase of people who need high quality, non-subscription info is too small to get this started.