Their integration of their own index was fairly recent when I added them. I’m not able to give details.
I’ve been thinking about splitting up the “semi-independent” section up to distinguish Brave and Kagi from Qwant and Neeva. Rationale:
Brave does have a truly independent index, but it’s designed to be extremely similar to Google. They published a paper on “goggles” to allow users to skew ranking behavior for different purposes. I think these two facts really do make it much more independent than Qwant and Neeva. I’m also tracking certain queries to see how SERPs change with time compared to Google, and have seen a few positive results.
Kagi is actually super interesting, and I think it offers the most unique results in the “semi-independent” category with its non-commerical index and bias: results from its independent sibling Teclis engine do show up, and they’re quite relevant. Kagi also offers “lenses” similar to Brave’s upcoming “goggles”, and its “non-commercial” lens is really quite something. I’m not sure I should give more details, but suffice to say that it really does go out of its way to properly integrate both mainstream sources and its Teclis index, rather than just inserting them randomly.
I should also add that reverse-engineering Google and Bing is something that lots of these engines engage in. Not receiving scoring data hasn’t stopped Brave (well, Cliqz at the time) and at least two other engines from coming up with their own way to determine ranking factors at play. A common strategy is to compare changes over time between their own SERPs and GBY’s and monitoring pages for changes that could trigger them, feeding these “change signals” into their own models. Cliqz’ tech blog detailed something similar: https://www.0x65.dev/