An Alternative to Bad Google Search Results: Kagi

I’ve suspected that Google search is getting worse, but now my suspicions have been confirmed. The rise of AI-generated content, reports Platformer, “has led to a flood of AI-generated spam that researchers say now outperforms human-written stories in Google search results.”

This is why I’m testing kagi.com as my search engine and liking what I see. To use Kagi, you pay for it. I’m amazed to hear myself say that I think it’s worth it. The fee is small, the results are better, are well-presented, without garbage. You can customize the layout. Extensions make Kagi easy to use with the Arc browser, Firefox, or Vivaldi. (Or with Chrome, if you enjoy web surveillance.) Kagi has a bit of AI itself, because you can ask the built in FastGBT (via API) to summarize results or pull highlights from the page you’re looking at. After it does this, it footnotes the results, so I can check for accuracy.

In my early fooling around with Claude.ai, and Descript’s AI tools, and also Podium.page for podcasts, I can say that Kagi is giving me results that I can depend on, because I can check them right on the page. I used Claude.ai for some research, and got hallucinations, including made up URLs. Podium.page is promising for podcasts, creating decent titles and show notes.

But if it’s one thing that these LLM-driven tools can do dependably well is summarize and provide highlights of web pages. Kagi delivers on that.