For the last two years, DuckDuckGo has been my default search engine. Before that, I bounced between Google and Bing, mostly driven by how bad search has become. DuckDuckGo is decent for 80% of the searches and the rest can be done through Google or Bing. Since the last year or so, AI summary is shown as the default answers on google and they are sometimes wrong by a wide estimate. Sometimes I really don’t need an AI summary and it seems there is no way to turn this off. Search quality has been in a long, steady decline over the last seven years. SEO spam, content farms, affiliate content, and now LLM-generated pages stacked on top of each other.
Research on SEO spam shows an Inverse relationship between affiliate marketing usage and content complexity: pages with higher concentrations of affiliate links tend to contain simpler, lower-quality text. The same research also shows that all major search engines fall victim to large-scale affiliate link spam campaigns, where coordinated networks of monetized pages successfully manipulate rankings across engines.
Another analysis of Google’s search results shows that 16 major media/publishing companies collectively own at least 588 individual brands and capture billions of search clicks per month. This highlights how a relatively small group of publishers shapes what most people see in Google search. The best example of this degradation is the slop that you get is if you search for a recipe on a search engine these days.
If I had programmatical control over what I can get out of the search indexes, I would block these as a whole with a block list. At some point, I might need to look at paid search alternatives like Kagi. In the last two years, AI-powered search tools felt like a real improvement for a while. In 2024–2025, Perplexity was genuinely useful. It cut through noise, gave good references, respected most source material, and saved time. But over the last six months of 2025, even Perplexity Pro has degraded. Results feel less reliable, more repetitive, and increasingly opaque. What really bugs me is how it often ignores my explicit choice of model, I pick Claude or GPT, and it quietly falls back to some default even if it does not show it. After doing the same searches across Claude, GPT, and Perplexity, I consistently noticed that GPT and Claude were just better and precise than going through perplexity pro. My perplexity pro usage fell down after Oct 2025 and I don’t think I will renew it when it is time this year.
I started using Ecosia in December, mostly out of curiosity. I didn’t expect much. But after a month or so, i feel it stuck with my browsing habits. It is better than DuckDuckGo. It has the vibes of using Google before 2019. Ecosia also looks like a good sustainable search alternative. I’ve also spent time with Qwant, the French search engine, and it leaves a similarly positive impression, it also has Qwant Junior, an interesting product. It’s not perfect, but it feels well designed in a way that many search products no longer do. Both of these have less aggressive AI integration, fewer dark patterns, and a clearer separation between search results and guesses. Ecosia is now my default search engine, with Qwant as a solid alternative. If neither of them don’t work. I still can go to the traditional search. A sweet bonus is that both of these companies are European, probably better for privacy.