Docker deploy in 3 minutes, configure in 10
`docker run -d -p 8080:8080 searxng/searxng` gives you a working search engine at localhost:8080. The defaults query Google, Bing, DuckDuckGo, Wikipedia, and a few others. The real power is in `settings.yml`: you can enable 70+ search engines, set rate limits, configure result ranking, and add custom engines. I enabled Google Scholar, Arxiv, GitHub, Stack Overflow, and Hacker News specifically for AI research. The config file is well-documented and changes take effect on reload, not restart.
How it helps with AI tool research
When I research AI tools for saas.pet reviews, Google gives me SEO-optimized listicles ('Top 10 AI Tools in 2026') that all link to the same 10 tools. SearXNG with Arxiv, GitHub, and Hacker News enabled surfaces tools that are actually being used by developers and researchers, not just marketed heavily. I discovered 3 tools through SearXNG that never appeared in Google's top 50 results: a niche video transcription tool with 200 GitHub stars, a Chinese OCR model with an active Discord community, and a browser automation library that 3 saas.pet readers recommended after I reviewed it.
The privacy model: what SearXNG strips
When you search Google directly, your IP, browser fingerprint, search terms, and click behavior are logged and used for ad targeting. SearXNG acts as a proxy: it sends the search query to Google on your behalf, strips your IP and cookies, and returns results without any tracking. Google sees the query coming from your server IP, not your device. SearXNG also removes sponsored results by default (configurable), removes URL tracking parameters (utm_source, fbclid, etc.), and does not store search logs by default. The instance logs are configurable: you can set log retention to 0 days for complete privacy.
Running on a $5 VPS: performance and limits
My HK server (1 vCPU, 2GB RAM) runs SearXNG alongside n8n and a few other services. SearXNG uses 150MB RAM idle, 400MB under load. Each search query takes 1-3 seconds because it queries multiple engines in parallel. The bottleneck is Google's rate limit: after about 100 queries per hour, Google starts returning CAPTCHAs. SearXNG handles this gracefully (falls back to other engines), but if you need high-volume Google results, you need multiple IPs or a proxy rotation. For personal use (10-20 queries per day), a single VPS is fine.
SearXNG vs Kagi vs DuckDuckGo vs Google
SearXNG: self-hosted, aggregates all engines, maximum privacy, requires server maintenance. Best for developers who want control. Kagi: paid ($5/month), high-quality results, no ads, no tracking. Best for people who want better search without self-hosting. DuckDuckGo: free, privacy-respecting, results are mostly Bing with some own index. Best for casual privacy-conscious users. Google: best results for most queries, worst privacy, most SEO spam. Best when you just need the answer fast. I use SearXNG for research, Kagi for daily search, and Google only when SearXNG and Kagi both fail.
