Kagi Sidekick (alpha)

Original Article ↗ Hacker News Discussion ↗

Kagi has introduced Sidekick, an embeddable site-search and optional AI assistant intended to help small and documentation-heavy websites while simultaneously feeding more targeted content into Kagi’s own search index. Paying users generally praise Kagi’s core search quality but are split on whether ventures like this and other AI features strengthen the product or risk distracting from its primary mission, raising questions about sustainability, privacy of on-site queries, and potential SEO abuse. The company frames Sidekick as a low-cost, win–win way to expand a high-quality index without ads, with safeguards promised against spammy or LLM-generated content.

Scope and Purpose of Sidekick

  • Sidekick is presented as an alpha “Labs” prototype, built quickly by one developer to gauge demand.
  • Main goals:
    • Crowdsource and expand Kagi’s own index, especially small, personal, niche, and documentation sites.
    • Give those sites better on-site search and visibility in Kagi results.
    • Increase Kagi’s brand exposure via embedded widgets.
  • Free use is planned for small/personal sites; details for large/nonprofit sites are unclear.

Focus on Core Search vs. New Features

  • Many paying users emphasize they want Kagi to focus on high‑quality web search and configurable summaries, not side products.
  • Others argue Kagi must invest in AI and adjacent tools or risk obsolescence if AI reshapes search.
  • A Kagi employee states the search business is sustainable and Sidekick/AI is about complementing search, not patching cashflow.

Data, Privacy, and “You Are the Product”

  • Some see Sidekick as a clever, low-cost way to discover and rank content via user queries and interactions on third‑party sites.
  • Others worry this means users are again “the product,” even with a paid service.
  • Questions are raised about whether widget queries follow Kagi’s existing privacy practices; no explicit answer is quoted, so this is unclear.

SEO Gaming and Spam Risk

  • Concern: embedding Sidekick and using user activity as a ranking signal could become a new SEO attack vector or pollution channel.
  • Responses:
    • Kagi’s small user base currently makes targeted gaming unlikely.
    • Kagi already penalizes ad/affiliate-heavy sites and allows user blocking/downranking.
    • If abuse grows, ranking algorithms could be adjusted.
  • There is debate over low‑quality SEO content vs. legitimate company blogs; some see most SEO as harmful, others note many helpful dev and product posts.

AI Functionality and Control

  • Sidekick’s AI Q&A/summarization is optional and off by default. Several potential adopters insist on non‑AI search only.
  • Some see AI features (Quick Answers, Wolfram|Alpha integration) as major benefits that reduce clicks; others distrust instant answers due to past errors on other engines.
  • Kagi mentions developing LLM‑content detection and penalizing LLM‑generated spam, but multiple commenters doubt reliable detection is possible.

Integration Questions and Comparisons

  • Comparisons are made to Algolia DocSearch and other site-search tools; users request plugins for MediaWiki, wikis, static site generators, and CLI integration.
  • Technical writers ask about index scope (domain vs. subdomain), indexing frequency, access to logs and embeddings, and how user queries are reused; answers are mostly not provided here.