Google, the search engine that's forgotten how to search

AI Overviews and User Experience

• Many dislike AI summaries: slow to load, occupy most of the mobile screen, add cognitive load, and often answer superficially or incorrectly.
• Some report increasing usefulness, especially for physics/programming and simple factual queries, and have switched back to Google for this.
• Users are split between finding them occasionally valuable vs. wishing they were optional or in a separate tool.

Ethics, Copyright, and Publisher Impact

• Strong concern that AI overviews monetize scraped human content while disincentivizing visits to source sites.
• Blocking Googlebot is the only way to avoid inclusion in AI summaries, which effectively means disappearing from search; some see this as no real choice, especially given Google’s market power and antitrust findings.
• Worry that traffic loss harms sites like Stack Overflow and small publishers most.

Search Quality, Ads, and Monetization

• Many perceive a long-term decline in organic result quality, with more content farms, brand bias, and UI clutter (knowledge panels, videos, “people also search for”).
• Others say Google still works well, especially compared to alternatives, and see complaints as overblown.
• Debate over whether Google is optimizing for ad revenue at the expense of user experience vs. needing strong organic results to sustain its ad business.

Alternatives and Workarounds

• Mentioned alternatives: DuckDuckGo, Brave Search, Yandex, Kagi, ChatGPT, RSS feeds, curated “awesome lists,” personal bookmark search.
• Some find DDG better for local/physical searches; others say DDG is now polluted by AI-generated pages.
• Yandex is praised for “old Google”-style simple ranking and better relevance on some technical queries.
• The udm=14 Google parameter and related tools are shared to restore “classic” results and remove AI overviews.

LLMs, Hallucinations, and Reasoning

• Multiple anecdotes of wrong or absurd AI answers (e.g., pyramids lit by electricity, misattributed authors, conflating fork travel with axle-to-crown).
• Technical debate on whether LLMs require exact word occurrences in training; consensus that they generalize but can overfit patterns, amplify bias, and hallucinate on niche topics.
• Recognition that RLHF and tuning improve behavior, but issues like sycophancy and susceptibility to crafted prompts remain.

Content Creation, SEO, and the “Dead Internet”

• Some creators have stopped blogging, feeling their work will just be “slurped” into LLMs without reward, threatening future high-quality content.
• Concern that if content becomes a commodity, creators will disengage unless platforms or AI firms start paying for it.
• Criticism that much of the SEO industry produces low-value “SEO content,” contributing to search degradation; others argue there is legitimate strategic/technical SEO work.
• Discussion of a broader shift from open blogs to algorithmic, walled-garden platforms, making deep, independent content harder to surface.

“Normie” vs. Power-User Perspectives

• One camp claims that for ordinary users—finding businesses, maps, products—Google is still excellent and dominant by merit, not lock-in.
• Others note “normies” around them also complain about degraded results, especially for shopping, hotels, and niche queries, and argue people stick with Google mainly because it’s the default and “good enough.”