Google's AI-generated search results keep citing The Onion

Rollout and User Experience

• Some users see AI overviews on Google; others (especially in parts of Europe and Brazil) report not having it yet, possibly due to regulatory caution.
• Several find the AI box unhelpful or distracting and look for ways to disable it, mentioning “Web” view tricks and browser tools.
• One person notes Google search already feels more “semantic,” sometimes ignoring critical negations or niche qualifiers in queries.

Hallucinations, Satire, and Misinformation

• The AI frequently cites satirical sources like The Onion and sarcastic Reddit posts as factual, leading to absurd advice (e.g., putting glue on pizza because of a troll comment).
• Users describe it as “pattern matching” that can’t reliably distinguish jokes, sarcasm, or dark humor from serious content.
• A popular example: earlier, it reportedly claimed Barack Obama was a Muslim president. Others tested later and got a corrected answer, suggesting hotfixes or blocking of certain queries.

Bias, Politics, and Content Controls

• Some commenters say the system appears to have swung from one political extreme to another; others argue US parties are not actually “far left vs far right.”
• Observations that Google may be blocking sensitive topics (e.g., “Muslim”) from triggering AI results after failures.
• Comparisons to other LLMs note that heavy RLHF and safety tuning can prevent many of these issues, something Google is perceived as reluctant to do at scale.

Data Sources and Reddit

• Concern that paid access to Reddit data is feeding sarcasm, trolling, and low-quality posts directly into AI answers.
• People wonder how sarcasm and dark humor in such datasets are handled; current failures suggest “they’re not.”

Impact on Search, the Web, and Monetization

• Worries that AI answers will reduce traffic to websites, undermining the content ecosystem the models depend on.
• Speculation that future monetization will involve “sponsored answers” and transactional flows (ordering food, booking services) directly inside AI results.
• Some see AI summaries as more work to fact-check than simply doing a traditional search; others argue generative answers can “cut through SEO bullshit” if done well.