Google is reshaping its AI Overviews again, this time by pulling in quotes from Reddit threads, web forums, blogs, and even a user’s own news subscriptions. According to TechCrunch AI, the update adds a preview of “perspectives from public online discussions, social media, and other firsthand sources” directly inside the AI-generated answer, alongside richer context like creator names, handles, and community labels.
This is Google’s attempt to fix a problem of its own making. AI Overviews, launched two years ago to push generative answers above the classic blue links, have a credibility tax. They’ve been caught recommending one small rock per day (citing The Onion) and suggesting glue on pizza to keep cheese stuck (citing Reddit). The fix, it seems, is to lean harder into Reddit, but with proper labels so you know what you’re reading.
What’s actually changing
- Forum and blog excerpts: Snippets from public discussions now appear inline as part of the AI response, not just buried in source links.
- Creator and community context: Each cited link shows who posted it, which subreddit or forum it came from, and the handle behind it.
- News subscription highlighting: If you pay for a publication, links from that outlet get visually elevated in the response.
- Firsthand-source framing: Google is positioning these as “perspectives,” not authoritative answers, which softens the accuracy bar.
TechCrunch AI notes the design rhymes with how ChatGPT and Claude already drop citation links to back up claims. The big difference: Google operates at trillions of queries a year, so even a small hallucination rate translates to staggering absolute volume.
Why this matters
The shift signals something interesting about how Google now sees its own AI product. For two years the pitch was that AI Overviews would answer your question. Now Google is admitting that for a huge slice of queries, there is no clean answer, just opinions from real humans who’ve tried the thing. That’s why “reddit” became one of the most appended words in Google search history.
What stands out here is the tension this creates. If the AI Overview is now serving you a buffet of forum perspectives, what exactly separates it from a regular search results page? Google is essentially admitting that AI synthesis works for some queries and fails for others, and rather than pick a lane, it’s stacking both inside the same UI element. TechCrunch AI raises this same concern directly.
The accuracy problem hasn’t gone away
A recent New York Times analysis cited in the report found AI Overviews are correct about 9 times out of 10. Sounds good until you do the math at Google’s scale. Hundreds of thousands of inaccurate results per minute is not a rounding error, it’s a feature with a known defect rate.
Adding forum quotes cuts both ways:
- Upside: Niche questions like “is this normal after surgery” or “does this car part fit my model” genuinely benefit from real human voices. Forum context is often where the truth lives.
- Downside: Reddit and forum content is unfiltered. Sarcasm, trolls, outdated advice, and confidently wrong takes are part of the package. The glue-on-pizza incident wasn’t a bug, it was Reddit working as designed.
The new context labels (creator, community, handle) are the safety net. Whether users actually read them before trusting the snippet is another question entirely.
Availability
Google is rolling the changes into existing AI Overviews, so anyone already seeing AI-generated answers in search will pick up the new format. No separate signup, no premium tier, no waitlist mentioned. The news subscription highlighting depends on whether you’re logged in and have active subscriptions linked to your Google account.
The practical advice from TechCrunch AI is the same one that applies to every LLM-powered tool: double-check that the AI isn’t hallucinating the validity of the citations themselves. A fake quote with a real-looking handle is still a fake quote.
More details at the original source.