A 28-day experiment tested what happens when a website blocks its content from Google’s AI Overviews and AI Mode while remaining eligible for ordinary search. The site’s AI-feature impressions fell by more than 99% within an hour, from roughly 500-1,000 a day to zero, even though it continued to rank first in regular results. After opting back in, AI impressions returned quickly. Comparing the opt-out and opt-in periods, the site’s owners estimated that links in AI answers had a click-through rate of about 0.9% - roughly twice the site’s 0.4% average rate for Google search overall.
The analysis also found a small “halo effect”: appearing in AI answers was associated with about 2.3% more clicks on ordinary search listings, especially for pages frequently cited by AI. The researchers estimate that AI citations accounted for around 21% of the site’s search clicks, with the halo effect bringing AI’s overall contribution to approximately 23%. They argue that citations can deliver meaningful traffic and reinforce visibility, making opting out a poor choice for their site. These figures come from one publisher and should not be generalized: the experiment used Search Console data from successive 28-day AI-off and AI-on periods, estimating AI clicks from differences in total clicks and checking results against matched queries and rankings. The site had also actively optimized for AI citations, which may have increased its gains.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers tested 16 LLMs on all 55 TIS-100 puzzles, giving each one attempt and time to refine its code. The models found 133 new performance tradeoffs across 18 puzzles and set three outright cycle records, though human players improved on most solutions within days. (tis-100.wincon.dev)
Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent harm caused by their models. The government is also weighing an AI safety watchdog with power to impose significant fines, following a hack of a Medicare statistics portal involving OpenAI. (abc.net.au)
AI is increasingly becoming a gateway through which people find information. Audiences still resist having AI produce their news. (niemanlab.org)
Nvidia holds a central position in the AI industry, making disruption to the company potentially catastrophic for the wider ecosystem. The race for advanced AI is also intensifying economic and geopolitical conflict, while concentrated corporate power and infrastructure face growing political resistance. (france24.com)
LLM inference can require reading about 26 GB of model weights to generate each token from a 13-billion-parameter model. The explanation shows how model size drives memory bandwidth demands during text generation. (youtube.com)
Deletion Tuning matches full fine-tuning across three LLM post-training methods by zeroing fewer than 0.05% of model weights. Its compact masks also preserve prior-task accuracy better than full fine-tuning, LoRA, and random sparse tuning. (arxiv.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.