DDoS

July 23, 2025

Pew Research - a pretty popular and rather objective US pollster - analyzed around 70,000 Google searches by 900 survey participants to figure out how AI changes searches. Turns out, by a lot. The most interesting finding was that on pages without an AI summary, users would click on a search result link 15% of the time. Post-AI summary, that percentage drops in about half, 8%. There's also a 1% of clicking on a link from the before-nonexistant AI summary, so let's say 9% total of link follow-through compared to 15%. Similarly, the browsing session ended right after a search in 16% of cases, but when an AI summary exists, that percentage increases to 26%. This can help us reasonably deduct that people read an AI summary and end their search there. Google disagreed with the research, called it flawed, and gave some non-statements that have little subtantive value (e.g. "We drive billions of clicks to websites" - a more cynical reader can read that as "ok you drive 9 billion clicks now when you used to drive 15, you're technically correct I guess"). This trend is more than likely to continue.

It is my personal opinion that AI summaries will lead to the closure of multiple smaller websites that used to depend on ad revenue. It's also very likely that the AI summary will prefer to cite the more popular websites with a higher SEO ranking in its AI summary - even when the content is not originally theirs, but they are republishing content from a smaller or more obscure website with a potentially lower ranking. Will the trend continue? If it becomes more profitable for Google or other search engines, it obviously will. Does that mean you'll see ads and sponsorships in your AI search summaries? Probably, eventually.

In my view there's two types of websites. Those made out of a genuine hobbyist interest for whatever subject they're about and a desire to share it with the world, and websites meant to generate revenue by leveraging SEO, ads everywhere, selling user data etc. I believe that the former are not threatened by this in the least, except potentially in the level of discoverability. But who knows, adversity drives innovation. As for the latter, I believe it should sink or swim. There are various schemes out there that do not involve ads that help to fund these websites already, from Brave's crypto scheme - however one feels about it - to more straightforward subscription/donation pipelines. There will probably be more to come in the future too.

Anyways, on to our favorite funny pages. A Hacker points out that AI overviews are wrong a lot of the time. This is objectively correct, but also either a bad faith argument or simply stupid. AI has, for better or worse, come leaps and bounds over the past few years. It is plain stupid to believe that any further improvements are impossible that would make this argument irrelevant - too costly, perhaps. That does not stop other Hackers from stating their favorite anecdotes of times that AI has been wrong before.

Hackers shill their favorite LLMs, in true YC fashion. Other Hackers embrace the no-reading future in which a machine decides for them, in true Hacker fashion. One can only hope that soon, the AI summaries nuke these Hacker threads, and the internet turns just ever so slightly less dumb.