HACKER Q&A
📣 ads_make_me_sad

Do LLM-based agents generally see ads while navigating the web?


If so, that seems like a vector for ad networks (or others) to influence the output of tools based on web-capable LLM agents.

If not, perhaps imitating an LLM is the ultimate ad block.


  👤 PaulHoule Accepted Answer ✓
Webcrawlers don’t see most ads because most ads are images and/or ads are inserted using JavaScript from another site. (A simple web crawler doesn’t see stuff inserted by JavasScript at all, although an advanced web crawler, like Googlebot, will execute JavaScript and then extract text from the DOM)

There are numerous reasons for this, but one of the most important is that everybody distrusts everybody else in the advertising universe. For instance as a publisher I have an incentive to overstate impressions and but as an advertisers I have an incentive to understate impressions. On top of that the numbers are not expected to match up on both sides because you will lose some users as they go through the funnel.

Having many third parties involved in handling ads means that everybody has to be lying to you for you to get successfully lied to and it’s some of the reason why a site might have 20, 30 or more trackers.