And it'll happen again when AI models start resorting to ads once again.
Yup. Any LLM recommendation for a product or service should be viewed with suspicion (no different than web search results or asking a commission-based human their opinion). Sponsored placements. Affiliate links. Etc.
Or when asking an LLM for a comparison matrix or pros and cons between choices ... beware paid placements or sponsors. Bias could be a result of available training data (forgivable?) or due to paid prioritization (or de-prioritizing of competitors!)
I don't think that will ever happen. All you need is a trivial browser extension with a locally run, very primite LLM, that takes the output of the commercial LLM, and removes all advertisement. And adBlocker AI, so to speak.
Yes, there will be people not using adblockers just as there are people today. But no adblocker ever was able to remove SEO spam from googles website, all they did was hiding obvious adds. They didn't improve the search experience.