Perplexity had a big advantage over the competition when model hallucination was bad. That gap has narrowed enough for now.

Perplexity beats Google, ChatGPT, and Claude if you want an answer with citations and want it fast. Claude deep research is more thorough, but that's going to be a wait. ChatGPT web search is slower, uses few citations, and looks like more of the answer is coming from the model than the results. It's also possible that the quality and speed of Perplexity would vanish with scale and the only reason they look so good right now is because they have so many fewer users.

The mid to long term problems I see are:

#1 Google could cut them off from Youtube and a big chunk of their value is gone without recourse.

#2 over time the open web is going to shrivel and switch to paid or even die as ad revenue drops and bot traffic increases (already at this point probably, just countermeasures haven't been fully adopted.)

#3 goes with the previous point, more content is going to be AI generated and not fact checked which will dramatically drop the value of the output. This of course is a problem for all LLMs. Google may be the one who has a big advantage here given their advanced AI research and that they already have an index of the pre-LLM internet.