Three sites made 215,128 “best software” pages for AI. Perplexity cites them

What protection do LLM search engines have against training off content generated by other LLMs?

Will we get to a point where AI-generated sites make up a majority of the internet, and LLMs are training upon their own regurgitations, with exponential amplification of all their lies and flaws?

Or will the pre-2022 corpus human knowledge be considered the low-background steel standard, and anything after that less and less reliable unless certified that it has been created by a human mind and untainted by hallucinations?

Makes sense. Manipulating training data so that models will recommend your product is undoubtedly a big industry.

It's difficult to read more than a few sentences, when this itself is clearly a Claude artifact.

I used one of the 12-month free Perplexity offers when they were everywhere. It felt slightly useful at first for simple queries where I didn’t want to go through the top 10 Google results manually. If I was looking for a specific recipe I remembered or a help page or user manual it would usually find it quickly.

Then they started optimizing for speed of responses over quality of results. I can enter a query and see my results appear in a second, but they’re garbage. The links and references it gives frequently don’t match the text right next to them. It feels like someone had a KPI to make responses as fast as possible and they optimized for that above all else.

They added a “Computer” option that’s supposed to do research for you. Half the time I can’t get it to trigger through the UI. Pressing the submit button doesn’t work. When I can get it to trigger, most of those sessions will work for a while and then just stop before an answer comes back.

The only reason I keep using it is to keep observing a company that has been heavily marketed and hyped, which should have had a market leading position for something. Even non-technical people I know who listen to Joe Rogan (where Perlexity is advertising heavily, I’m told) are asking me about it.

Now there are reports of people being billed at the end of their trial period without warning, despite them saying that they will warn before this happens. There are some alarmingly bad customer support screenshots where the customer support agent (AI? Probably) acknowledges that they didn’t send the email they promised but refuse to help anyway. It takes escalating it on Twitter to get it corrected.

If I want to do actual research or AI assisted web searching I have Claude or ChatGPT do it. The results are so much higher quality and it does exactly what I ask. It may take 45 seconds instead of the instant response from Perplexity but I save time overall because the response and links are more likely to be correct

They’ll train on prompts and anything else you send in. Many LLM responses are sorta finger printable: I assume this is intentional