“After feeding perplexity about 500 in credits to work on a project I swapped to Claude. So glad I did because I did a complete overhaul for 30 dollars.”
“They even nerfed the fucking browser use feature by gating it under Max. Idk when they changed it but that was the one thing that MAY have kept subscribed because the convenience was nice every once in a while.”
“\- for citations specifically (perplexity, google AI overviews), some tools scrape the actual source links shown and cross-reference against your domain, since that's more reliable than trying to parse it out of the generated text itself”
“That also answers your Perplexity point, and not the way I expected: Perplexity isn't inflating relative to ChatGPT (5.2 vs 5.6 products named per answer). Gemini is the outlier for terseness.”
“The Perplexity scores consistently higher makes sense given it's search-augmented, but I'm curious if that inflates name-checking over actual usefulness. Did any product completely drop off a single model's radar that you'd have expected to dominate?”
“Stop researching and start testing, open ChatGPT/Perplexity/Gemini yourself and ask the actual questions your customers would ask, see if you show up and who does, that 30 minutes tells you more than another week of reading conflicting advice.”
“Perplexity is a search product with a model attached, so it wins when the answer lives in documents you have not read yet. Claude wins when the material is already in front of it and the work is reasoning over it, which is what code review and long-form drafting actually are.”
“I would then test these questions using ChatGPT, Google AI Overviews/AI Mode and Perplexity, among others. Then, look at which companies are being mentioned, which sources are being used, and what those companies are doing differently.”
“Can't make up my mind as my work is mainly git repo code review, blogging and research and my wife is a teacher so leaning towards perplexity pro but need suggestions on what others are using”
“Write 40 prompts a real buyer would actually type, hit the OpenAI, Gemini and Perplexity APIs on a weekly cron, and log three things per run: brand mentioned yes/no, where in the answer, and which sources got cited. Costs a few dollars a month.”
“we recently started using Shiprocket AITLAS, and its been surprisingly useful. instead of just tracking Google rankings, it shows how our brand appears across ChatGPt, Gemini, Google AI overviews, AI mode, and perplexity, seeing the actual AI responses made a huge difference.”
“Its still better than gemini native app for me, in terms of quality of output and keeping context. But it's restricted in terms of usage. I miss what perplexity used to be!”
“I'd start by running 20 to 30 buyer questions with no brand name in them ("best X for Y", "X alternatives") through ChatGPT, Perplexity and Gemini, and tracking three things per prompt: are you named, is your URL actually cited, and who shows up instead. Do that on a schedule, because a one-off check goes stale in a week.”
“Anfänglich fand ich Perplexity klasse, solide Ergebnisse, dann die Paypal-Aktion - 1 Jahr kostenlos. Aber: In dieser Zeit wurden die Ergebnisse immer schlechter im Vergleich zu Claude oder Google Gemini. Inzwischen ist Perplexity von meinem Desktop verschwunden und ich nutze überwiegend Claude und Gemini.”
“Perplexity is just a harness for better AI. It had its time, but ChatGPT and Claude don’t really need harnesses like Perplexity anymore. So why still use essentially a watered down version?”
“I get the out of credits message when i don’t notice that computer is switched on and accidentally start a “search” with it. Did you double check if you’re using Search instead of Computer?”
“I use pro for research on various topics. It does a great job synthesizing information and finding good, legitimate references and creating APA style citations. I've been able to train it to write in various voices and it does a fantastic job at initial drafts with specific instructions.”
“Perplexity hängt halt viele monate zurück mit den Modellen die sie anbieten, da hast du mehr bei den Top frontier Modellen wie Claude oder chatgpt, der rest ist whack”