For this experiment, we turned to Perplexity because ChatGPT immediately said that it had little or no memory across separate chats. Perplexity, by contrast, was less quick to rule this out.
So we asked what our year had looked like from Perplexity’s perspective. According to the tool, our main topics were urban planning, commercial areas and transport, accounting for 40% of our queries, followed by citizens’ initiatives, participation formats and council information systems at 25%.
And what would we want for Christmas, according to Perplexity, based on our searches?
- Access to specialised information in a clean, searchable structure
- Our own specialised AI agents to take over recurring research, flag sticking points and identify legal or planning obstacles at an early stage
- Enough time and energy to continue carrying out in-depth research without burning out – including a few completely work-free days over the festive period
Perplexity also gave us two fun facts:
- We ask about citizens’ initiatives at least as often as other people ask about restaurants.
- Our combination of search queries is highly unusual and, statistically speaking, falls somewhere within a fraction of a percent of all queries submitted to Perplexity.
So apparently we are less interested in material things than in information. The topics we research also seem to be fairly unusual. Thank you, Perplexity! We feel honoured.
But are these claims actually accurate? Our conclusion:
It is an entertaining feature, and it did produce a few observations that felt surprisingly accurate. On closer inspection, however, we found that some aspects and topics were missing entirely. Even when we asked specifically about individual topics from other chats, the answers were not always correct.
For recurring topics, it can therefore be useful to ask Perplexity to draw on information across different chats. However, it is worth keeping in mind that this function is not yet fully reliable.
By Ida Kandler
