If you connect file connectors such as Google Drive and Dropbox in Perplexity, and after changing the original file, it still answers according to the old content, which is usually not because the cache has not been refreshed, but because you understand "connector file selection" as "continuous real-time synchronization". Perplexity currently states this very clearly: if the original file changes, you usually need to re-upload or re-select it, and the system will not automatically catch up with the latest version for you.
This is also the most misunderstood by many people. Seeing that it can select files from cloud disks,Naturally think that the original document will be changed in the future,Perplexity will automatically update accordingly。 However, at this stage, many file connection scenarios are more like "bringing the selected content into this thread as an attachment or context" rather than giving you a knowledge base that is always synchronized in real time.
In this case, the simplest way to deal with it is very simple:
1. After changing the original file, upload it again in a new thread or select it again from the connector.
2. If you are still asking in the old thread, don't default to the new version of the system.
3. If the data is to be checked repeatedly for a long time, prefer a file space that is more suitable for continuous management rather than relying solely on one-time thread attachments.
There is also a common mispositive that is "I can see the same file name, so it should have been updated". The same file name does not mean that the context content has been refreshed, especially in threaded Q&A. The system is actually using the version you sent in at that time, not the real-time state behind the file name in the cloud.
If you're in a teamwork scenario, this pit will be more obvious. One person changed the Drive documentation, and the other person was still asking about the old thread, and it was easy to think that Perplexity was hallucinating, but in fact it was just referencing the old version context.
So when you find that it is still answering old content, don't rush to question the model. Do a minimal test first: open a new thread, re-select the updated file, and ask the same question. If the answer changes, the question is not about reasoning, but about what you expect from how the file is connected.
In a word, Perplexity files are easy to connect, but don't automatically think of it as "real-time two-way syncing". Once the original file is updated, you often have to manually resend it.
