Docognito catches up a long conversation automatically. The moment you open one, it scrolls the real page so older messages mount, then keeps every message it has rendered even after the site later drops it from the page. A gap that survives a few seconds means that sweep is still running or has hit its bounded limit, and a manual scroll or reload is the rare fallback.
What is happening when a thread looks incomplete
Opening a conversation with hundreds of messages does not pull every one of them onto the page at once. Both sites load a long thread in stages, and the part that loads first is not the same part on both sites. What looks like a missing chunk of your conversation is usually a chunk that has not been asked for yet, not a chunk that has been lost.
This matters because the document view Docognito draws is not a separate record of anything you have written. It is a redraw of whatever is already sitting in the page at the moment it checks, which means the two are always in step with each other. If the underlying page has not loaded a stretch of messages, the document view cannot show that stretch either, because there is nothing there yet to show.
ChatGPT drops turns, Claude loads from the top
The two sites handle a long thread in close to opposite ways, which is worth knowing so you read the gap correctly rather than guessing.
| Symptom | What the site is doing | What to do |
|---|---|---|
| Only the newest few messages are on the page, older ones are gone | ChatGPT keeps just the recent stretch of a long thread mounted, and removes earlier turns from the page to save memory | Scroll upward inside the conversation, which asks the site to bring an earlier stretch back |
| Only the earliest messages are on the page, and it fills in gradually | Claude loads a long thread starting from the top, adding newer stretches as it goes | Give it a few seconds, or scroll down toward the end of the thread |
The row worth remembering is that ChatGPT drops from the top and reloads it on demand, while Claude fills in from the top on the way to the bottom. They are not the same behaviour wearing two names, and Docognito's own catch-up sweep, covered next, is built to run both directions rather than assuming one.
How Docognito fits into this
Docognito does not wait for you to notice a gap. The moment you open an existing conversation, it invisibly scrolls the real page's own scroll container the way a reader would, pulling the site into loading whatever it has not mounted yet. The sweep runs in both directions, because the two sites are lazy in opposite ways: it hammers the top of the page first, for a site that keeps only the newest turns mounted and reloads older ones on scroll, then steps back down toward the bottom for a site that mounts a long thread starting at the top. Every message it finds along the way gets rendered into the document.
That rendering is also kept, not just refreshed. Once Docognito has drawn a message into the document, it stays there even after the site later unmounts that same message from its own page to save memory. The document view is not only catching up to whatever currently sits on the page, it is remembering everything it has already seen, so a message does not disappear from the document a second time just because the site removed it a second time.
The sweep is bounded rather than endless, and it steps aside the moment you start typing or a reply starts streaming, so it never fights a conversation in progress. On a very long thread it can still be running when you first look, or it can reach its round limit before the site has produced everything it has. That is the case a manual scroll or a reload now actually covers, not the general one.
Telling still loading from actually stuck
Give the automatic sweep a moment first
Docognito starts scrolling the real page for you the instant you open an existing conversation, so most gaps close before you would even reach for the mouse.
Scroll toward the gap if it is still open
A short manual scroll near the top or bottom of the visible messages covers the rare case where the automatic sweep is still working through a very long thread or has reached its round limit.
Reload if nothing changes
A hard refresh restarts the loading from the beginning and clears out any stalled request. It does not touch your account or delete anything you sent.
When a reload is genuinely faster than waiting
If the automatic sweep and a short manual scroll do not change anything, a reload is usually quicker than continuing to wait, since it restarts whichever request stalled rather than leaving it hanging. This is close to the same troubleshooting step you would reach for on the plain site, disguise or not, because the cause sits with how the page loads its own history, not with anything the extension adds on top of it.
A genuinely stuck load, one that the automatic sweep, a manual scroll, and a reload all fail to fix, is uncommon and usually points to a connection problem or a temporary fault on the assistant's side rather than anything about the document view itself.
What to try, in order, before assuming something is lost
Treat a partly loaded thread as a loading state first, not a fault, and let Docognito's automatic sweep run before assuming anything is missing. Scroll toward the gap yourself only if it is still open after a few seconds, and reload only if neither changes anything. The document view keeps every message it has already rendered, so catching up an incomplete thread is about the real page finishing its own load, not about anything Docognito has lost.
Use Docognito and read a gap in a long conversation as a page Docognito is already working to catch up, not one you need to nurse along by hand.