Environment
- Surface: ChatGPT Web
- Operating system: Windows 10, 64-bit
- Browser: Google Chrome
- Affected area: Very long ChatGPT conversations
- Archiving software used: FireShot Pro for Chrome v2.1.4.18 (64-bit)
- FireShot license: Registered Lifetime license
Bug
I regularly use ChatGPT for very long conversations and archive important individual chats completely as PDF files. These archives serve as permanent reference documents so that information from long conversations can be reliably accessed and used again later.
For exactly this purpose, I purchased FireShot Pro with a Lifetime license for approximately €104.
This workflow used to work:
Open an individual ChatGPT conversation → FireShot “Capture entire page” → save the complete conversation as a PDF.
This no longer works reliably with long ChatGPT conversations. Problems with displaying and archiving long conversations have existed in different forms for several months, but the situation has now become significantly worse for my existing archiving workflow.
When I try to capture a long ChatGPT conversation today, FireShot starts the full-page capture but stops automatically after approximately 1–2 seconds. Only a small part of the conversation is captured.
Even more problematic is what happens when scrolling through very long conversations:
Messages or larger parts of the conversation that were previously displayed disappear again. In their place, large white/empty areas sometimes appear.
As a result, software that automatically scrolls through a long webpage in order to capture it can no longer capture the complete conversation.
I specifically tested whether FireShot itself was generally malfunctioning. As a control test, I captured a very long Wikipedia page about Minecraft using the same FireShot “Capture entire page” function.
That control test worked without any problems. The complete long Wikipedia page was captured successfully.
Therefore, on my system, the problem can be reproduced specifically with long ChatGPT conversations.
I understand that modern websites may load content dynamically. Based on my tests and the observed behavior, it seems possible that the current rendering of long ChatGPT conversations involves dynamic loading or virtualization of the scrolling area. However, I cannot confirm whether this is actually the technical cause, and I therefore explicitly do not present this assumption as a confirmed diagnosis.
Expected behavior:
An existing ChatGPT conversation should remain reliably rendered while scrolling.
Messages that have already been loaded, or that are required for displaying the conversation, should not disappear in a way that causes large empty areas during normal scrolling, printing, or automated capture.
It should remain possible to print, save as PDF, or archive a complete individual conversation locally using webpage-capture software, without parts of the conversation being missing.
For me, this is not a minor convenience feature. Archiving my conversations is an important part of how I use ChatGPT.
Reproduction
I have performed numerous tests to narrow down the problem.
1. Standard FireShot full-page capture
I open a very long ChatGPT conversation and select:
“Capture entire page”
FireShot begins capturing but stops automatically after approximately 1–2 seconds.
Result: Only a small part of the conversation is captured.
2. Control test on another very long webpage
To rule out a general FireShot problem, I opened a very long Wikipedia page about Minecraft and used the same “Capture entire page” function.
Result: The complete long Wikipedia page was captured successfully.
FireShot therefore continues to work correctly with long conventional webpages.
3. Manual selection capture
As another test, I used FireShot’s “Capture selection” function.
Manually selecting a very long ChatGPT conversation is not practical. More importantly, while scrolling through the conversation, ChatGPT content that had previously been displayed disappears. Parts of the intended capture area then become white/empty.
Result: Complete archiving is not possible.
4. FireShot – capture a specific HTML element using CTRL
FireShot provides the following function within “Capture selection”:
“Hold CTRL to capture a specific HTML element.”
I tested this as well.
FireShot successfully detects the large chat/scrolling area and then automatically scrolls through the conversation.
This means that the automatic scrolling itself works.
However, the main problem remains:
As FireShot continues scrolling through the long conversation, ChatGPT content further down disappears or parts of the conversation become completely white/empty.
FireShot can consequently only capture these empty areas at those positions.
Result: The resulting capture does not contain the complete conversation.
5. Different positions within the conversation
I also attempted the capture from different positions within a long conversation.
Result: The problem remains. Parts of the conversation are not continuously rendered while scrolling.
6. Investigation of the ChatGPT scrolling container
During troubleshooting, I also investigated whether the problem was simply related to focus on the current ChatGPT scrolling container.
In technical discussions, the current scrolling area has been referred to as:
.thread-scroll-container
Focusing this area may affect certain navigation problems, but in my case it does not solve the problem of content disappearing while scrolling and therefore does not restore complete archiving.
7. Other investigated workarounds
I additionally investigated possible workarounds involving:
- CSS/overflow modifications
- JavaScript or userscripts
- Incognito/logged-out browser operation
- alternative capture methods
- other archiving software
So far, none of these has provided a reliable solution that restores my previous workflow.
Switching to another program is also not a satisfactory actual solution for me. I purchased FireShot Pro with a Lifetime license specifically for reliable archiving of long webpages and ChatGPT conversations.
8. OpenAI data export is not an equivalent replacement for this use case
I am aware of OpenAI’s account data export.
However, my goal is not to request the complete data set of my ChatGPT account and then search through it for individual conversations.
I want to continue doing what was previously possible: open one specific conversation and archive that individual conversation directly and locally.
In addition, during a previous OpenAI data export attempt, the supplied download link failed within the stated availability period and returned only:
{"detail":"Not found."}
Therefore, the account data export is not an equivalent replacement for my specific archiving workflow.
9. FireShot support has also been contacted
On October 6, 2026, I also sent a detailed report to FireShot support.
That report documented, among other things:
- the FireShot Pro version used
- the ChatGPT full-page capture stopping after approximately 1–2 seconds
- the successful Wikipedia control test
- manual selection capture
- CTRL/HTML-element capture
- automatic scrolling
- the subsequent white/empty areas
I am deliberately trying to have the problem investigated from both sides.
Diagnostics
Time of current tests: October 6, 2026, Germany (CEST / UTC+2)
Frequency: Reproducible with my very long ChatGPT conversations.
Observed behavior:
- Full-page capture stops after approximately 1–2 seconds.
- The long conversation is not captured completely.
- Previously displayed content disappears while scrolling.
- Large white/empty areas appear.
- Automated scrolling capture therefore cannot capture the missing content.
- The same capture procedure works on a long conventional webpage (Wikipedia).
Screenshots showing the behavior and additional technical information can be provided if required.
Final request to OpenAI
I would specifically ask that this problem not be treated merely as an issue involving third-party software.
FireShot makes the problem particularly visible to me, but the critical behavior occurs within the rendering of very long ChatGPT conversations: content is not rendered reliably throughout scrolling, or empty areas appear.
Reliable rendering, navigation, printing, and archiving of my own long conversations are fundamental functions for my use of ChatGPT.
I use these conversations as important reference documents and have additionally invested money in software specifically to make this workflow reliable.
It is therefore particularly disappointing to me that problems involving very long conversations have existed in different forms for several months while ChatGPT continues to receive new features.
I have nothing against new features. However, as a paying user, I would like to see existing fundamental functions and workflows – particularly those involving long conversations – receive at least the same priority as the introduction of new features.
Could OpenAI please clarify:
- Is this specific problem already known and internally tracked as a bug?
- Is there an internal bug/ticket number for it?
- Is a fix currently being worked on?
- Is there an officially recommended workaround that allows a single very long conversation to be rendered completely and archived locally as a PDF?
- Could this complete bug report please be forwarded to the team responsible for the ChatGPT web interface and the rendering of long conversations?
I would greatly appreciate a concrete response, as this workflow is genuinely important to the way I use ChatGPT.
Thank you.
P.S. Thank you very much for taking the time to read my suggestion. I may not always be able to reply immediately, but I truly appreciate every response and will answer as soon as I can.