NoCap
Account

AI frustrations

#1

Ignores explicit instructions

12 threads

Users must repeat constraints that the assistant was already given.

Answer qualityChatGPT · Copilot · Gemini
Read 12 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#2

Forgets earlier conversation context

9 threads

Details established earlier disappear from later answers.

Memory & continuityChatGPT · Claude Code · Copilot · Gemini
Read 9 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#2

Quality changes after an update

9 threads

A previously useful workflow becomes less dependable after a product change.

Reliability & accessChatGPT · Copilot · Cursor
Read 9 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#4

Changes more than the user requested

5 threads

An edit alters unrelated content or a much larger portion of a project.

Workflow & controlChatGPT · Claude Code · Cursor · Gemini
Read 5 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#4

Invents facts with confidence

5 threads

Plausible answers contain unsupported or fabricated information.

Answer qualityChatGPT · Perplexity
Read 5 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#4

Pays but cannot access the paid features

5 threads

An account is charged while the interface offers free-tier access or missing features.

Cost & limitsChatGPT
Read 5 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#7

Behaves inconsistently between attempts

4 threads

Similar requests yield different behavior across chats or sessions.

Reliability & accessChatGPT · Gemini
Read 4 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S02 · Reddit
    It’s incredibly disappointing ↗
    Gemini · Publication date not recorded

    The author reports different outcomes across accounts.

    Reported; current resolution not established
  • S03 · Reddit
    Is Gemini’s image gen just totally cooked right now? ↗
    Gemini · Publication date not recorded

    The same edit succeeds in one chat but fails in another.

    Reported; current resolution not established
  • S08 · OpenAI Community
    Please don’t retire GPT-5.1 Thinking — reply 36 ↗
    ChatGPT · Source date 2026-03-03

    Behavior changes inside the conversation.

    Reported; current resolution not established
  • S49 · Reddit
    Model Swapping ↗
    ChatGPT · Publication date not recorded

    Response style and quality change within a session.

    Reported; current resolution not established · Perceived model switching is reported by the user; backend routing has not been independently verified.
#7

Requires repeated corrections and retries

4 threads

Users spend extra turns restating requests or repairing answers.

Workflow & controlChatGPT · Gemini
Read 4 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#7

Support does not resolve the problem

4 threads

Repeated support contacts or troubleshooting steps fail to restore the workflow.

Reliability & accessChatGPT
Read 4 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Agrees or flatters instead of correcting

3 threads

Users want critical feedback rather than automatic validation.

Answer qualityChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Apologizes but repeats the mistake

3 threads

Promises to correct behavior do not prevent the same failure.

Conversation & toneChatGPT · Gemini
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S02 · Reddit
    It’s incredibly disappointing ↗
    Gemini · Publication date not recorded

    Apologies do not improve instruction adherence.

    Reported; current resolution not established
  • S04 · OpenAI Community
    Catastrophic Failures of ChatGpt — discussion page 3 ↗
    ChatGPT · Source date 2025-06-20

    Accountability language accompanies repeated errors.

    Reported; current resolution not established · Findings refer to visible replies on this page, not the original topic date.
  • S32 · OpenAI Community
    GPT-5 has a temperature issue ↗
    ChatGPT · Source date 2025-09-08

    A reply says apologies are followed by the same mistakes.

    Reported; current resolution not established
#10

Claims work was done without an artifact

3 threads

Completion claims conflict with the resulting files or records.

Workflow & controlChatGPT · Claude Code · Gemini
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Gets trapped in a login loop

3 threads

Successful sign-in attempts return to login or account creation.

Reliability & accessChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Interface freezes or becomes unresponsive

3 threads

Typing, scrolling or viewing responses hangs during a chat.

Reliability & accessChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Long chats become painfully slow

3 threads

Large conversation histories make the interface difficult to use.

Reliability & accessChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Mobile and desktop behave differently

3 threads

The same account or conversation has different restrictions across clients.

Reliability & accessChatGPT · Copilot
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Project or knowledge files stop working

3 threads

Persistent document collections become unreadable or fail to upload.

Files & researchChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Refuses requests users consider harmless

3 threads

Users report safety refusals on requests they see as ordinary or permitted.

Conversation & toneChatGPT · Gemini
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Repeats answers to old topics

3 threads

Already answered points keep appearing after the task has moved on.

Conversation & toneChatGPT · Claude Code · Copilot
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#10

Sources or citations do not support the answer

3 threads

References are fabricated, misplaced or irrelevant to the associated claim.

Files & researchChatGPT · Perplexity
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S30 · OpenAI Community
    ChatGPT generating Fake? references ↗
    ChatGPT · Source date 2023-01-20

    The author cannot locate any of the supplied references.

    Reported; current resolution not established
  • S50 · Reddit
    Citations are wrong ↗
    Perplexity · Source date 2024-10-07

    The author reports incorrect or misplaced references after checking.

    Reported; current resolution not established
  • S51 · Reddit
    Gives Wrong Citation/Link Most the Time? ↗
    Perplexity · Source date 2025-03-02

    A digital-twin answer links to an unrelated robotics paper.

    Reported; current resolution not established · A reply disputes some examples; the catalog records the complaint, not an independently reproduced failure.
#10

Stops mid-task without saving progress

3 threads

A limit or interruption ends work before state can be saved.

Workflow & controlChatGPT · Claude Code · Copilot
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S05 · GitHub
    Rate limit terminates active session with no save opportunity ↗
    Claude Code · Publication date not recorded

    The report says there is no final turn to save progress.

    Reported; current resolution not established · User-submitted issue drafted with model assistance; incident not independently reproduced.
  • S17 · OpenAI Community
    ChatGPT 4o bug — keeps on crashing ↗
    ChatGPT · Publication date not recorded

    A reply reports continue-generation failures or restarts.

    Reported; current resolution not established
  • S59 · Reddit
    The new Copilot update is terrible ↗
    Copilot · Publication date not recorded

    A reply says a mobile chat ends after a message threshold.

    Reported; current resolution not established
#10

Usage limits are hard to predict

3 threads

Users cannot easily anticipate when their allowance will run out.

Cost & limitsChatGPT · Claude Code · Cursor
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S05 · GitHub
    Rate limit terminates active session with no save opportunity ↗
    Claude Code · Publication date not recorded

    The author requests progressive warnings and usage visibility.

    Reported; current resolution not established · User-submitted issue drafted with model assistance; incident not independently reproduced.
  • S40 · OpenAI Community
    Increase ChatGPT’s Memory — constantly full ↗
    ChatGPT · Source date 2024-06-20

    A reply reports no warning before memory fills.

    Reported; current resolution not established
  • S55 · Cursor Community
    Manual mode missing ↗
    Cursor · Source date 2025-07-23

    The user worries about agent context reads increasing token costs.

    Team reply confirms removal and explains changed default context behavior.
#10

Voice replies interrupt the speaker

3 threads

The assistant treats thinking pauses as the end of a spoken request.

Voice & accessibilityChatGPT
Read 3 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Cannot find or complete subscription cancellation

2 threads

Account or billing problems make cancellation difficult.

Cost & limitsChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Cannot upload files

2 threads

Upload controls or transfers fail, including on paid subscriptions.

Files & researchChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Chat history disappears or cannot be opened

2 threads

Saved conversations become unavailable or vanish from the interface.

Memory & continuityChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Does not follow saved preferences

2 threads

Saved user preferences fail to influence later responses.

Memory & continuityChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Does not preserve the requested writing style

2 threads

Tone, vocabulary or character conventions drift away from the brief.

Conversation & toneChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

File links or analysis sessions expire

2 threads

Previously generated files become inaccessible after session expiry.

Files & researchChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Generated file downloads fail

2 threads

The assistant supplies an artifact that the user cannot download.

Files & researchChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Image edits change the original unexpectedly

2 threads

A local image edit produces a newly composed or broadly changed picture.

Images & mediaChatGPT · Gemini
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S02 · Reddit
    It’s incredibly disappointing ↗
    Gemini · Publication date not recorded

    The assistant changes content beyond the requested edit.

    Reported; current resolution not established
  • S35 · OpenAI Community
    Add items to existing image, do not replace ↗
    ChatGPT · Source date 2024-04-05

    A request to add objects instead produces a new picture.

    Replies suggest inpainting as a workaround; results vary.
#24

Makes arithmetic errors even after checking

2 threads

Calculations or totals remain wrong after the user asks for verification.

Answer qualityChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S46 · OpenAI Community
    ChatGPT calculations are wrong ↗
    ChatGPT · Source date 2023-01-16

    Users report wrong sums and derived statistics.

    Reported; current resolution not established
  • S47 · OpenAI Community
    How come Chat cannot do simple math? ↗
    ChatGPT · Source date 2025-01-13

    A business-plan total remains wrong after double-checking.

    Reported; current resolution not established
#24

Misreads or omits uploaded content

2 threads

Answers skip or distort information present in source documents.

Files & researchChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Misses exact word or character counts

2 threads

Output length and reported counts fail explicit numeric constraints.

Answer qualityChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S38 · OpenAI Community
    Cannot count words or produce word-count-limited text ↗
    ChatGPT · Publication date not recorded

    Users report inaccurate word counts and length targets.

    Reported; current resolution not established
  • S39 · OpenAI Community
    4o model does not follow the prompt ↗
    ChatGPT · Source date 2025-05-14

    Social captions miss the specified word ranges.

    Reported; current resolution not established · Findings checked in the indexed first-person excerpt; full-page fetch failed during review.
#24

Paid plans still interrupt work with caps

2 threads

Subscription users hit usage ceilings during their normal work.

Cost & limitsClaude Code · Copilot
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Speech is transcribed in the wrong language

2 threads

Dictation is translated or assigned an unintended language.

Voice & accessibilityChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Starts image generation when text was requested

2 threads

Mentioning an image triggers generation instead of a text-only response.

Images & mediaGemini
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#24

Upgrade or subscription causes duplicate charges

2 threads

Users report overlapping subscriptions or repeated billing.

Cost & limitsChatGPT
Read 2 sources

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

A useful manual workflow disappears

1 thread

A removed interaction mode forces users to change how they work.

Workflow & controlCursor
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S55 · Cursor Community
    Manual mode missing ↗
    Cursor · Source date 2025-07-23

    A user asks why Manual mode was removed.

    Team reply confirms removal and explains changed default context behavior.
#39

Adds unwanted follow-up questions

1 thread

Conversational questions continue despite requests to stop asking them.

Conversation & toneChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Background sound disrupts voice conversations

1 thread

Ambient audio can trigger unwanted interruptions or input.

Voice & accessibilityChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Cannot apply a code suggestion to the editor

1 thread

The expected control for applying generated code is absent or fails.

Coding & agentsCursor
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Cannot conveniently share project chats

1 thread

Project conversations require manual copying instead of the desired share flow.

Workflow & controlChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Cannot open the sources needed for research

1 thread

Web tools fail when the user needs access to supporting pages.

Files & researchChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Compaction loses useful working state

1 thread

Summarizing a long session discards details needed to continue.

Memory & continuityClaude Code
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Connector permissions are too broad

1 thread

Users want to restrict connected document access to selected folders or files.

Workflow & controlChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S44 · OpenAI Community
    Granular Google Drive Permissions ↗
    ChatGPT · Publication date not recorded

    The author wants selected folder or file access instead of broad Drive access.

    Later support reply describes some workspace-admin scoping, not a personal folder-only connection.
#39

Copying a table loses its formatting

1 thread

A useful table is difficult to transfer into another document.

Files & researchRaycast AI
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S60 · Reddit
    Copying table from AI chat ↗
    Raycast AI · Publication date not recorded

    Table formatting is lost when copying it elsewhere.

    Reported; current resolution not established
#39

Copying text corrupts non-English characters

1 thread

Copied output becomes garbled in other applications.

Files & researchChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Deletes files without the requested scope

1 thread

Reported agent actions remove user files beyond the intended task.

Coding & agentsClaude Code
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Edits the wrong image or returns it unchanged

1 thread

The output uses an old reference or shows no requested edit.

Images & mediaGemini
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Failed generations consume allowance

1 thread

Users report spending generation credits on outputs that fail the request.

Cost & limitsGemini
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S02 · Reddit
    It’s incredibly disappointing ↗
    Gemini · Publication date not recorded

    Unsuccessful attempts reportedly use Pro allowance.

    Reported; current resolution not established
#39

Generates broken code

1 thread

Suggested code fails to run or does not fit the existing project.

Coding & agentsChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Gives excessively long answers

1 thread

Long explanations bury the requested answer or crowd out useful context.

Conversation & toneChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

High memory use crashes the app or browser

1 thread

Users report excessive RAM consumption and out-of-memory crashes.

Reliability & accessChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S17 · OpenAI Community
    ChatGPT 4o bug — keeps on crashing ↗
    ChatGPT · Publication date not recorded

    Replies describe high RAM use and out-of-memory crashes.

    Reported; current resolution not established
#39

Image quality or visual diversity declines

1 thread

Users report repeated compositions, artifacts or reduced realism.

Images & mediaChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Interface updates disrupt familiar workflows

1 thread

Frequent layout or mode changes make routine work harder.

Workflow & controlCursor
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S53 · Cursor Community
    Cursor layout and UI feedback — page 4 ↗
    Cursor · Source date 2025-12-16

    Users object to repeated layout changes.

    A team reply says UI changes are being paused; this is historical feedback.
#39

Invents functions or dependencies

1 thread

Code references APIs, modules or parameters that do not exist.

Coding & agentsChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S04 · OpenAI Community
    Catastrophic Failures of ChatGpt — discussion page 3 ↗
    ChatGPT · Source date 2025-06-20

    A developer reports invented modules, functions and parameters.

    Reported; current resolution not established · Findings refer to visible replies on this page, not the original topic date.
#39

Misspells text placed inside images

1 thread

Words embedded in generated artwork do not match the supplied text.

Images & mediaChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S41 · OpenAI Community
    DALL-E text in image creation is misspelled ↗
    ChatGPT · Publication date not recorded

    Users need accurate words inside generated images.

    Historical DALL-E discussion; replies distinguish newer image generation and manual editing workarounds.
#39

Model choice changes or is unclear

1 thread

Users report unexpected model changes or uncertainty about the active model.

Workflow & controlChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S49 · Reddit
    Model Swapping ↗
    ChatGPT · Publication date not recorded

    The author reports unexpected changes to the chosen model.

    Reported; current resolution not established · Perceived model switching is reported by the user; backend routing has not been independently verified.
#39

Printed transcripts omit part of the chat

1 thread

Saving a conversation as a PDF loses earlier messages.

Files & researchChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S57 · OpenAI Community
    Session Transcript Printing/Saving ↗
    ChatGPT · Publication date not recorded

    A reply reproduces missing earlier messages in a printed PDF.

    Reported; current resolution not established
#39

Promises background work that never arrives

1 thread

The assistant describes future work without delivering a finished file.

Workflow & controlChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Repeated permission prompts block progress

1 thread

An approval loop leaves the agent waiting rather than completing the task.

Coding & agentsClaude Code
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Saved memory fills up too quickly

1 thread

Users repeatedly clear saved memories to continue using personalization.

Memory & continuityChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Sounds patronizing or moralizing

1 thread

Answers lecture or scold rather than engage with the intended question.

Conversation & toneChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Switching models breaks the interface

1 thread

Changing the selected model causes loading errors or interrupts access.

Reliability & accessChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S48 · OpenAI Community
    Content failed to load when switching models ↗
    ChatGPT · Source date 2026-04-19

    Selecting another model causes a loading failure.

    Later replies report the loading bug fixed; action behavior remains a separate complaint.
#39

Undo and checkpoint controls are hard to find

1 thread

After an unwanted edit, users struggle to locate a way to restore their work.

Coding & agentsCursor
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

  • S52 · Cursor Community
    Reverting back to previous code? ↗
    Cursor · Source date 2025-03-05

    The author initially cannot find a restore control.

    Reporter later found Restore last checkpoint and reported it helped.
#39

Unlabeled controls hinder screen-reader use

1 thread

Blind users cannot reliably identify and operate interface buttons.

Voice & accessibilityChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Updated knowledge files are ignored

1 thread

Replacing source files does not reliably update the assistant’s answers.

Files & researchChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Uses outdated technical information

1 thread

Answers rely on obsolete software behavior instead of current sources.

Files & researchChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

#39

Voice mode stops listening after interruption

1 thread

Stopping a spoken response leaves voice input unresponsive.

Voice & accessibilityChatGPT
Read 1 source

Paraphrased user reports. Interpretations and suspected causes have not been independently verified.

Agent Portal