Persistent custom instructions
DocumentedUser-authored preferences can apply across ChatGPT conversations.
View source detailAgent intelligence
ChatGPT
Official ChatGPT evidence covers model selection, custom instructions, projects, memory, and tools without establishing a published full system prompt.
ChatGPT is documented as a multi-surface product with selectable models, cross-conversation custom instructions, project-scoped context, controllable memory, and product tools. The selected official sources do not publish a complete current base system prompt or one universal runtime configuration.
Scope: ChatGPT web · ChatGPT apps · Selectable OpenAI models · reviewed Aug 30, 2026
Reviewed product surfaces
Product Surface Schematic
Each layer links to the detailed evidence map. Its state applies only to the reviewed evidence.
Reviewed surfaces ChatGPT web · ChatGPT apps
- Base product promptNot established
- Custom instructionsDocumented
- Project instructionsDocumented
- MemoryDocumented
- Tools and modesDocumented
- Not established
- Documented
Source-backed answers
Key behaviors
Each answer is scoped to the selected evidence. “Not established” means the evidence does not support the claim; it does not assert absence.
Project instructions and context
DocumentedProjects can carry instructions alongside chats and uploaded files.
View source detailMemory / persistent guidance
DocumentedMemory is separately controllable; Temporary Chat starts non-personalized but can optionally use existing memories without creating new ones.
View source detailTools and modes
DocumentedSearch, research, files, data analysis, images, voice, and canvas are documented product capabilities.
View source detailBase prompt visibility
Not establishedThe selected official sources do not publish a complete current ChatGPT base system prompt.
Review evidence scopeEditorial guidance
What this means in practice
These conservative editorial implications come from the linked evidence. They add no vendor claims or generic prompt advice.
Editorial implication
Choose the instruction scope that matches the work.
Custom instructions can apply across conversations, while Projects combine project-specific instructions, chats, and uploaded files.
Supporting evidenceEditorial implication
Treat memory as a separate, user-controlled context mechanism.
Temporary Chat starts non-personalized, but optional personalization can use existing memories, custom instructions, and plugins without creating new memories. Saving converts it into a regular chat governed by account-level settings.
Supporting evidenceEditorial implication
Do not assume one fixed tool or prompt configuration.
Models, tools, modes, plans, and surfaces can vary, and the selected evidence does not establish the complete base prompt.
Supporting evidenceUseful next actions
Continue your research
Compare agent behaviorsReview the same evidence-backed questions across all six products.
How does Claude publish product-prompt history?Compare ChatGPT's documentation evidence with Anthropic's dated Claude product-prompt publications.
Why is there no ChatGPT prompt route?Read the source, completeness, and publication gates used across Promptcred.
Review source and provenance detailsVerify origin, retrieval, completeness, publication limits, and rights.
Product scope
ChatGPT is the product; its selected model and enabled capabilities are separate context.
- Organization
- OpenAI
- Product
- Multi-surface conversational AI product
- Surface
- ChatGPT web · ChatGPT appsThe browser-based ChatGPT product surface.
- Model context
- Selectable OpenAI modelsOfficial product documentation treats model choice as separate from the ChatGPT product; availability can vary by plan and surface.
Evidence status
Coverage and review
Official OpenAI documentation covers product capabilities, user-facing context layers, and current Temporary Chat controls. The review does not claim a complete current ChatGPT system prompt.
- Coverage
- partial
- Primary source
- Tier A · vendor-published
- Review freshness
- currentOfficial OpenAI documentation was reviewed for capabilities, custom instructions, projects, memory, and current Temporary Chat personalization controls. Targeted research planning evidence was consulted on September 14, 2026; see the separately dated sources. Other evidence was not re-reviewed.
- Reviewed
- Aug 30, 2026
Detailed instruction layers
Instruction-system map
The map separates the unestablished base prompt from custom instructions, project context, memory, model selection, and product tools.
Base product promptnot established
The selected official sources do not publish a complete current ChatGPT base system prompt.
Source detailCustom instructionsdocumented
User-authored preferences can apply across ChatGPT conversations.
Source detailProject instructionsdocumented
Projects can carry instructions alongside chats and uploaded files.
Source detailMemorydocumented
Memory is separately controllable; Temporary Chat starts non-personalized but can optionally use existing memories without creating new ones.
Source detailTools and modesdocumented
Search, research, files, data analysis, images, voice, and canvas are documented product capabilities.
Source detailSupporting instruction evidence
Documentation and source coverage
Official sources document the product but do not support a standalone public system-prompt page.
- Selected snapshot
- ChatGPT instruction architecture · reviewed August 30, 2026
- Evidence coverage
- Through Aug 30, 2026ChatGPT instruction architecture · reviewed August 30, 2026
- Latest known vendor version
- Not established by selected evidence
- Current product state
- Not established by selected evidence
- Publication
- Full prompt unavailable
- Instruction page
- Not availableNo verified current full first-party ChatGPT prompt artifact was established in the selected official sources.
Evidence-backed capabilities
Capabilities with evidence
Each capability links to its supporting prompt section or first-party source and keeps its stated confidence level.
| Capability | What the source establishes | Confidence | Evidence |
|---|---|---|---|
| Research planning | Deep research in Chat proposes a research plan that the user can review and modify before research. Availability and usage depend on plan and workspace controls. This does not establish a general coding-agent Plan mode or identical behavior in Work and Codex. | high | Primary source |
| Selectable models | Official documentation distinguishes the ChatGPT product from the models paid users can select. | high | Primary source |
| Custom instructions | Users can provide persistent response preferences through product settings. | high | Primary source |
| Project context | Projects organize chats, files, and project-specific instructions. | high | Primary source |
| Memory controls | Memory can be managed separately. Temporary Chat starts non-personalized, can optionally use existing memories and other personalization inputs, and does not create new memories while temporary. | high | Primary source |
| Product tools and modes | Search, deep research, images, files, data analysis, voice, and canvas are documented capabilities. | high | Primary source |
Version evidence
Snapshots and meaningful changes
The selected evidence establishes no compatible first-party full-prompt snapshots. Help-center updates are not treated as prompt versions.
- selected ChatGPT instruction architecture · reviewed August 30, 2026Reviewed Aug 30, 2026Documentation coverage
Provenance
Source ledger
Every source lists its origin, retrieval date, completeness, publication limits, and rights separately.
Tier A · vendor-published
Deep research in ChatGPT — research planning
- Publisher
- OpenAI
- Retrieved
- Sep 14, 2026
- Upstream ref
- Live official documentation consulted 2026-09-14
- Rights posture
- Original summary; no prompt text republished.
Deep research in Chat proposes a research plan that the user can review and modify before research. Availability and usage depend on plan and workspace controls. This does not establish a general coding-agent Plan mode or identical behavior in Work and Codex.
Open primary source Tier A · vendor-published
ChatGPT capabilities overview
- Publisher
- OpenAI
- Retrieved
- Aug 11, 2026
- Upstream ref
- Live official help article reviewed 2026-08-11
- Rights posture
- Original summary of official documentation; no prompt text copied.
Describes product capabilities and model selection, not one universal runtime configuration or prompt.
Open primary source Tier A · vendor-published
ChatGPT custom instructions
- Publisher
- OpenAI
- Retrieved
- Aug 11, 2026
- Upstream ref
- Live official help article reviewed 2026-08-11
- Rights posture
- Original summary of official documentation.
Documents custom instructions, not the complete product prompt or context assembly.
Open primary source Tier A · vendor-published
Projects in ChatGPT
- Publisher
- OpenAI
- Retrieved
- Aug 11, 2026
- Upstream ref
- Live official help article reviewed 2026-08-11
- Rights posture
- Original summary of official documentation.
Documents project context but not the underlying base system prompt.
Open primary source Tier A · vendor-published
ChatGPT memory FAQ
- Publisher
- OpenAI
- Retrieved
- Aug 11, 2026
- Upstream ref
- Live official help article reviewed 2026-08-11
- Rights posture
- Original summary of official documentation.
Describes user-facing memory behavior; implementation details and a complete runtime context are not established.
Open primary source Tier A · vendor-published
Temporary Chat FAQ
- Publisher
- OpenAI
- Retrieved
- Aug 30, 2026
- Upstream ref
- Live official feature FAQ reviewed 2026-08-30
- Rights posture
- Original summary of official documentation.
Documents default and optional Temporary Chat personalization, memory creation, plugins, custom instructions, and saving behavior; account and workspace restrictions still apply.
Open primary source Tier A · vendor-published
ChatGPT release notes — More controls in temporary chat
- Publisher
- OpenAI
- Retrieved
- Aug 30, 2026
- Upstream ref
- ChatGPT release note published 2026-08-27
- Rights posture
- Original summary of official documentation.
Records the August 27, 2026 rollout of optional Temporary Chat personalization and saving; it does not establish one universal ChatGPT runtime configuration.
Open primary source