ChatGPT custom instructions for QA received a practical capacity upgrade on July 15, 2026. OpenAI says Pro, Enterprise, Business and Education users can now save up to 5,000 characters, up from 1,500.
What changed
The release is a limit increase, not a new test-generation model. It gives eligible users more room to define response style and behavior in persistent custom instructions. OpenAI’s help documentation says custom instructions are applied across chats and can be edited or removed for future conversations.
Why this matters for QA engineers
The larger limit can hold a compact, reusable QA review policy instead of forcing teams to repeat the same expectations in every prompt. A tester could specify rules such as:
- Separate confirmed behavior from assumptions and missing evidence.
- Include positive, negative and boundary cases when suggesting tests.
- Prefer stable user-facing locators and meaningful business assertions.
- Never claim a test passed unless execution evidence was provided.
That can make initial drafts more consistent across bug triage, test-case design and automation review. It does not validate product behavior: QA engineers still need to inspect source requirements, run tests and verify evidence.
A sensible rollout
Start with one short team-approved instruction set and compare its output against your existing review checklist. Keep secrets, customer data and environment credentials out of the instructions. For managed workspaces, confirm that the guidance matches organizational policy before standardizing it.
Sources
- OpenAI: ChatGPT release notes – July 15, 2026.
- OpenAI: Custom Instructions for ChatGPT – accessed July 20, 2026.