ChatGPT Just Gave You 5,000 Characters of Context. Here's What to Actually Write in Them.

Every ChatGPT conversation used to start cold. You'd explain your job, your audience, your tone, what not to do — and the next session, you'd do it again. The custom instructions field existed to fix this, but 1,500 characters left room for maybe three sentences. On July 15, OpenAI expanded the limit to 5,000 characters for all paid users. That's enough for a real briefing document.

What changed and why it matters

Custom instructions apply to every ChatGPT conversation automatically. They sit before the conversation starts — more persistent than a system prompt you type each session, more flexible than a GPT you have to remember to open. The expansion didn't change what the field does; it changed what you can actually fit in it.

The previous 1,500-character limit was enough for a few sentences if you were disciplined. The new 5,000-character limit is enough for a structured brief: who you are, what kind of work you're doing, how you want responses framed, what formats you use, and what to skip. The difference between those two lengths is the difference between a sticky note and a one-pager. A sticky note is a reminder. A one-pager is a brief. A brief gives the model enough context to make good default decisions without being told every session.

What to do differently on Monday

Most people re-brief ChatGPT on every conversation. They open a new chat with "I'm a marketing manager at a software company, and I need..." — and type the same sentence again tomorrow. Custom instructions exist to absorb that overhead, but only if you write them like a brief, not a preference setting.

"Be concise" is a preference. It doesn't tell ChatGPT what you work on, who you're writing for, or what a good output looks like in your context. "I write for sales directors at mid-market SaaS companies who scan documents before they read them — lead with the number or the decision, not the rationale" is a brief. The second version makes every response better without you saying another word.

A content strategist at a 60-person fintech startup

She writes for compliance-aware financial professionals — readers who know the technical terms, don't need them explained, and will stop reading if the tone is generic. Her previous custom instructions: "Write clearly and avoid jargon."

With 5,000 characters, she rewrites them as a proper brief: her role, her primary audience, two sentences about the company's product and what problem it solves, a list of terms her readers use versus the jargon they don't, her most common deliverable types (short-form email, one-page summary, LinkedIn post), and a note that she usually pastes a rough draft for refinement rather than asking for a first pass cold.

The next conversation she opens doesn't start cold. She pastes a rough email, types "tighten this," and gets a revision that already knows who it's written for. The brief did the setup work once.

A project manager at a 200-person professional services firm

He runs three simultaneous client engagements and uses ChatGPT mostly for status updates, risk logs, and client-facing summaries. His custom instructions used to say "be professional and structured."

He rewrites them with specifics: his role, the three document types he produces most, the fact that his clients want tables and bullets over narrative prose, the risk categories he tracks, and a line that every status update must end with a sentence stating what action is required and by whom.

Every summary now arrives pre-formatted for how his clients read. He no longer rewrites the structure from scratch — only the content changes. The parts that were fixed (format, voice, ending) stay fixed in the brief.

The one thing

5,000 characters is enough to stop explaining yourself every session. Write the briefing document once — your role, your audience, your format, and one firm constraint — and let every conversation inherit it.

<BlogPracticeSection />