A Custom GPT is a saved ChatGPT configuration that bundles a name, system Instructions, optional Knowledge files, and a checklist of enabled Capabilities into a reusable assistant. Build one by clicking + Create on the Explore GPTs page, filling the Configure tab (Name, Description, Instructions), attaching Knowledge files, ticking Capabilities (Web Search, Canvas, Image Generation, Code Interpreter), then choosing visibility: Only me, Anyone with the link, or GPT Store. Paid plan required. Verified May 2026.

How to Create Custom GPTs Step-by-Step: 2026 Walkthrough

Custom GPTs turn a long system prompt and a stack of reference files into a single clickable assistant. Instead of pasting your brand voice, output format, and source documents at the start of every chat, you bake them into a Custom GPT once and call the GPT directly. This guide walks every step of the current GPT Builder with screenshots from May 2026, flags the paid-plan requirement upfront, and shows where AI Toolbox (formerly ChatGPT Toolbox) complements Custom GPTs with a prompt library you can paste into the Instructions field via the // shortcut, folders to organize the chats your GPTs generate, and full-text search across every Custom GPT conversation. With 20,000+ active users and a 4.5/5 Chrome Web Store rating, AI Toolbox is the most installed productivity layer for ChatGPT.

What Is a Custom GPT?

A Custom GPT is a saved package of Name, Description, Instructions, Knowledge files, Capabilities, and Actions that behaves like a dedicated assistant for one purpose. Once you save it, the GPT appears in My GPTs and you can open it from the sidebar or the Explore GPTs page. Every conversation with that GPT runs with the same system prompt and the same toolset, so you do not have to re-paste instructions or re-attach files for every new chat.

Custom GPTs are the right surface for repeatable work where the inputs change but the framing stays the same: an outline generator that always returns the same H2 structure, a code reviewer that always uses your team's style guide, a translation assistant scoped to one industry. For one-off prompts, regular chat is faster.

Plan Requirement: Creating a GPT Needs a Paid Plan

Creating Custom GPTs requires a paid ChatGPT plan: Plus, Pro, Team, or Enterprise. Free accounts can browse the GPT Store and use Custom GPTs that other people have published, but cannot open the GPT Builder or save a new one. Publishing publicly to the GPT Store also requires a verified builder profile. If you only need to use someone else's Custom GPT, free is fine. If you want to build your own, you need to upgrade.

For details on the rest of the plan ladder, see the ChatGPT limits guide. The break-even for the Plus upgrade is usually 2-3 GPTs you would actually build, since each one removes a recurring "paste the system prompt + attach the files" loop from your day.

Step 1: Open the GPT Builder from + Create

Click Explore GPTs in the ChatGPT sidebar to open the GPTs discovery page, then click the black + Create button in the top-right of the page header. The page header reads "GPTs", with a "My GPTs" link and the "+ Create" button on the right. The + Create button is what launches the GPT Builder. The Builder opens in a split view: the Create / Configure tabs on the left for editing the GPT, and a live Preview pane on the right where you can chat with the GPT as you build it.

The Create tab is the conversational builder where you can describe what you want and ChatGPT generates a starting config for you. The Configure tab is the manual builder with discrete fields for Name, Description, Instructions, Knowledge, Capabilities, and Actions. For precise control, use Configure directly. For a quick draft, Create can scaffold something to refine in Configure.

Step 2: Configure Name, Description, Instructions

In the Configure tab, fill three top fields: Name (the label users see), Description (one-line summary), and Instructions (the full system prompt). The Name shows in My GPTs and at the top of every chat. The Description shows under the Name in lists and search results, so make it scannable. The Instructions field is the system prompt that runs at the start of every conversation with this GPT, so spell out exactly what it does, what tone and format to use, what data to prioritise, and what to avoid.

Write Instructions prescriptively. Bad: "Be helpful and friendly." Good: "For every outline request, return: 1) Proposed H1, 2) Meta description under 160 characters, 3) 5-7 H2 sections each with a one-sentence summary, 4) Three FAQ questions to answer. Always include the year 2026 in titles. Never use em dashes as punctuation." Anything you would otherwise paste at the start of every chat belongs in Instructions. To reuse the same instructions across many GPTs (or to keep a master copy outside the Builder), save them as a prompt template in the ChatGPT module of AI Toolbox and paste with the // shortcut.

Save the prompts behind your best Custom GPTs so you never lose them. AI Toolbox adds folders, full-text search across every chat (including the chats your Custom GPTs generate), a prompt library with the // shortcut for pasting GPT Instructions verbatim, and prompt chaining for multi-step workflows. Free plan covers 2 folders, 2 pins, and 2 saved prompts. Unlock unlimited everything on Premium ($9.99 per month) or pay once for Premium Lifetime ($99). Working across ChatGPT, Gemini, and Claude? All Access Lifetime ($149) covers all three modules and saves $148 versus three single-platform Lifetimes ($99 × 3 = $297). Add AI Toolbox to Chrome free →

Step 3: Pick Knowledge Files and Capabilities

Scroll past Conversation starters to Knowledge (Upload files) and Capabilities (Web Search, Canvas, Image Generation, Code Interpreter and Data Analysis). Knowledge attaches reference files that the GPT can pull from in every conversation. Anyone who uses the GPT benefits from the same knowledge base without re-uploading, so this is where brand voice docs, style guides, product specs, and FAQ libraries belong. Keep file sizes reasonable for fast retrieval.

The Capabilities checklist is on by default for Web Search, Canvas, and Image Generation; Code Interpreter and Data Analysis is off by default. Enable only what your GPT needs. Each enabled capability adds latency and broadens what the model can do, so a code-only GPT with Web Search disabled stays focused, and a writing GPT with Code Interpreter off avoids unnecessary tool calls. Below Capabilities, the Actions section lets you wire up custom JSON-schema API calls (an OpenAPI spec for the external service). Skip Actions for a first GPT; come back to it once the basics are working.

Step 4: Share, or Keep It Private

Click Save (or Update, when editing) in the top-right and pick one of three visibility options: Only me, Anyone with the link, or GPT Store. The Share GPT modal shows the three options as rows with icons, and a selection dot on the right. Only me keeps the GPT private to your account, useful while you iterate. Anyone with the link makes it shareable via URL but invisible to discovery, useful for team or client handoff. GPT Store publishes it publicly to OpenAI's directory where anyone can find and use it; this option also requires that your builder profile is verified.

Pick Only me while testing instruction wording or refining the Knowledge base. Switch to Anyone with the link the moment you hand a draft to a collaborator. Reserve GPT Store for the final polished release once the GPT has earned its public listing through repeated use. You can change visibility later from My GPTs, so the first choice is not permanent.

Editing a Custom GPT After You Create It

Open My GPTs from the Explore GPTs page, click the GPT you want to change, then click the pencil icon to re-open the GPT Builder. Any field is editable: Name, Description, Instructions, Knowledge, Capabilities, Actions, visibility. Click Update in the top-right to save. Existing conversations with the GPT continue to work, and the next time someone opens it, they get the new version.

Common edits worth scheduling: tighten Instructions whenever a recurring output goes wrong, swap stale Knowledge files when source docs change, toggle Capabilities on or off as the GPT's scope shifts (for example, turn Code Interpreter on once you start asking it to draft scripts). Custom GPTs improve a lot in the first 2-3 edits as you watch real users hit edge cases; lock in those iterations early.

Get More From Custom GPTs with AI Toolbox

Custom GPTs handle the per-GPT system prompt and toolset. What they do not do is help you search across every Custom GPT conversation, save the prompts that produced your best Custom GPT chats for reuse in fresh chats, or group GPT-generated chats into folders. That is what AI Toolbox (formerly ChatGPT Toolbox) adds on top.

• Full-text search across every conversation, including chats from every Custom GPT you own or have used. Use Cmd+Shift+F (Ctrl+Shift+F on Windows) to search by message content, role, or date range with exact-match toggle.
• Folders and subfolders for grouping chats by GPT, by client, or by stage of work. Folders use search-based adding: you search the conversation list and add matches to the folder, no drag-and-drop required. Free plan supports 2 folders. Unlimited on Premium.
• Prompt library with the // shortcut. Save your Custom GPT Instructions as a reusable template with {{placeholder}} variables, then paste into a fresh chat to test wording before committing changes to the GPT. Free plan: 2 saved prompts. Unlimited on Premium.
• Prompt chaining with the .. shortcut. Chain up to 10 prompts that mimic a multi-step Custom GPT workflow without needing to build a full GPT for every variation.

AI Toolbox is also available for Gemini and Claude as part of the same unified install. The Gemini module adds full-text search across every synced Gemini conversation with date, role, and exact-match filters. The Claude module adds full-text search with exact-match toggle and message-level bookmarks. Pricing details are on the AI Toolbox pricing page.

Frequently Asked Questions

What is a Custom GPT?

A Custom GPT is a packaged version of ChatGPT that bundles a name, description, system instructions, optional knowledge files, and a set of enabled capabilities (Web Search, Canvas, Image Generation, Code Interpreter, custom Actions). You build one through the GPT Builder, and once saved, it appears in My GPTs and behaves like a dedicated assistant for that purpose without you re-pasting the instructions every time.

How do I create a Custom GPT?

Open the Explore GPTs page from the ChatGPT sidebar, click + Create in the top-right to launch the GPT Builder, switch to the Configure tab, and fill in Name, Description, and Instructions. Add reference files under Knowledge, pick which Capabilities to enable, then click Create at the top-right to save.

Do I need to pay to make a Custom GPT?

Yes. Creating Custom GPTs requires a paid ChatGPT plan: Plus, Pro, Team, or Enterprise. Free accounts can browse and use Custom GPTs that other people have published, but cannot open the GPT Builder or save a new one. Publishing to the public GPT Store also requires a verified builder profile.

What goes in the Instructions field of a Custom GPT?

Instructions is the system prompt for the GPT. Spell out what it does, what tone and format to use, what data to prioritize, and what to avoid. Aim for prescriptive language and concrete examples instead of vague directives. Anything you would otherwise paste at the start of every chat belongs here, since it applies automatically to every conversation with the GPT.

What are the Capabilities I can enable on a Custom GPT?

The Capabilities checklist in Configure exposes Web Search, Canvas, Image Generation, and Code Interpreter and Data Analysis. Enable only the ones your GPT actually needs, since each one adds latency and broadens what the model can do. Custom Actions are listed below Capabilities and let the GPT call third-party APIs via JSON schemas you define.

What does each visibility option on a Custom GPT mean?

Only me keeps the GPT private to your account. Anyone with the link makes it shareable via URL but invisible to discovery. GPT Store publishes it publicly to OpenAI's directory where anyone can find and use it. Pick Only me while iterating, switch to Anyone with the link for team use, and reserve GPT Store for polished public releases.

Can I add files to a Custom GPT for it to reference?

Yes. The Knowledge section in Configure has an Upload files button that attaches reference documents to the GPT. The GPT can pull from these files in every conversation. Files persist with the GPT, so anyone who uses the GPT can benefit from the same knowledge base without re-uploading. Keep file sizes reasonable for fast retrieval.

How do I edit a Custom GPT after creating it?

Open My GPTs from the Explore GPTs page, click the GPT you want to change, then click the pencil icon to re-open the GPT Builder. Edit any field (Name, Description, Instructions, Knowledge, Capabilities, Actions, visibility) and click Update in the top-right to save. Existing conversations with the GPT continue to work after the edit.

Conclusion

Custom GPTs collapse "paste the system prompt + attach the files + remember the tone" into a single saved assistant. Open the GPT Builder from + Create, use Configure for precise control, write prescriptive Instructions, attach Knowledge files for repeated context, pick only the Capabilities you actually need, and pick visibility based on who should reach the GPT. Iterate by re-opening from My GPTs. Free accounts can use existing Custom GPTs but need to upgrade to Plus or higher to build their own. To save the Instructions behind your best Custom GPTs as reusable prompts, search across every Custom GPT chat, and organise GPT-generated conversations into folders, install AI Toolbox for Chrome: free plan covers 2 folders, 2 pins, 2 saved prompts, and full-text search with up to 5 results per query. Premium ($9.99 per month) or Premium Lifetime ($99 one-time) unlocks unlimited everything plus bulk export.

References and Further Reading

• OpenAI Help Center: GPTs FAQ (Retrieved May 2026)
• OpenAI ChatGPT Release Notes (Retrieved May 2026)
• How to Use ChatGPT Projects (2026) — Projects vs Custom GPTs, when to use each
• How to Use ChatGPT Tasks (2026) — pair a Custom GPT with a Task to run it on cadence
• How to Use ChatGPT Canvas (2026) — Canvas is one of the Capabilities you can enable on a Custom GPT
• ChatGPT Models Explained (2026) — which GPT-5.x models run inside Custom GPTs
• How to Search ChatGPT History — full-text search across Custom GPT chats
• How to Export ChatGPT Conversations — bulk export of Custom GPT outputs

Last updated: May 31, 2026