ChatGPT Dots vs Custom GPTs vs Projects (OpenAI Dots)

Snapshot

ChatGPT Dots (OpenAI Dots), Custom GPTs, and ChatGPT Projects all live inside ChatGPT—but they solve different problems. A Custom GPT is a configured assistant you open and talk to. A Project is a workspace that keeps related chats, files, and instructions together. A Dot is an always-on agent with its own cloud computer that keeps work moving between conversations and asks when a decision needs you.

OpenAI is also retiring Custom GPTs in favor of plugins (Enterprise timeline published through Dec 11, 2026; other plans follow in-product notices). Migrating a GPT to a plugin is not the same as adopting a Dot. For other comparisons, see Dots vs Claude · Dots vs Cursor Projects · Dots vs Grok Bot · Dots vs Codex vs ChatGPT Work. Dots setup: getting started · features · what is Dots.

Product framing

• ChatGPT Dots — Always-on personal agent in ChatGPT (GPT-6 Astra). Per-Dot cloud computer and browser; works toward goals between chats; plugins into thousands of apps; message or call in ChatGPT; also Slack and Microsoft Teams; optional connect of one personal computer; can start or manage Work and Codex tasks. Conversations with your Dot do not count toward ChatGPT usage limits; Work/Codex metering still applies when those tasks run.
• Custom GPTs — Saved ChatGPT configurations: instructions, knowledge files, conversation starters, and optional apps/actions. You open the GPT and ask; it responds in that chat. It does not keep running autonomous cross-app work after you leave. New GPT creation/publishing is already unavailable on personal Free/Go/Plus/Pro accounts; Business/Enterprise/Edu can still create when workspace settings allow. OpenAI plans to retire custom GPTs and migrate creators to plugins.
• ChatGPT Projects — Smart workspaces for ongoing work: group chats, upload reference files, add project instructions, optional app links (for example Drive/Slack sources). Project memory stays anchored to that project. Available to signed-in users (plan/workspace gates apply); Business/Enterprise/Edu can share with teammates. Projects are staying; they are the common “keep context together” home while GPTs retire.

Indie explainers (not official docs) often put it as: GPT answers when you chat; Project holds the filing cabinet; Dot keeps executing. Treat that as framing only—facts below follow OpenAI Help and Learn.

Nature of the work

• Custom GPT / plugin successor: Best when the job is respond on demand—FAQ tone, document Q&A, reusable review checklist, a packaged assistant others open by link or Store. You start each conversation; it does not chase deadlines while you sleep.
• ChatGPT Project: Best when the job is accumulate context—client folder, research thread, proposal library. Chats, files, and instructions live together so you (or teammates) pick up where you left off. Tools in project chats match regular ChatGPT plus connected apps; this is still chat-driven work, not a managed always-on agent.
• Dot: Best when the job is delegate continuity—recurring follow-ups, inbox/calendar watch, parallel background agents, Slack/Teams updates, optional Work/Codex handoffs. You define responsibility, sources, and permissions; the Dot tracks progress and brings you results or decisions.

If your Custom GPT only answers well, OpenAI’s path is migrate to a plugin (and/or move instructions/files into a Project)—not automatically create a Dot. Indie guides (Hora de Codar, PCWorld) make the same split: retirement migration ≠ always-on agent adoption.

Continuity, memory, and channels

• Dots: Continuity between conversations is the product default. Memory/notes carry across ChatGPT, Slack, and Teams while messages stay in their channels. Activity View for delegated tasks; Custom Rules and built-in review for approvals; Pause / Scheduled controls.
• Custom GPTs: Each chat starts from the GPT’s instructions and knowledge—not ongoing autonomous goals. Existing GPT conversations remain accessible per Help through retirement; chats created with a GPT generally cannot be moved into a Project.
• Projects: Project memory draws from chats and files inside that project. Move eligible chats into a project; save useful responses as project sources. Sharing creates a live context hub for a team when your plan supports it.

Computers, apps, and controls

• Dots: Own cloud computer/browser; optional local computer connection; ChatGPT plugins with existing permissions; private website sign-in / Take over; built-in safeguards plus optional Custom Rules (allow / require approval / forbid). See integrations and Help privacy FAQs.
• Custom GPTs: Tools historically included web search, canvas, code interpreter, and apps or custom actions (with plan limits). Custom actions do not transfer automatically in the planned GPT→plugin migration—rebuild with a supported connector or MCP where needed.
• Projects: Use the same ChatGPT tools and connected apps inside project chats; add Drive/Slack links as sources where supported. No separate “cloud computer agent” product surface—Projects organize work; they do not replace Dots.

Access, retirement, and pricing notes

• Dots access: Rolling out to eligible Pro tiers (region/age gates apply; published Help notes Pro exclusions for EEA/Switzerland/UK), Business Premium, and Enterprise when an admin enables Use dots (Beta). First Dot included on eligible plans; deeper-work allowances apply (extended limits called out for the first month after launch). Details: pricing · availability · EU/UK note.
• Custom GPT retirement (Help): Retirement scheduled Dec 11, 2026 (Feb 11, 2027 only for Enterprise workspaces with an approved deferral). Migration experience targeted around mid-Sep 2026 and may not appear for every account at once. After migration, the original GPT stays usable until retirement but becomes read-only. Personal accounts: new GPT create/publish already off; existing GPTs remain usable/editable under existing permissions until retirement.
• Projects: Available to signed-in ChatGPT users globally subject to plan/workspace settings—no “always-on agent” eligibility gate like Dots.

Re-check Help notices in your own account before migrating or buying; this page is a map, not a substitute for OpenAI’s current plan pages.

When to pick which

• Pick a Custom GPT / plugin when you need a reusable, shareable assistant that answers with fixed instructions and files—and follow OpenAI’s migration path before retirement.
• Pick a ChatGPT Project when you need a durable workspace for ongoing chats + files + instructions (solo or shared), including as a landing place for instructions/files after GPT retirement.
• Pick ChatGPT Dots (OpenAI Dots) when you need a managed always-on agent that keeps responsibilities moving across apps and channels—and your plan/region is eligible.
• Use Project + Dot together when that matches your stack: Project for the living context hub; Dot for delegated follow-through. Use plugin + Project when you only needed a configured assistant, not an always-on agent.

Related: Dots vs Claude · Dots vs Cursor Projects · Dots vs Grok Bot · Dots vs Meta Muse · Dots vs Codex vs ChatGPT Work.

Sources

OpenAI: Introducing dots, Meet dots (Learn), Getting started / Help, Dots feature page, computers/apps · channels · controls · tasks Learn pages, Plugins Learn, Custom GPT retirement FAQ, Creating and editing GPTs, Projects in ChatGPT, Work/Codex usage notes, Pro tiers. Indie framing only (label as non-official): Hora de Codar “Dots vs GPTs,” PCWorld on Custom GPT retirement / Projects alternative. Re-check before buying, migrating, or rolling out.