ChatGPT Sites

From PRIMO.ai
Jump to navigation Jump to search

YouTube ...Google Search ...Google News ...Bing ...Reddit


ChatGPT Sites is an artificial intelligence-powered website and web application creation service developed by OpenAI. It allows users to describe a website or interactive application in natural language, have ChatGPT and Codex construct it, preview its behavior, refine the design, and publish it to an internet-accessible address.

Unlike conventional website builders that require users to assemble predefined components manually, ChatGPT Sites uses Generative AI and AI-assisted software development to interpret requirements and generate the code, interface, and application logic needed to create a functioning website.

The service supports both relatively simple websites and interactive applications. Examples include personal portfolios, business landing pages, project dashboards, calculators, resource libraries, interactive educational materials, games, and internal business tools.

A Site can also store information in a database, accept file uploads, support user authentication, and connect to approved external services. The available capabilities depend on the account, workspace configuration, supported runtime, and permissions.

ChatGPT Sites is more than a website-design assistant. It combines AI-assisted development, application hosting, deployment, persistent storage, and access management within the ChatGPT environment.

The central workflow is conversational. Users describe what they want, evaluate the generated results, request refinements, and deploy the approved application. They don't have to write the underlying code themselves, although developers can use Codex and compatible source projects for more technical control.

As of November 2026, ChatGPT Sites is in public beta. It is available on eligible paid ChatGPT plans, with feature availability, geographical restrictions, and usage limits varying by plan and workspace.

For the latest product information, see ChatGPT Sites and the official documentation.

Describe an idea, turn it into a functioning website, refine it through conversation, and publish it without managing a separate hosting environment.
ChatGPT Sites Profile (November 2026)
Attribute Details
Developer OpenAI
Product ChatGPT Sites
Product Category AI website builder and hosted web application platform
Public Beta 2026
Primary Interface Natural-language conversation
Development Technology ChatGPT and Codex
Main Functions Website generation, application development, hosting, deployment, sharing, editing
Website Types Static websites, interactive websites, lightweight web applications, games
Programming Required No, although technical users can work with source projects
Hosting Managed hosting through ChatGPT Sites
Persistent Database Supported; up to 10 GB per Site
File Storage Supported; subject to usage policies and allowances
Authentication Workspace authentication and optional Sign in with ChatGPT
Custom Domains Supported where available
External Integrations Supported plugins, APIs, and private-network tunnels where eligible
Scheduling Supported Site automations where available
Collaboration Workspace editors and controlled visitor access
Supported Plans Plus, Pro, Business, Enterprise, and eligible Edu workspaces
Free and Go Plans Not supported as Site creators
Pricing Model Plan allowances, Codex usage, and hosting resource consumption
Official Website ChatGPT Sites
Documentation ChatGPT Sites Documentation

History and Development

From AI Code Generation to Website Creation

AI-assisted web development developed alongside improvements in large language models capable of understanding programming instructions, producing source code, and explaining software behavior.

Early Generative AI tools primarily produced code snippets. Users still needed to configure development environments, assemble application components, test functionality, purchase hosting, and deploy their projects.

As AI coding systems improved, users could delegate increasingly large software development assignments. Rather than requesting individual functions, they could describe an application and have an AI agent create files, install dependencies, diagnose errors, and prepare a functioning project.

OpenAI's Codex capabilities helped make this approach practical within a broader agentic development environment. ChatGPT Sites extends that progression by making hosting and publication part of the same workflow.

Introduction of ChatGPT Sites in 2026

OpenAI publicly documented ChatGPT Sites during 2026, initially emphasizing lightweight websites and applications created with Codex for organizational use.

An early emphasis was on internal tools: operational dashboards, launch trackers, onboarding materials, project hubs, and other applications that teams could share within their workspace.

The initial model addressed a common business problem. An employee might know precisely what information or workflow a useful application should contain but lack the programming skills or infrastructure needed to create it.

ChatGPT Sites made it possible to describe that application in ordinary language and produce a hosted interface.

OpenAI subsequently expanded the service into a broader website-building platform supporting individual creators, business users, educators, and developers.

Expansion Into Public Publishing

On July 9, 2026, OpenAI announced an expanded public beta for ChatGPT Sites. Business and Enterprise customers gained additional public-sharing capabilities, while access expanded toward eligible individual and educational plans.

The public beta broadened the service beyond internal dashboards and organizational applications.

Users could create websites intended for customers, community members, prospective clients, students, or the general public.

The expanded product included controls for private access, named viewers, workspace audiences, public publishing, custom domains, and collaboration.

Some capabilities were introduced through staged rollouts, so their availability varied across accounts and regions.

Continued Development

During the second half of 2026, OpenAI continued expanding Site management, publishing controls, external viewer access, and integration capabilities.

October 2026 updates included the removal of a fixed Site-count limit for Plus and Pro users, while continuing to enforce resource and usage limits.

The service also evolved toward more sophisticated applications with persistent storage, scheduled updates, and connections to users' authorized services.

These changes demonstrate a broader shift from generating website code toward operating complete hosted applications.

How ChatGPT Sites Works

Natural-Language Development

The distinguishing feature of ChatGPT Sites is its use of ordinary language as the primary development interface.

A user can describe an application without specifying every programming language, component, or infrastructure service.

For example:

Build a website for a neighborhood club. Members should be able to view upcoming events, register for activities, and see announcements. Administrators should be able to manage events. Use a clean, mobile-friendly design and save registrations between visits.

ChatGPT interprets the request and develops an initial application.

The user can then request changes such as adding filters, changing the layout, modifying data fields, or refining permissions.

The process resembles collaborative development with a software engineer, except much of the implementation is performed by an AI coding system.

The Development Process

A simplified ChatGPT Sites development workflow includes:

  1. Requirements interpretation. ChatGPT identifies the website's purpose, audience, content, functionality, and constraints.
  2. Application planning. The system determines the needed pages, interface components, application logic, and storage requirements.
  3. Code generation. AI-assisted development creates or modifies the necessary application files.
  4. Preview. The website runs in a preview environment before public release.
  5. Testing. The user examines appearance, navigation, forms, calculations, and other functions.
  6. Refinement. Natural-language instructions guide subsequent changes.
  7. Version preparation. A deployable version is saved for review.
  8. Deployment. The approved version is deployed to a hosted Site URL.
  9. Access configuration. The owner specifies who can view or edit the application.
  10. Maintenance. The owner can revise, republish, and manage the Site over time.

This is a conceptual description rather than a guarantee that every project follows an identical internal sequence.

From Prompt to Application

A useful starting prompt should specify more than visual appearance.

For example, requesting "a modern-looking dashboard" describes a style but not the application's purpose or behavior.

A stronger request identifies the people using the system, the information they need, the actions they should be able to perform, and the expected results.

An effective specification addresses:

  • Audience: Who will use the website?
  • Purpose: What problem does it solve?
  • Content: What information should appear?
  • Interactions: What can visitors do?
  • Data: What information must be stored or retrieved?
  • Permissions: Who can view, change, or administer information?
  • Design: What should the website look like?
  • Validation: How will successful behavior be verified?

The clarity of these requirements has a direct effect on the usefulness of the generated application.

See Prompt Engineering (PE), Requirements Management, and Development.

Getting Started

Creating Your First Site

Users with access can begin through ChatGPT Work on the web or through Work or Codex in the supported ChatGPT desktop application.

  1. Open ChatGPT and sign in.
  2. Select Work on the web, or Work or Codex in the supported desktop application.
  3. Ask ChatGPT to build a website, or mention @Sites.
  4. Describe the purpose, audience, information, and required functionality.
  5. Provide supporting files, example data, reference links, or visual specifications when appropriate.
  6. Examine the generated preview.
  7. Ask ChatGPT to correct problems and refine the design.
  8. Save a version for review.
  9. Deploy the Site.
  10. Review the published URL and configure sharing permissions.
  11. Test the published experience as an intended visitor.

The Sites management interface is available through chatgpt.com/sites for eligible accounts.

The interface and available controls may differ between the web application and desktop application.

First Project Example: A Resource Directory

A practical introductory project is a searchable resource directory.

Use @Sites to create an interactive resource directory from the documents I provide.

Organize resources by topic. Include keyword search, category filters, descriptions, and links to the original materials.

Make the interface readable on a phone and desktop.

When a search returns no results, display a helpful message.

Do not invent resources or descriptions unsupported by the source materials.

Save the initial version for review before publishing.

This project introduces several useful concepts without requiring a complicated backend.

Users can test navigation, search, filtering, layout, and content accuracy before considering more advanced capabilities.

Refining the Application

Once the initial Site exists, changes can be requested conversationally.

Examples include:

  • Move navigation to a collapsible sidebar.
  • Add a date-range filter to the dashboard.
  • Display an explanation when no results match the search.
  • Require confirmation before deleting records.
  • Make tables usable on narrow mobile screens.
  • Add a help page explaining each field.
  • Save records so they remain available after the browser closes.
  • Restrict administrative functions to authorized users.
  • Create a separate printable view.

Precise requests generally produce more testable changes than broad instructions such as "make it better."

Website and Application Capabilities

Static Websites

A static website primarily displays information.

Examples include portfolios, event announcements, instructional pages, business descriptions, product showcases, and project presentations.

A static Site may contain multiple pages, navigation menus, images, and responsive layouts without requiring a persistent database.

Static websites are often suitable as first projects because their behavior is relatively easy to inspect.

Interactive Web Applications

A web application allows visitors to perform actions and receive responses.

Examples include:

  • Calculators that update results when inputs change.
  • Dashboards with interactive charts and filters.
  • Forms that collect and validate information.
  • Educational exercises that respond to learner choices.
  • Games that maintain scores and progress.
  • Project trackers that manage records and statuses.
  • Decision-support tools that compare alternatives.

These applications generally require additional testing because correct appearance doesn't establish correct behavior.

Persistent Data

ChatGPT Sites supports persistent storage for applications that must remember information between visits.

Examples include completed tasks, project requests, game scores, registrations, saved preferences, and uploaded resources.

Persistent data differs from temporary browser state. A filter selection might not need permanent storage, while an event registration usually does.

When describing an application, specify exactly which information must be retained and which users should be able to retrieve or change it.

File Uploads

Supported Sites can store uploaded files such as documents, images, and other permitted resources.

An application might allow users to attach supporting documents to a request, upload project images, or maintain a collection of downloadable resources.

Site owners should establish appropriate rules for accepted file types, file-size validation, retention, sharing, and access.

Files uploaded by visitors are subject to the Site's application logic, hosting controls, and applicable policies.

Responsive Design

ChatGPT can generate interfaces intended for desktop computers, tablets, and phones.

Responsive design adapts layouts to different screen sizes.

Typical adaptations include stacked panels, simplified navigation, readable typography, flexible images, and scrollable tables.

However, requesting a mobile-friendly design doesn't guarantee that the result works well on every device. Important layouts should be tested at multiple screen sizes.

See Web Design, Accessibility, and Visualization.

Applications and Use Cases

Examples of ChatGPT Sites Applications
Category Example Application Useful Capabilities
Business Operations Project request dashboard Status tracking, owners, filters, persistent records
Education Interactive learning portal Lessons, questions, progress tracking
Personal Productivity Task and goal tracker Forms, saved data, dashboards
Community Organizations Club activity website Events, registration, announcements
Research Searchable reference library Search, categorization, document links
Finance Budget scenario calculator Calculations, charts, assumptions
Marketing Product landing page Responsive design, information presentation
Events Conference information portal Agendas, session directories, navigation
Gaming Browser-based game Interactive controls, game rules, scoring
Creative Work Portfolio website Media presentation, project descriptions
Business Analysis Performance dashboard Charts, filters, data summaries
Customer Support Resource and FAQ hub Search, navigation, support information

Education and Training

Educators can use Sites to turn instructional material into interactive learning experiences.

An instructor might provide a lesson outline and request quizzes, explanations, simulations, or guided exercises.

A learner could explore concepts through an interactive interface rather than reading a conventional document.

For example, a probability lesson might include a simulation demonstrating how outcomes change over repeated trials.

The educational value depends on accurate subject matter, appropriate difficulty, and effective feedback.

AI-generated exercises should be checked for incorrect answers and misleading explanations.

Sites must not be designed to target children under 13 or the applicable age of digital consent, consistent with OpenAI's applicable restrictions.

Games and Interactive Experiences

ChatGPT Sites can be used to prototype browser-based games.

Potential applications include word games, trivia contests, puzzles, strategy demonstrations, simulations, and simple multiplayer experiences when supported by the application design.

Game creators can describe rules and ask ChatGPT to implement interfaces, scoring systems, and player interactions.

For example:

Create a browser-based word game for adult players.

Display a daily puzzle, accept guesses, provide feedback, and maintain each signed-in player's progress.

Explain the rules before play begins.

Provide a practice mode with unlimited attempts and a separate daily challenge.

Test that scoring is consistent and that players cannot submit the same result twice.

For complex games, developers should explicitly define the state model, scoring rules, error conditions, and persistence requirements.

The platform may support a playable prototype without necessarily providing every capability required for a commercial multiplayer gaming service.

See Gaming, Game Development with Generative AI, Game Design, and Game Theory.

Business Dashboards

A dashboard can turn a spreadsheet or approved data source into an interactive presentation.

Users may ask ChatGPT to generate charts, summary cards, filters, comparison views, and explanatory text.

A sales dashboard, for example, might display revenue by month, sales by region, and the status of outstanding opportunities.

The most important requirement is that calculations be traceable to the underlying data.

Generated charts should not be considered reliable merely because they look professional.

See Analytics, Visualization, Spreadsheets, and Data Quality.

Data Storage and Application Architecture

Managed Application Hosting

ChatGPT Sites combines website creation with managed hosting.

In a traditional development process, a developer may need to arrange a web server, deployment system, database, file storage, SSL certificates, and access controls.

Sites handles supported hosting and deployment operations within the platform.

This reduces initial setup work, particularly for small applications and prototypes.

However, Site owners remain responsible for application correctness, content, security-sensitive design decisions, and compliance.

Database Storage

The Sites runtime supports a relational database capability identified in its technical documentation as D1.

A database allows an application to store structured records and retrieve them when needed.

For example, a project tracker might maintain records containing a task identifier, title, owner, priority, status, and due date.

The published storage limit is 10 GB of database data per Site.

Database features and limitations should be evaluated before using Sites for a large or business-critical application.

See Database, SQL, and Data Management.

Object and File Storage

The platform also supports object storage identified as R2 in its technical documentation.

Object storage is suited to files such as photographs, documents, and other uploaded materials.

OpenAI's documentation lists no fixed per-Site file storage limit, although overall usage allowances, platform restrictions, and applicable billing still matter.

Database storage and file storage serve different purposes.

A database might store the title and owner of a document, while object storage holds the document itself.

Application State

Web applications often distinguish among three types of state:

  • Temporary interface state: Information needed only during the current interaction, such as an open menu.
  • Session or identity state: Information associated with a signed-in user.
  • Persistent application state: Information that should remain available after a session ends.

This distinction matters when specifying a Site.

For instance, a quiz might need to remember a student's answers across visits, while a temporary tooltip doesn't require database storage.

Clearly describing persistence requirements helps avoid unnecessary data storage and unpredictable behavior.

Runtime Restrictions

ChatGPT Sites isn't a substitute for every hosting environment.

OpenAI documents support for HTTP, HTTPS, and WebSockets, but not arbitrary inbound or outbound raw TCP connections.

Some frameworks, database systems, private-network configurations, background processes, and specialized hosting patterns may not be supported.

Developers should verify runtime compatibility rather than assuming any locally working application can be deployed unchanged.

Sign-In, Identity, and User Access

Private and Public Sites

A newly created Site is private by default, with access limited to its owner and applicable workspace administrators.

The owner can later configure sharing according to the available account and workspace options.

Supported access patterns include:

  • Owner and workspace administrators.
  • Selected authorized individuals or groups.
  • Members of the same workspace.
  • Named external viewers.
  • Anyone on the internet, when public publishing is permitted.

Available settings depend on plan, role, geography, and administrator policy.

Sign in with ChatGPT

Public Sites can optionally use Sign in with ChatGPT to recognize visitors and support personalized functions.

This can help implement saved progress, individual records, and interfaces that change according to the signed-in user.

Public access and personalized sign-in are distinct concepts.

A website may allow anonymous visitors to browse general information while requiring sign-in for particular features.

The application must still implement authorization correctly.

Authentication establishes a user's identity; authorization determines which information and actions that user may access.

Workspace Authentication

Sites restricted to a ChatGPT workspace use account identity to enforce workspace-level sharing.

A workspace Site could provide a project dashboard to employees while denying access to people outside the organization.

However, merely having workspace access doesn't mean every person should be permitted to see every application record.

The application may need additional record-level authorization.

Collaboration and Editing

Site owners can grant eligible members of the same workspace editing access.

Editors can modify the Site and save versions. After the owner completes the first publication, editors may publish updates to the existing Site URL.

The owner controls sharing and other management permissions.

Editing rights should be granted carefully because editors may be able to inspect application data and publish code that interacts with configured secrets.

Connected Applications and External Services

Connecting to Other Systems

In supported organizational environments, ChatGPT Sites can interact with approved connected services.

This makes it possible to develop applications that display information from tools already used by a team.

A Site might display assigned project issues, search permitted documents, or present information from supported organizational systems.

The exact integration capabilities depend on the available plugins, workspace configuration, service permissions, and the Site's implementation.

Visitor-Specific Permissions

An important design principle is that a Site's users don't automatically inherit the creator's connected accounts.

Each visitor authorizes supported connections using their own identity and existing permissions.

For example, an employee opening an issue-tracking Site may see only the issues that their connected account permits them to access.

A colleague using the same Site could see a different set of records.

Where supported, a connector may request read or write permissions. Users should examine the requested access carefully before approving any ability to change information.

Connected-app use is subject to organizational and tenant policies and is intended for supported private workspace Sites.

Private-Network Connections

OpenAI also documents private-network tunnels for eligible organizational deployments.

A tunnel allows a Site to communicate with an approved service that isn't otherwise publicly accessible.

For example, an organization might connect a Site to an internal inventory API.

The organization's IT team must configure the connection, and the underlying service's authentication and authorization requirements continue to apply.

Tunnels shouldn't be treated as unrestricted access to the organization's private network.

Hosting Tools for ChatGPT Plugins

A ChatGPT Site can also host an MCP server that exposes application functions as tools usable through a ChatGPT plugin.

This introduces another relationship between websites and AI agents.

Instead of using only a graphical web interface, a user may interact with selected application capabilities conversationally through a plugin.

For example, a Site hosting a project-management workflow might expose a tool for retrieving approved project status information.

The permissions of the Site, plugin, user, and connected service must all be considered.

See Agents/Assistants, ...build an agent, and Web Automation.

Custom Domains and Publishing

Hosted Site Addresses

When a Site is deployed, ChatGPT generates a hosted web address.

The owner can share that address according to the Site's audience settings.

Supported Site URLs may contain a descriptive name and an account or workspace identifier.

Users should inspect the complete address before making a Site public, particularly when account names or workspace names could reveal information they didn't intend to publish.

Changing the URL

Eligible Site owners can rename their ChatGPT-hosted address.

OpenAI documents redirects from the previous address to the new one when the applicable workspace address remains unchanged.

Custom domains and workspace-address changes may follow different rules.

A URL change should be tested to ensure that bookmarks, internal navigation, and shared links still work as expected.

Custom Domains

Sites supports custom domains where the feature is available.

A custom domain allows a website to use an address owned by the creator rather than the default hosted address.

For example, an organization might publish a project resource hub under a subdomain it controls.

The owner must possess the domain and have permission to update its DNS records.

Typical configuration involves entering the desired hostname, adding the required DNS records, waiting for verification, and testing the resulting address.

ChatGPT Sites doesn't register a domain on the owner's behalf.

Preview, Saving, and Deployment

A preview allows the user to evaluate changes before making them available to the published audience.

Saving a version and deploying a version are different operations.

A saved version can be reviewed without immediately replacing the live application.

Deploying an updated version changes the live Site.

Because every deployment URL represents a production deployment, users should carefully distinguish testing from publication.

Updating a Published Site

Users can return to an existing Site and describe the changes they want.

Typical updates include revising text, adjusting layouts, changing calculations, improving accessibility, and adding functionality.

An updated version should be tested before publication.

Changes involving database structure require additional caution.

Restoring an earlier version of a Site restores its code, not necessarily its database or uploaded files.

Scheduled Updates and Automation

ChatGPT Sites supports scheduled work associated with a Site where the feature is available.

A Site owner may configure an automation to perform recurring work using supported resources.

For example, a scheduled process could prepare a recurring update for a project dashboard.

Site schedules run in the cloud and are managed through the Site's automation settings.

The owner must have the required permissions and access to any information the scheduled task needs.

An important limitation is that a visitor's temporary authorization to a connected application doesn't automatically become available to background tasks.

A scheduled task must have its own supported access to the data it needs.

This is important when designing dashboards intended to remain current without manual intervention.

See Agents/Assistants, Robotic Process Automation, and Personal Productivity.

ChatGPT Sites Compared with Other Development Approaches

Website and Application Development Approaches
Capability ChatGPT Sites Traditional Website Builder Custom Software Development
Primary Interface Natural-language instructions Visual editing and templates Source code and development tools
Programming Knowledge Optional for many projects Usually unnecessary for basic sites Generally required
Application Logic AI-generated and refined Depends on platform features Developer-controlled
Interactive Applications Supported within runtime limits Varies by builder Broadly supported
Persistent Data Supported Depends on platform Configurable
Hosting Integrated Often integrated Separate or integrated
Custom Domains Supported where available Commonly supported Configurable
Deployment Managed through Sites Managed by builder Developer-defined
Source-Level Control Available in compatible Codex workflows Varies Extensive
Advanced Infrastructure Limited to supported runtime Platform-dependent Potentially extensive
Main Strength Rapid conversational application creation Accessible visual website design Maximum architectural control
Main Limitation Runtime, platform, and usage restrictions Template and platform constraints Development and maintenance complexity

ChatGPT Sites Versus ChatGPT Code Generation

Ordinary ChatGPT code generation can produce HTML, CSS, JavaScript, and other programming files.

However, generated code alone isn't necessarily a deployed website.

The user may still need to create a project, configure dependencies, establish hosting, and maintain a deployment process.

ChatGPT Sites adds a managed environment in which compatible applications can be previewed, deployed, shared, and maintained.

This distinction is important for people who want a working website rather than source code to implement independently.

ChatGPT Sites Versus Codex

Codex is a broader AI-assisted software development environment.

It can work with codebases, development tasks, testing, and other programming workflows.

ChatGPT Sites focuses on turning a suitable project into a hosted web experience.

These capabilities are complementary.

A developer can use Codex to build or refine an application and Sites to host and share it.

Some software projects will exceed the supported Sites runtime and require another deployment platform.

ChatGPT Sites Versus No-Code Platforms

Traditional no-code platforms often use visual editors, predefined components, database interfaces, and workflow designers.

ChatGPT Sites relies more heavily on natural-language instructions and AI-generated implementation.

This can make it easier to begin with a custom idea instead of selecting from an existing library of components.

However, visual builders may provide more predictable controls for certain repetitive tasks, while a custom development environment may provide more flexibility for advanced applications.

The most appropriate approach depends on application complexity, user skill, governance requirements, and expected maintenance.

Pricing, Availability, and Resource Limits

Subscription Eligibility

As of November 2026, ChatGPT Sites is available through eligible Plus, Pro, Business, Enterprise, and Edu accounts, subject to regional availability and workspace controls.

Free and Go plans don't provide Site creation access.

In managed environments, administrators may limit who can create Sites, publish them, invite external viewers, or connect approved services.

Enterprise public publishing is disabled by default unless an administrator enables the relevant permissions.

Development Usage

Building and editing Sites counts toward the user's applicable Codex usage.

This is separate from the resources consumed when a published application receives visitors or stores information.

A particularly complex application may consume more development resources than a simple static website.

Site owners should distinguish development consumption from ongoing hosting consumption.

Hosting Allowances

OpenAI's November 2026 documentation describes included hosting allowances that depend on the subscription.

Hosting resource consumption can include traffic, database storage, file storage, custom domains, and supported private connections.

For Plus and Pro users, OpenAI has removed the fixed numerical limit on Sites, but shared usage and storage limits still apply.

A large number of small Sites may therefore be permitted while resource-intensive applications remain constrained by the account's allowance.

Hosting Pricing Transition

OpenAI's published documentation states that Sites hosting is free through December 31, 2026.

Beginning January 1, 2027, OpenAI plans to introduce metering with included monthly allowances and potential overage charges.

The published rates are proposed and may change before billing begins.

Proposed Hosting Resource Rates for January 2027
Resource Billing Unit Approximate Credits
Site without storage Per Site per month 0
Site with storage Per Site per month 25
Traffic 100,000 requests 16
Database storage 1 GiB for one month 150
File storage 1 GiB for one month 3.75
Custom domain Per domain per month 25
Private connection Per tunnel per month 125

These amounts represent proposed usage credits rather than universal dollar prices.

The actual cost will depend on plan allowances, applicable credit pricing, resource consumption, and contractual terms.

A Site with an empty database may still incur a database-related base charge when metered hosting begins.

Enterprise billing and temporary waivers are governed by the organization's agreement and current OpenAI policies.

For current rates and allowances, consult official Sites documentation.

Privacy, Security, and Governance

Responsibility for Published Content

Site creators are responsible for reviewing the content and behavior of their published applications.

This responsibility extends to generated text, code, uploaded materials, visitor submissions, and information displayed through integrations.

AI-generated code can contain mistakes or security weaknesses, and generated content can contain unsupported factual claims.

Publication doesn't mean OpenAI has independently reviewed or endorsed the Site.

Owners must ensure that they have permission to use and distribute the content included in their applications.

Personal Information

Sites that collect personal information require particular attention to privacy.

Examples include registration forms, customer inquiries, user profiles, and saved progress associated with a signed-in account.

Creators should explain what information they collect, why it is needed, who can access it, and how it is retained or deleted.

Applicable privacy and data-protection laws may require additional notices, consent procedures, or administrative controls.

The existence of an authentication feature doesn't automatically establish compliance.

Sensitive and Prohibited Uses

OpenAI places restrictions on how Sites may be used.

These include restrictions involving Protected Health Information, prohibited financial transactions, payment-card handling, malicious software, phishing, impersonation, and applications targeting children below the applicable minimum age.

Permitted e-commerce activity may use a third-party payment processor.

Payment-card data must be collected and processed by the approved third-party provider rather than the Site itself.

The Site owner remains responsible for the commercial integration, customer support, fulfillment, refunds, and applicable taxes.

See the ChatGPT Sites Terms.

Security Testing

Site owners should inspect authentication, authorization, input validation, error handling, and access to stored information.

A form that functions correctly for one user may still contain an authorization error that exposes another user's records.

Testing should therefore include multiple user roles and scenarios.

Examples include:

  • Submitting invalid or incomplete form data.
  • Attempting unauthorized access to records.
  • Verifying that private pages can't be accessed anonymously.
  • Confirming that deleted or revoked users lose access.
  • Testing links and file downloads using a separate account.
  • Checking whether a database update affects unrelated records.
  • Verifying that secret values aren't exposed through the user interface.

See Cybersecurity, Privacy, Prompt Injection Attack, and Software Testing.

Organizational Controls

Business and Enterprise administrators can manage aspects of Sites availability, sharing, and integration permissions.

Enterprise environments support role-based controls for creation and public publication.

Administrators can also manage which supported plugins may be used within Sites.

Organizations should establish policies for:

  • Who may create and publish Sites.
  • What information may be stored or displayed.
  • Which external connections are permitted.
  • Who reviews applications before public publication.
  • How suspected security or privacy problems are reported.
  • Who is responsible for ongoing maintenance.
  • How ownership is transferred when employees change roles.

Data Residency

OpenAI documents regional hosting behavior for eligible organizational workspaces.

New Sites in workspaces with supported United States or European Union data residency use the workspace's applicable region for supported storage and hosting.

Other regional configurations may not apply their workspace residency settings to Sites.

Existing Sites are not necessarily migrated when regional hosting capabilities change.

Organizations with data residency requirements should verify the specific hosting and inference arrangements that apply to their workspace.

Limitations and Risks

AI-Generated Code Can Be Incorrect

ChatGPT can produce a functioning-looking interface while misunderstanding the intended logic.

A calculator might use the wrong formula, a game might implement an incorrect scoring condition, or a dashboard might aggregate records improperly.

Testing must address the application's actual requirements rather than its visual appearance alone.

Requirements Can Be Ambiguous

Natural-language specifications may omit important details.

For example, "allow members to edit events" doesn't explain which members, which events, or whether edits require approval.

Developers should define permissions and business rules explicitly.

Platform Dependence

Sites relies on OpenAI's supported hosting runtime, account access, deployment system, and usage policies.

An application that needs unsupported infrastructure may require another hosting environment.

Commercial projects should evaluate long-term maintenance, integration requirements, data portability, and operational risk before committing to a platform.

Database Changes Require Care

Updating application code can affect persistent data.

OpenAI warns that some database changes may occur during publication even when the deployment doesn't complete successfully.

Restoring old code doesn't automatically restore an earlier database state.

Applications containing valuable information should have an appropriate backup and recovery strategy.

Usage and Cost Can Change

Traffic, uploaded files, and stored records can increase resource consumption.

A public website may experience usage very different from its initial testing conditions.

Creators should monitor limits and understand the applicable subscription allowances and hosting policies.

Best Practices

Define a Minimum Viable Product

Begin with the smallest application that proves the central idea.

For a project tracker, the first version might allow users to create tasks, assign owners, and update statuses.

Additional features such as analytics, notifications, complex permissions, and integrations can be introduced after the basic workflow has been tested.

This approach reduces development complexity and makes failures easier to diagnose.

Use Representative Test Data

Provide sample data that resembles real application usage without unnecessarily exposing private information.

Test records should include normal cases, missing values, duplicates, unusual characters, and boundary conditions.

For analytical applications, include examples with known expected results.

Separate Content from Application Logic

When possible, distinguish the information displayed by the application from the rules controlling its behavior.

For example, a game should have a clearly defined scoring system that can be independently checked.

A dashboard should have documented metric definitions.

A registration system should have explicit rules for eligibility, capacity, and duplicate submissions.

This reduces the likelihood that later interface changes will accidentally alter core behavior.

Save Versions Before Major Changes

Preserve an approved version before requesting substantial revisions.

This provides a clearer recovery point if an update introduces a defect.

Remember that restoring application code doesn't necessarily reverse database or file changes.

Verify the Published Experience

A Site may behave differently for a visitor than it does for its creator.

Test the production address using the intended account type and access permissions.

Check navigation, authentication, persistent records, file uploads, calculations, and mobile layouts.

For private applications, verify that unauthorized visitors can't access the content.

Example Prompts

Example 1: Business Dashboard

Use @Sites to build an interactive business performance dashboard from the spreadsheet I provide.

Show monthly revenue, expenses, operating margin, and year-to-date totals.

Include a date filter and comparisons against the prior period.

Use the formulas and definitions from the source spreadsheet. Do not invent missing values.

Add a data dictionary explaining each metric.

Verify every displayed total against the source data.

Keep the Site private while I review it.

Example 2: Community Event Website

Create a mobile-friendly website for a community organization.

Include a home page, upcoming events, event details, and a registration form.

Allow visitors to register for events and save their registrations.

Provide an administrator interface to create, edit, and cancel events.

Prevent duplicate registrations and display a useful message when an event reaches capacity.

Require authorization for administrative functions.

Prepare a private preview before publishing.

Example 3: Interactive Learning Application

Use @Sites to create an interactive introduction to machine learning for adult beginners.

Include short lessons explaining supervised learning, unsupervised learning, and reinforcement learning.

Provide interactive examples and five-question quizzes after each lesson.

Explain incorrect answers and allow learners to retry.

Avoid unnecessary mathematical jargon, but preserve technical accuracy.

Make the application responsive and accessible.

Test the quiz scoring before publication.

Example 4: Game Prototype

Build a browser-based prototype of a strategy game using the rules I provide.

Use the supplied rules as the authoritative specification.

Implement player actions, score calculations, win conditions, and a reset function.

Include a visible explanation of the rules.

Separate the game logic from interface presentation.

Create automated tests for the scoring and victory conditions.

Do not introduce new mechanics without my approval.

Save a version for review before publishing.

Example 5: Searchable Knowledge Hub

Use @Sites to build a searchable knowledge hub from an approved collection of documents.

Organize the information by subject, provide filters and keyword search, and link to the original sources.

Display concise descriptions for each resource.

Do not generate unsupported facts or invent references.

Provide an empty-results message and clear navigation.

Ensure that visitors can access only the resources they are permitted to view.

Test search results using a representative collection of queries.

Troubleshooting

Common ChatGPT Sites Problems
Problem Possible Cause Recommended Action
Sites isn't available Plan, regional, or workspace restriction Review eligibility and administrator settings
Site creation is blocked Shared usage or resource limit Inspect the limit notification and usage information
Site preview is incorrect Ambiguous requirements or generated implementation error Describe the defect precisely and provide expected behavior
Published Site differs from preview Wrong version deployed or environment difference Review the saved version and deployment status
Visitor can't open the Site Sharing permissions or account mismatch Verify audience settings and the visitor's account
Connected app doesn't work Connection, plugin, or tenant permission issue Verify app permissions and workspace policies
Stored data disappears Persistence not implemented or incorrect database behavior Inspect storage configuration and test after reload
Custom domain returns an error DNS or verification problem Compare DNS records with current Site settings
Update breaks existing records Database schema or migration problem Review migrations and recovery options
Old code is restored but data remains changed Code and data have different restoration behavior Restore data separately when a valid recovery mechanism exists
Public Site exceeds limits Traffic or storage consumption Review usage and applicable hosting allowances

Frequently Asked Questions

Do I need programming experience?

No. Users can create many supported websites by describing their requirements in natural language.

However, programming knowledge can help when diagnosing complex behavior, integrating external services, or reviewing security-sensitive code.

Can ChatGPT Sites host a complete application?

Yes, when the application fits the supported runtime and platform capabilities.

The service supports interactive interfaces, persistent storage, authentication, and certain external integrations.

It doesn't support every framework, infrastructure pattern, or backend requirement.

Can I publish a Site publicly?

Yes, where public publishing is available and permitted by the account and workspace settings.

Enterprise workspaces require administrator authorization for public publishing.

Can other people edit my Site?

Eligible members of the same workspace can be granted editing permissions.

External viewers don't receive editing rights merely by being invited to view the Site.

Can I use my own domain?

Yes, where the custom-domain feature is available.

You must control the domain and configure the required DNS records.

Can visitors save information?

Yes. Applications can use persistent database and file storage, subject to supported capabilities and limits.

Can Sites connect to business applications?

Yes, in supported workspace environments using approved plugins and connections.

Each visitor's access is governed by their own permissions and the Site's configuration.

Can I use Sites to sell products or services?

Permitted e-commerce activities can use an external payment processor, subject to the ChatGPT Sites Terms.

Site owners remain responsible for their commercial activities and legal obligations.

Can I export or move a Site to another host?

Compatible source projects can be developed through Codex workflows, but the portability of a complete application depends on its framework, database, storage, authentication, and platform-specific services.

Users shouldn't assume that exporting source code automatically reproduces the hosted environment elsewhere.

What happens if I delete a Site?

Deleting a Site permanently removes it. OpenAI's documentation states that deleted Sites cannot be restored.

Owners should preserve any information they need before deletion.

Glossary

Important ChatGPT Sites Concepts
Term Meaning
ChatGPT Sites OpenAI's hosted website and application creation service
Codex AI-assisted software development capabilities used to implement and modify code
Prompt Natural-language instruction describing an intended result
Preview A version used to inspect an application before publishing
Deployment Publication of a prepared application version to a hosted environment
Runtime The execution environment in which the application operates
D1 Relational database storage supported by the Sites runtime
R2 Object storage supported for application files
Persistent State Information retained beyond a single browser interaction or session
Authentication The process of verifying a user's identity
Authorization Rules that determine which actions and information a user may access
Custom Domain An owner-controlled web address associated with a published Site
Plugin An integration providing approved tools or access to external services
MCP Model Context Protocol, a mechanism for exposing tools and information to AI systems
Tunnel A controlled connection between a Site and a service on a private network
Automation A scheduled or otherwise triggered process that performs supported work
Vibe Coding Informal term for developing software through natural-language instructions to AI systems
No-Code Development Building applications without manually writing most underlying source code

Learning Resources

Official Documentation

Tutorials and Guided Learning

Further Research

Related PRIMO.ai Topics

Conclusion

ChatGPT Sites represents a significant development in AI-assisted software creation because it combines application generation with hosting, publishing, persistent storage, and controlled sharing.

The technology reduces the practical distance between describing a software idea and making a working application available to others.

It can be particularly useful for individuals, educators, community organizations, and business teams that need relatively focused web applications without establishing a conventional development and hosting environment.

However, AI-generated applications still require careful specifications, functional testing, security review, and ongoing maintenance.

ChatGPT Sites should therefore be understood not as a replacement for all software engineering, but as an increasingly capable platform for translating natural-language requirements into usable software.

Its importance extends beyond website design. It illustrates the broader movement toward agentic AI systems that can create, deploy, and maintain functional digital tools rather than merely explain how those tools could be built.

Retrieved from "https://primo.ai/w/index.php?title=ChatGPT_Sites&oldid=53050"