Creating a site configuration with Copilot in SharePoint—Review permissions and costs before confirming

About this article
This article was generated using an automated workflow powered by generative AI. Official Microsoft information was re-verified on November 2, 2026, and organized into procedures for planning a fictitious project site using 'Create site with AI' in Copilot in SharePoint and reviewing configurations, permissions, and costs before and after creation. Signing into a Microsoft 365 tenant, creating a site, consuming Copilot Credits, and displaying on WordPress were not executed.

Validation Status: 📘 Official Information Verified – Real Tenant Unverified

Usage Requirements at a Glance (Official Verification: November 2, 2026)

First Test: Reviewing the Plan Without Creating a Site

1. Open 'Create site' -> 'Create a site with AI' in SharePoint, or 'Build' in the new interface. If not displayed, check the provisioning status with your administrator.

2. Enter the prompt for the fictitious project below,and view the plan and architecture diagram onlyis displayed.

3. Compare the counts of pages, lists, and libraries,without pressing Confirmand stop.

Check here after the operation:Verify whether 3 pages, 2 lists, and 1 library are lined up as proposals. This is an expected value and not a success record in an actual tenant. If the configuration differs, adjust it using "Change only one condition" described later.

If you simply ask Copilot to "create a project site," you can shape it in a short time. However, what is important for enterprise use is not speed, but confirming what can be done before creation and determining who can view it after creation.

According to Microsoft's official documentation, Copilot in SharePoint can propose sites, pages, lists, libraries, and more from natural language. When creating a site, instead of generating assets immediately, it displays a plan, and builds them after a human reviews and clicks Confirm. This "time when it is not yet created" is used as a review process.

Test with Fictional Business Data

Do not enter actual department names, customer names, email addresses, or project information. Assume a fictional "Company A New Service Verification Project."

User names should be fictitious names like User A, and documents should be restricted to dummies such as sample-plan.docx Avoiding the creation of production sites or real projects during the Preview feature evaluation is the primary safety measure.

Step 1: Do not create yet, have it output a plan first

From the SharePoint start page, go to "Create site with AI" or select "Build" on the new screen, and paste the following prompt.

架空の「A社・新サービス検証プロジェクト」用の
SharePointチームサイトを計画してください。

必要な構成:

- ホーム、進捗、FAQの3ページ

- タスク一覧

- 課題一覧

- 検証資料ライブラリ

タスク一覧の列:
件名、担当者、期限、状態、優先度

課題一覧の列:
課題ID、内容、影響、担当者、対応期限、状態

外部共有、個人情報、実在する会社・人物・案件のデータは使いません。
サンプルデータは架空と明示してください。

まだ作成しないでください。
最初にページ、リスト、ライブラリ、列、ビューの構成案だけを示してください。

Here, do not perform operations equivalent to the final "Create", "Build", or "Confirm". Record what appears on the planning screen.

Expected observation

Even if the AI's suggestion looks good, if the column types do not match the use case, you will run into trouble with aggregation and notifications. Check that the assignee is Person, the deadline is Date, and the status and priority are Choice candidates. Even if the prompt specifies column names, the types may not always be as expected.

Step 2: Change only one condition and observe the difference

After saving or screen-recording the initial proposal, change only the FAQ page to "Decision Log". Do not change any other conditions.

現在の計画のうち、FAQページだけを削除し、
代わりに「意思決定ログ」リストを追加してください。

列は、決定日、決定事項、理由、決定者、関連課題IDです。
それ以外のページ、リスト、ライブラリ、列は変更しないでください。
まだ作成しないでください。変更前後の差分を示してください。

The expected difference is simply "one less FAQ and one more Decision Log." If the task column disappears or the library name changes, do not adopt that proposal.

Changing a single condition helps test whether the AI can maintain localized fixes. Changing multiple conditions at the same time makes it impossible to determine what caused the structure to break.

Review table prior to confirmation

The operational rule is not to confirm while leaving items marked as "Unverified". The AI generating a draft is separate from satisfying governance conditions.

Step 3: Minimum Pilot for Creation

Build as a private validation site only after completing self-QA for P1 through P5 and administrator approval. According to official information, assets are not created until confirmation, but the process cannot be stopped once building has started. Save the screen plan and cost confirmation before execution, and record the start time.

After creation, verify the following in order:

1. Site name, URL, Owner, Member, Visitor

2. public/private and external sharing

3. Number of pages, lists, and libraries

4. Column names, types, required fields, Choice values, default views

5. Whether sample data is fictional

6. Unexpected assets added by Copilot

7. Sensitivity labels, retention, DLP

8. Credit usage and administrator-side records

9. Deletion or isolation procedure upon failure. Because this actual operation has not been performed, the result is not treated as "success".

Detailed target environment and constraints (Officially confirmed: November 2, 2026)

Failure examples and reasons

"Create a site with AI" is not displayed.

The main troubleshooting areas include targeted release, Preview settings, Microsoft Copilot licenses, site creation permissions, and the rollout status of the new SharePoint interface. According to the official prerequisites for creating sites with AI, the tenant setting for Preview must be set to either "All Sites" or "Exclude Selected Sites," and it is not available when set to "Selected Sites" or "No Sites." Simply having an E3 or E5 license does not guarantee that Copilot is automatically assigned to the user. Administrators should verify the licenses and feature settings before users repeatedly reload the page.

Stops midway due to insufficient credits.

Routine tasks in Copilot in SharePoint and advanced work that creates an entire site have different cost boundaries. If administrators have not configured consumption-based billing and spending policies, advanced work may not be able to be completed. Read the displayed confirmation and do not approve it if the budget owner is unknown.

The plan is correct, but permissions are too broad.

The quality of the architecture proposal and access control are separate issues. Microsoft lists unsupported items during the Preview phase as permission and sharing settings, complex Power Automate flows, column and row formatting, advanced web parts, and third-party integrations. Humans must verify the visibility scope, groups, and Owners/Members/Visitors upon creation, and not expect AI to automatically determine the principle of least privilege for business operations. Especially in a Preview environment, it is necessary to decide not to use test sites as storage locations for production content.

Found to be unnecessary after creation.

Since generation cannot be canceled after it has started, pre-confirmation review is the primary preventive measure. After creation, the owner decides whether to isolate, review the content, or delete the site before adding users and continuing to use it. If sample data exists, process it before it mixes with actual data.

Practical decision-making in operations

Sites where "Create site with AI" is easily adopted are short-term projects, training, FAQs, and test sites consisting of standard pages, lists, and libraries. This is because it reduces the time required to create the structural proposal and allows humans to review column types and permissions.

On the other hand, for sites that initially require complex approvals, external integrations, advanced formatting, strict record management, or custom web parts, it is safer not to consider the initial AI-generated proposal as final. Retain traditional design and administrator reviews, and limit Copilot to drafting the initial outline.

Pilot evaluations should compare not only "creation time," but also the following totals.

We determine that the adoption impact is low if faster generation leads to an increase in corrections and cleanup.

Summary

Site creation with Copilot in SharePoint is practical because it generates plans for pages, lists, and libraries from natural language and allows modifications before confirmation. However, Create sites with AI is currently in Preview at the confirmation stage, and creating entire sites or solutions requires meter-based billing and Copilot Credits as advanced work.

A safe way to test is to generate a plan using dummy data, modify a single condition, and verify the structure, permissions, visibility, and cost before creation. AI suggestions are a starting point for design and do not replace administrator approval or the principle of least privilege.

Official and Primary Sources

• Microsoft Learn — Get started with Copilot in SharePoint

• Microsoft Learn — Create sites with AI

• Microsoft Support — How to create a SharePoint site with AI

• Microsoft Learn — License options for Microsoft Copilot