How do you set up Attack simulation training in Microsoft 365?

Microsoft 365 has a built-in phishing simulator that sends realistic but harmless lures, records who clicks or reports, and assigns short training. Used well, it shows where your people need help. Used badly, it teaches staff to hide mistakes. This guide covers licensing, every simulation technique, the setup wizard, reports, and how to run tests fairly.

Who can use Attack simulation training?

Microsoft says Attack simulation training "requires a Microsoft 365 E5 or Microsoft Defender for Office 365 Plan 2 license" (Microsoft Learn). Business Premium includes Plan 1, not Plan 2, so most small businesses need an add-on. See Defender for Office 365 Plan 1 vs Plan 2.

Two other options at the time of writing (October 2026):

• E3 trial subset. Microsoft offers E3 customers a trial with "the ability to use a Credential Harvest payload and the ability to select 'ISA Phishing' or 'Mass Market Phishing' training experiences." Nothing else is included.
• 90-day Plan 2 trial, from the trials hub in the Defender portal.

The tool is at Email & collaboration > Attack simulation training in the Defender portal (https://security.microsoft.com/attacksimulator).

Permissions

Use the narrowest Microsoft Entra role that fits:

There are no PowerShell cmdlets for Attack simulation training. Automation goes through the Microsoft Graph API.

Which attack techniques can you simulate?

Microsoft offers seven social engineering techniques. All except How-to Guide come from the MITRE ATT&CK framework (Microsoft Learn).

The compromise definitions come from Microsoft's reporting documentation. Credential Harvest, Link to Malware, Drive-by URL and OAuth Consent Grant can use QR codes instead of links (Microsoft Learn), which is useful practice for QR code phishing. OAuth simulations map to the real attack in OAuth consent phishing.

What are payloads, login pages and landing pages?

• Payload: the simulated email, with its link or attachment. Microsoft provides many built-in payloads. You can copy and edit one, or write your own. Built-in payloads can't be edited directly.
• Login page: the fake sign-in page used by Credential Harvest and Link in Attachment payloads.
• Landing page: where users end up after they act on the payload, usually a short explanation.
• Payload automation: turns real phishing detected in Microsoft 365 into harmless simulations.

Each payload shows a predicted compromise rate, which uses data across Microsoft 365 to estimate the share of people likely to fall for it. Use it to pick a payload that's realistic rather than impossible.

Two content rules to know: payloads containing terms such as "IRS" or "Internal Revenue Service" are blocked, and Microsoft says customers remain responsible for deciding the appropriate use of any payload, including ones that show third-party brands (Microsoft Learn).

How do you launch a simulation?

Go to Attack simulation training > Simulations > Launch a simulation (Microsoft Learn):

1. Select technique. Start with Credential Harvest; it's the most common real attack.
2. Name and describe the simulation. You can Save and close as a draft at any point after this.
3. Select a payload and login page. Choose one that matches lures your business really receives, such as a shared-document notice.
4. Target users. Include everyone or choose specific users and groups. Microsoft finds Entra groups "easier to manage". Shared mailboxes aren't supported.
5. Exclude users who shouldn't be tested, such as people on leave.
6. Assign training. The default, Microsoft training experience (Recommended) with Assign training for me (Recommended), picks modules based on each user's history. You can choose modules yourself, redirect to your own training URL, or pick No training. The due date is 30 days after the simulation ends by default, or 15 or 7 days.
7. Select a landing page.
8. Select end user notifications: training assignment and reminder messages, plus an optional positive reinforcement notification for people who report the simulation.
9. Launch details. Launch now or schedule. The simulation ends after two days by default, which is also the minimum, and you can set up to 30 days. Enable region aware time zone delivery sends messages during each user's working hours.
10. Review, use Send a test to see the message yourself, then Submit.

Microsoft recommends that you "assign training for each simulation, as employees who go through training are less susceptible to similar attacks." Turn on positive reinforcement too. Thanking people who report is how you build the habit you actually want.

Things that break simulations

• Other security tools clicking links. Microsoft says email security applications, endpoint software, Outlook add-ins and automated triage playbooks can create false clicks and compromises. It recommends exempting all simulation URLs and domains in those tools, and notes "we don't send simulation messages from a static list of IP addresses."
• Browsers blocking simulation URLs. Google Safe Browsing may show "Deceptive site ahead" for some simulation URLs. Test the URL in your browsers first.
• Reporting mailbox detonation. If reported messages go to a custom mailbox without the right exclusions, Safe Links or Safe Attachments can detonate them and assign training to users who reported correctly. See user reported message settings.
• Audit logging off. Reports rely on the audit log. Microsoft says that if it's off, reports appear empty and training assignments are blocked.

Third-party simulation tools need a different setup, described in our guide to the advanced delivery policy for phishing simulations.

What do the reports show?

The Overview and Reports tabs summarize results, and the Attack simulation report gives detail (Microsoft Learn):

Simulation data is kept for 18 months unless you delete it (Microsoft Learn). Two reporting details matter: simulation messages reported through non-Microsoft report buttons aren't counted, and on-premises mailboxes don't show read, forward, delete or report data.

The most useful number isn't the click rate. It's how many people reported the message, and how fast. A falling click rate can just mean easier payloads; a rising report rate means your team is acting as an early warning system.

How do you run simulations ethically?

The UK's National Cyber Security Centre warns about "unhelpful 'us vs them' scenarios that phishing simulations may engender, where staff may feel they are being tested by the security team" (NCSC). It also says training is "the layer that is often over-emphasised in phishing defences." Follow these rules:

1. Tell leadership first. Agree the purpose, the schedule and who sees named results. Executives get phished too, so include them in the targets.
2. Tell staff that simulations happen, in general terms. Don't announce dates.
3. Never punish a click. No public lists, no disciplinary action, no performance reviews. People who fear blame stop reporting real attacks.
4. Avoid cruel lures. Fake bonuses, layoffs, pay changes or health scares damage trust. Microsoft's payload restrictions and your HR team are a useful check.
5. Celebrate reports. Use positive reinforcement notifications and mention good catches in team meetings.
6. Limit who sees named data. Share aggregate results widely, individual results narrowly.
7. Fix systems, not just people. If a payload fools half the company, ask why that kind of message reaches inboxes, and whether a process such as phoning to confirm bank changes would stop the real version.

Our guide to email security awareness training for small businesses covers building a reporting culture in more depth.

A simple first-year plan

1. Month 1: send a How-to Guide simulation on reporting phishing. No test, just teaching.
2. Month 2: run a Credential Harvest simulation with an easy payload, training assigned and positive reinforcement on.
3. Every 2 to 3 months: rotate techniques: Link in Attachment, QR code, OAuth Consent Grant.
4. Each time: review report rate, training completion and repeat offenders. Offer extra help, not penalties.
5. Yearly: share an aggregate summary with leadership and adjust.

Next steps

• Check your license: Attack simulation training appears only with E5 or Plan 2.
• Write a one-page simulation policy with leadership and HR before the first launch.
• Exempt Microsoft's simulation URLs in any other security tools that scan links, so clicks are real.
• Make reporting easy with the built-in Report button, and keep real phishing out of inboxes. iBlockSpam filters phishing and spoofing at the MX; see how it works.

Browse our other setup guides.