Outlook Outage: AI Cited StatusGator Over Microsoft Learn

During the November 12, 2026 Outlook outage, AI engines cited the status monitor StatusGator in 15.3% of answers with retrieval evidence, more than any single Microsoft page. 5W's audit of 288 answers from four AI engines found Microsoft named in 70.1% of answers to Microsoft-specific prompts. The sources behind those answers show why a company's outage update needs to be easy to quote.

By Ronn Torossian, founder and chairman of 5W Public Relations. Last updated: November 13, 2026.

What did the Outlook outage audit find?

The Outlook outage audit found that what an AI engine said depended on which engine answered and which sources it pulled in. On November 12, 2026, 5W ran 24 prompts three times each on Perplexity, ChatGPT, Gemini and a Claude API model, using Profound's tools, and recorded 288 answers. The main results:

• Microsoft was named in 152 of 288 answers (52.8%). That rises to 143 of 204 answers (70.1%) on the 17 Microsoft-specific prompts and falls to 9 of 84 (10.7%) on seven general monitoring prompts.
• StatusGator was the most cited publisher, in 44 of 287 answers with retrieval evidence (15.3%), ahead of AP News at 36 answers (12.5%).
• Microsoft-owned domains were cited in 57 of 287 answers (19.9%) across the full test.
• Perplexity named Microsoft in 79.2% of its answers, ChatGPT and Gemini in 65.3% each, and the Claude API model in 1.4%.
• Only Gemini included Microsoft's incident identifier, EX1550543, in 20 of its 72 answers (27.8%).

What happened during the Outlook outage?

The Outlook outage began on November 12, 2026, when Microsoft said it was investigating reports of users having trouble reaching Exchange Online mailboxes through multiple connection methods in the Americas. According to an Associated Press report carried by WOSU, the company pointed administrators to incident EX1550543 in the Microsoft 365 admin center.

ABC News reported that Microsoft first pointed to infrastructure under elevated load across multiple regions, then said the issue could affect Microsoft 365 connectivity for users in any region.

A customer who asks an AI engine what is happening during an outage gets an answer the company did not write. The audit recorded what four engines returned to that customer.

Why did AI engines cite a status monitor before Microsoft's pages?

AI engines cited StatusGator more often than any single Microsoft page because third-party monitors and news outlets in the test were at least as easy to retrieve and check as Microsoft's own documentation. That explanation is my reading of the pattern. The audit measured the pattern, not the reason for it.

Source: 5W analysis using Profound's tools. Each domain counts once per answer, and an answer could cite several domains.

Microsoft was not absent. Its domains together appeared in 19.9% of answers, and Microsoft Learn alone in 10.8%. But no single Microsoft page was cited as often as StatusGator. For a communications team at any technology company, that is the number to study: the official source was present, and it was one voice among several.

Why did four AI engines tell four different stories?

The four AI engines told different stories because they named Microsoft at very different rates: Perplexity in 57 of 72 answers, ChatGPT and Gemini in 47 each, and the Claude API model in one.

Source: 5W analysis using Profound's tools. Each engine has 72 answers.

The Claude figure needs context. The model in the audit was Anthropic's claude-sonnet-4, called through the API with Anthropic's web search tool, so it does not describe the Claude chat product. The audit counted recorded brand mentions without reading answer text, so it cannot say whether that model wrote "Outlook" or "Exchange" instead of "Microsoft." The gap should not be read as a ranking of the engines.

The broader point holds either way. Apple appeared in 24.0% of all answers and Zoho in 18.8%, mostly as comparisons or alternatives in ChatGPT and Gemini answers. One incident produced several public versions of the story, and the company controlled none of the choices about which version a customer saw.

What did the engines return when people asked what to do?

When the question moved from what happened to what to do next, Microsoft's lead over other brands narrowed. Each prompt below produced 12 answers (three trials on four engines).

• "When will Outlook be back up?" Microsoft led at 83.3%.
• "What should I do when Outlook is down?" Microsoft led at 58.3%, with Apple close behind at 50%.
• "How do I email my team if Outlook is down?" Google led at 50%, with Microsoft and Slack at 41.7% each.

These figures count mentions of any kind, including alternatives and comparisons, so they do not show that engines sent customers elsewhere. They do show that action questions pull other brands into the answer. An update that says only what happened leaves the next-step question open for other companies to fill.

What should a company publish in the first hour of an outage?

A company should publish one timestamped update that serves customers, administrators, reporters and status monitors at once. These steps follow from the source mix above. They are my suggestions based on the data, and the audit did not test whether they change what engines say.

1. State which services are affected and which are not, because a specific scope is easier to repeat accurately than a general apology.
2. Tell users what they can do right now, naming the workaround or alternate access path, since action questions were where other brands gained ground.
3. Give the time of the next update, so a monitoring service or reporter can show when the information was last confirmed.
4. Name the incident identifier and say where administrators can find the technical record.
5. Send reporters and status monitors the same details that appear on the company page, so each source can be checked against the others as the incident develops.

I have written before about how SaaS outages test communications, and the CrowdStrike outage case study covers the same first-72-hours discipline. The audit adds one question to both: if a stranger asks a chatbot about this incident tonight, which sources will it find?

Why does an incident identifier matter during an outage?

An incident identifier matters because it is the one exact, checkable fact that every source can repeat the same way. Microsoft used EX1550543 to send administrators to the official incident record. Only Gemini carried it into its answers, in 20 of 72 (27.8%). ChatGPT, Perplexity and the Claude API model did not include it.

A plain identifier in the page title and the opening line of an update gives an engine something precise to retrieve and repeat. A general statement that the company is investigating gives it nothing to match against. Companies that already run crisis communications services or an internal incident process can add the identifier step to their template at no cost.

What can this audit not show?

This audit cannot show what customers actually saw or did, because it is a one-day snapshot of 24 prompts on four engine configurations. It did not poll users, grade the accuracy of answers, read answer text, or measure how often the public asked each question. Each engine uses a different model, retrieval tool and answer format, so a gap between engines can reflect system design as much as brand visibility.

The full data, with confidence intervals and a prompt-by-prompt table, is on the 5W Outlook outage research page and in Everything-PR's audit report. The next step is a repeat run of the same 24 prompts a week after the outage, to see whether the answers and the cited sources change once the incident is over.

Research disclosure: 5W conducted the audit using Profound's tools. Everything-PR News LLC, which published a companion report, is owned by a shared interest group that includes Ronn Torossian, who founded and chairs 5W.

Frequently asked questions

Which source did AI engines cite most during the Outlook outage?

StatusGator was cited most, in 44 of 287 answers with retrieval evidence (15.3%). AP News followed at 12.5%, according to 5W's audit using Profound's tools.

Did every AI engine name Microsoft?

No. Perplexity named Microsoft in 79.2% of answers, ChatGPT and Gemini in 65.3%, and the Claude API model in 1.4%. The audit did not review answer text.

Were Microsoft's own pages cited?

Yes, in a minority of answers. Microsoft-owned domains appeared in 19.9% of answers, and Microsoft Learn alone in 10.8%.

What should a company publish during an outage?

One timestamped update that names what is affected, what users can do now, the next update time and the incident identifier. This is a suggestion drawn from the data, not a tested result.

Does the audit prove AI changes customer behavior?

No. The audit records what four engines returned on one day. It does not measure what customers did after reading the answers.