September 2026 Ad Platform
Reliability Report
Nine incidents across Google, Microsoft, Amazon, Shopify and OpenAI in the first half of September. Meta logged zero - one year to the day after the outage that opened last year's Q4 cluster.
The Numbers
Source: AdStatus monitoring, September 1-15, 2026. "110h" is the total delay window Microsoft announced across its four reporting incidents.
Meta's Status Page Has Been Empty for 25 Days
Meta's last logged Ads Manager incident was August 21, 2026. Between September 1 and September 15, its status page recorded nothing at all.
That is worth flagging because of what happened last year. On September 15, 2025, Meta logged an Ads Delivery outage. It was the first of eleven incidents in the eight weeks that followed - the pre-BFCM cluster that made Q4 2025 the worst stretch of Meta reliability we have on record.
| Date | Service affected |
|---|---|
| Sep 15, 2025 | Ads Delivery |
| Sep 24, 2025 | Ads Reporting |
| Sep 24, 2025 | Ads Creation and Editing |
| Oct 9, 2025 | Ads Delivery |
| Oct 16, 2025 | Ads Creation and Editing |
| Oct 21, 2025 | Ads Delivery |
| Oct 23, 2025 | Ads Delivery |
| Oct 27, 2025 | Ads Delivery |
| Nov 6, 2025 | Ads Delivery |
| Nov 7-8, 2025 | Ads Creation and Editing |
| Nov 10, 2025 | Ads Delivery |
Meta Ads Manager incident history, September 15 - November 10, 2025. Source: metastatus.com/ads-manager/history.
Read this as one of two things
Either Meta genuinely had a clean first half of September - which would be a real improvement on 2025 - or its status page is not reporting what advertisers are experiencing.
We have measured the second failure mode directly. During the July 16, 2026 Ads Delivery outage, Meta's own dashboard showed the incident at 09:40 UTC while its RSS feed did not publish the entry until roughly 11:11 UTC - a 90-minute gap in which every RSS-based monitor, including ours at the time, saw nothing. An empty status page is evidence about the status page, not proof about the platform.
Key Findings
Microsoft announced 110 hours of delayed data in two weeks
Four incidents, all reporting delays: 3 hours on general reporting, 77 hours on Quality Score, 21 hours on Share of Voice, and 9 hours on Offline Conversions. Every notice carried the same line - "This is only affecting the reporting and there is no impact on the Ad delivery." Your ads ran. Your numbers were wrong.
September's failures were measurement and billing, not delivery
Of the nine incidents, seven hit reporting, attribution or payments: Microsoft's four delays, Amazon's sponsored ads reporting interruption, and Shopify's Klarna gateway degradation. Only Google's Ad Manager DAI incident and Shopify's storefront outage touched live serving. Advertisers could spend. They could not see, or in some cases could not get paid.
ChatGPT Ads Manager logged its first outage on our monitor
On September 15 at 16:54 UTC, OpenAI resolved an incident titled "Ads Manager login issues," with Ads Manager listed as the affected component. ChatGPT Ads moved to self-serve access across India, Europe, the Middle East and North Africa this month. A new ad platform has joined the outage inventory, and it broke in its first weeks of wider availability.
Complete September 1-15, 2026 Incident Timeline
Google Ad Manager DAI disruption. Affected users could reach Ad Manager DAI but saw error messages, high latency and unexpected behaviour. Partially restored by the time Google posted its update.
Reporting data delayed. Reports, UI and API could return stale or inaccurate data. No ad delivery impact.
Sponsored ads reporting interrupted. Opened in Europe, the Middle East and North Africa, then downgraded to degraded across North and Latin America at 02:22 UTC on September 2. Amazon confirmed no campaign metrics were lost.
Quality Score delayed up to 77 hours. The single largest announced data delay of the period. Quality Score feeds bid decisions, so three days of stale values is not a cosmetic problem.
Klarna payment gateway degraded. Checkouts using Klarna were affected, along with related performance issues. If you were driving paid traffic to a Klarna-enabled store, conversion loss here looks exactly like a creative or audience problem in your ad reporting.
Share of Voice data delayed. The third Microsoft reporting incident in 48 hours.
Admin, storefront and checkout errors. A backend service issue produced errors for some stores. Fix implemented at 09:31 UTC, monitoring closed at 11:04 UTC. A full hour of broken checkouts on a Sunday morning.
Offline Conversions data delayed. Offline conversion lag distorts exactly the campaigns that rely on it most - lead gen and anything with a sales cycle longer than a click.
Ads Manager login issues. OpenAI listed Ads Manager as the affected component and confirmed full recovery. The first ChatGPT Ads incident to appear in our monitoring.
Patterns & Takeaways
Microsoft's reporting pipeline is now a standing risk, not an incident
We flagged the same pattern in the March 2026 report: seven Microsoft incidents that month, all data delays, zero outages. Six months later nothing has changed, and the delay windows are getting longer. A 77-hour Quality Score lag means bid decisions made on Wednesday were using Sunday's data. If you run Microsoft Advertising, stop treating reported numbers as same-day truth and build a buffer into any automated rule that reads them.
Payment-layer failures look like ad failures in your dashboard
Shopify's Klarna degradation on September 2 did not touch a single ad platform. But for any store sending paid traffic to a Klarna checkout, it shows up as a conversion rate drop with no obvious cause - the kind of signal that gets a working campaign paused. When conversions fall and delivery metrics look normal, check the checkout before you touch the campaign.
The ad platform list just got longer
ChatGPT Ads went self-serve in India, Europe, the Middle East and North Africa this month, and logged an Ads Manager login incident on September 15. Every new channel is a new status page, a new failure mode, and a new thing nobody is watching yet. Budget moving into ChatGPT Ads carries the same blind spot early Meta advertisers had in 2012.
The Q4 window is open
Last year Meta's incident run started September 15 and did not let up until November 10. If 2026 follows the same curve, the next eight weeks are when it shows. Right now Meta's page is clean - which is either good news or an early-warning gap. Either way, the second half of September is the part worth watching.
What Advertisers Should Do
- Re-pull your Microsoft Advertising numbers for September 1-4 and September 13-14. Quality Score, Share of Voice, reporting and offline conversions were all delayed in those windows, and anything you decided from them was decided on stale data.
- Check any automated bidding rule that reads Microsoft Quality Score. A 77-hour delay is long enough for a rule to act three times on the same stale value.
- If you sell through Shopify with Klarna enabled, reconcile September 2 conversion data before you judge any campaign that ran that day.
- Do not read Meta's quiet fortnight as an all-clear. Compare your own delivery and spend curves against the status page, and treat a divergence as the signal.
- If you are testing ChatGPT Ads, add status.openai.com to whatever you already watch. It is a live ad platform now, with live incidents.
Methodology
All incident data comes from AdStatus monitoring of official status sources: metastatus.com for Meta, ads.google.com/status for Google, status.ads.microsoft.com for Microsoft Advertising, status.ads.amazon.com for Amazon Ads, shopifystatus.com for Shopify, and status.openai.com for ChatGPT. Feeds are polled every 5 minutes and each incident is stored with its canonical ID, detection timestamp and severity.
Counting rules: the nine incidents are distinct incidents first detected between 2026-09-01 and 2026-09-15 inclusive, on ad, commerce and AI-advertising platforms. Multi-update incidents - Amazon's two-region report and Shopify's three-stage September 6 entry - are counted once each. The 110-hour figure is the sum of the maximum delay windows Microsoft stated in its own four notices (3 + 77 + 21 + 9); it is announced exposure, not measured downtime.
The 2025 comparison table is taken from Meta's published Ads Manager incident history. All times are UTC.
Get the Next Incident in Your Slack
Stop checking status pages. AdStatus monitors Meta, Google, Microsoft, Pinterest, Amazon, Shopify and the AI platforms every 5 minutes and sends Slack or Teams alerts the moment something breaks. 7-day free trial.
Add to Slack - Free for 7 DaysRelated Reports
Meta Ads Delivery Outage July 16: Why Most Monitoring Missed It
The measured 90-minute gap between Meta's dashboard and its RSS feed
March 2026 Ad Platform Reliability Report
12 incidents, and the first time we flagged Microsoft's reporting pipeline
Meta's Black Friday Crisis
What the Q4 2025 incident cluster cost advertisers