Janitor AI Down Explained: The November Outage, the Domain Hold and What to Check
Key Takeaways
- Janitor AI really did go down — and it was not a server crash. On 19 November 2026, the domain janitorai.com was placed on registry "client hold" through its registrar, Namecheap, with the record last changed at 10:46 UTC. A client hold removes a domain from DNS, so the site stopped resolving for everyone.
- The outage took down the website, the app's backend, the official status page and the help centre at the same time, because all of them live under the same domain.
- The Janitor team said on Discord that the domain had been affected by what they considered an erroneous abuse report, and a banner on the service described the hold as the result of a false abuse report, explicitly noting it was not a hack.
- The disruption lasted roughly 15 hours. The Android app was reportedly working again in the early hours of 20 November, and the team announced a backup domain, jani.com, for users who still could not reach the main site.
- Searches for "janitor ai down" rose 2,400% and "is janitor ai down" rose 1,750% over the past seven days worldwide, according to Google Trends data recorded on 21 November 2026.
- As of the morning of 21 November 2026, independent monitors indicate the service is reachable again — UptimeRobot's latest check found no issues, and HostTracker recorded no incidents in the previous 24 hours. But no fresh official confirmation of a full restoration of janitorai.com was found at the time of writing, so treat the "back" verdict as monitor-based, not team-confirmed.
What Janitor AI is — and why so many people noticed it was gone
Janitor AI is an AI-powered chatbot platform that lets users create and interact with virtual characters for roleplay, storytelling and creative conversations. It grew very quickly, and its audience is large and unusually engaged: people maintain ongoing conversations with characters they have built or chosen, and they return daily. When a service like that disappears without warning, its users do not wait politely. They search.
Understanding the outage also requires understanding how Janitor AI is built, because that architecture is what makes its outages confusing. The site hosts the characters and the conversations, but the actual text generation is frequently produced by a model backend the user configures themselves. Janitor has its own model, known as JLLM, but many users connect their own API keys against third-party providers — OpenRouter, DeepSeek and others — through a proxy setup. That means there are at least three different things that can break: the Janitor site itself, the JLLM model, or the user's own proxy route. A user whose proxy replies fail while JLLM works has a provider problem, not a Janitor outage. Only when the site will not load at all, on the web and in the app, on two different networks, is it most likely a platform outage. Most Janitor incidents in 2026 were partial — one model, one feature, one route — and the November event was the exception: everything went at once.
What happened on 19 November 2026
The timeline of the November outage, pieced together from the public registry record, independent monitors and the Janitor team's own announcements, runs as follows. Times are given as reported by the sources cited; where sources disagree on the precise minute, this article says so rather than picking one silently.
The root event was a change to the registration record for janitorai.com. The public .com registry record shows the domain on "client hold", set through its registrar, Namecheap, with the record last changed at 10:46 UTC on 19 November 2026. That fact is verifiable in the Verisign RDAP record for the domain, and it was checked and reported by Companion Scout on the day. A client hold is a registrar-level block: it removes the domain from DNS, so browsers report that the site does not exist. It is not a server crash, not a hosting failure, and not a case of the domain expiring — the registration itself runs to 2030.
Independent monitoring picked up the failure shortly afterwards. HostTracker recorded disruptions beginning at 12:16 UTC on 19 November and a single incident lasting 823.6 minutes, or roughly thirteen and three-quarter hours. StatusGator recorded 9,074 user-submitted outage reports in the 24 hours around the event and logged a major outage on the official status page at 11:05 AM on 19 November. Entireweb recorded 3,299 user reports in 24 hours as of 20 November, with none in the final hour it displayed — a pattern consistent with an outage that had peaked and was fading.
The Janitor team announced a service disruption at about 9:45 p.m. WIB on 19 November (Western Indonesia Time), according to the timeline compiled by Bittime — which corresponds to the early afternoon in UTC — and stated on Discord that the domain had been affected by what they considered an erroneous abuse report. A banner shown on the service, captured in a video update posted on 20 November, put it more directly: janitorai.com had been placed on hold due to a false abuse report, and the banner explicitly noted that this was not a hack. That distinction matters, because a domain hold caused by an abuse report and a security breach are very different events with very different implications for users' accounts and data — and in the first hours of an outage, users reasonably fear the second.
Recovery came in stages rather than all at once. The Android application was reportedly functioning again at about 12:33 a.m. WIB on 20 November, and a backup domain was announced to users at about 1:30 a.m. WIB on 20 November: jani.com, presented as an alternative address while recovery continued. As of the morning of 20 November, monitoring indicated that the primary domain was still disrupted, and there was no confirmed timeline for restoring janitorai.com. A video update posted on 20 November described the outage as lasting approximately 15 hours and the service as back — but the sources found at the time of writing do not pin down a single moment at which the main domain was fully restored for everyone.
Why a search spike is not the same as a confirmed outage
It is worth pausing on the method here, because "janitor ai down" is a search query, and a search query is evidence of people worrying — not, by itself, proof that a service is down. Search interest can spike for many reasons: a rumour, a viral post, a problem affecting one region. A responsible outage article has to distinguish the signal from the confirmation.
In this case, the confirmation exists from several independent directions at once. The registry record itself shows the client hold, with a timestamp. Two independent monitoring services — HostTracker and StatusGator — recorded the failure from their own probes and from thousands of user reports. The Janitor team itself announced a disruption and named the cause. And UptimeRobot later showed the service reachable again, consistent with a real outage followed by a real recovery. The search spike and the confirmed outage agree. A spike alone would not have been enough to write this piece; the piece was written because the underlying event is documented.
The same standard cuts the other way for the question users are actually asking now: is it back? Here the evidence is thinner and more mixed, and this article does not smooth that over. UptimeRobot's page for Janitor AI, checked on 21 November, states that its monitoring probes can reach janitorai.com and complete a full check without errors, with the latest automated check run at 01:58 GMT on 21 November 2026 from North America detecting no unusual response times or error codes. It shows the service as up for 23 hours and 54 minutes, with 99.961% uptime over the last 24 hours. HostTracker's live probes likewise show HTTP, ping and page-speed checks all returning OK, with zero incidents and zero minutes down in the last 24 hours. Entireweb, as of 20 November, recorded no reported problems in the last hour or the last 20 minutes, after 3,299 reports in the preceding 24 hours.
Those monitors point in the same direction — the service is reachable again — but they are monitors, not the Janitor team, and none of the reporting found at the time of writing includes a fresh, timestamped statement from Janitor confirming that janitorai.com is fully restored for all users. It is also possible that some users still see problems while monitors see a healthy site, because DNS changes propagate unevenly or cached records linger.
How to check Janitor AI's status properly
The November outage exposed a practical problem: most users' first instinct when a site fails is to refresh, clear their cache, reinstall the app or switch on a VPN. None of those steps can fix a domain-level hold, and during this outage none of them did. A better check, drawn from Janitor's own help guidance and from the status trackers that covered the event, runs in this order.
First, check the official status page. Janitor's help centre points users to status.janitorai.com, which tracks two components — Backend and Frontend. When it was checked on 25 October 2026, both were marked operational, with 45-day uptime of 99.962% and 99.987%. A green status page does not prove every model is healthy, but it is the authoritative first check — and during the November outage its unreachability was itself a signal.
Second, read the changelog. Janitor posts short incident notes there, such as "jllm was having a bunch of errors earlier today". It is the best place to tell a model-specific problem from a platform-wide one.
Third, try the website and the app separately. If one works and the other does not, it is probably not a full outage. During the November event both failed together, which is what marked it as platform-wide.
Fourth, switch networks — for example, from Wi-Fi to mobile data. If the failure follows you across networks, it is not your connection.
Fifth, test JLLM without your proxy. Start a new chat on Janitor's own model. If that works, the problem is your proxy route — your key, balance, model ID or provider — not Janitor. The same logic applies in reverse: if only one bot or one chat fails while others work, it is usually a chat or bot problem, not an outage.
Sixth, check an independent monitor — UptimeRobot, HostTracker, StatusGator or Entireweb — and compare what they report with what you see. If the monitors say the site is up and you still cannot reach it, the problem is local to you: DNS caching, a stale route, or an account-level issue.
A short diagnostic table summarises the pattern:
- Site will not load on web or app, on two networks: most likely a platform outage — check the status page and changelog.
- Pages load, but JLLM replies fail or time out: most likely a JLLM incident — look for a changelog notice about JLLM errors.
- JLLM works, but proxy replies fail: most likely your key, balance, model ID or provider — check your provider's error code and status page.
- Only your account is blocked or redirected: most likely an account, age or moderation action — read the exact notice, then contact support.
What not to do while Janitor AI is down
Outages attract opportunists, and this one was no exception. The November event produced a wave of lookalike domains, "fix" sites and supposed mirrors, and the advice from the coverage is consistent and worth repeating plainly.
Do not paste your API key into "fix" sites or private messages. Nobody legitimate needs your key to restore Janitor. Your key is a credential for your own provider account, and handing it to a stranger's website is how accounts get drained. The same applies to any site that asks you to log in with your Janitor credentials outside janitorai.com, jani.com or the official app — those are the only addresses the team's own announcements have named, and any other address claiming to be Janitor should be treated as a scam.
Do not confuse janitor.ai with janitorai.com. Janitor.ai is a different site — an unrelated writing-tools site — and it is not Janitor AI. Entering your Janitor login or API key there would hand your credentials to a stranger.
Do not delete your account to reset things. Since 13 August 2026, deletion starts a 30-day countdown, and Janitor's terms cancel unused Router credits when you delete. Deleting an account cannot fix a domain hold.
Do not buy janitor+ to escape an outage. Paid capacity has had its own incidents, and a platform-wide outage affects paying and free users alike.
Do not keep regenerating on a paid route during errors. Janitor's terms say Router credits can be used up even when a reply is cancelled or fails after processing begins, and your own provider may bill failed requests too. If replies are failing, stop and diagnose rather than burning credits on retries.
If the status page is green, the changelog is quiet, and the problem survives a refresh, another device and another network, contact support. Janitor's help centre offers live chat and email at support@janitorai.com, with billing@janitorai.com for subscription or payment problems.
Janitor AI's outage history in 2026: mostly partial, occasionally total
The November domain hold was the most severe Janitor incident of 2026, but it was not the first. Earlier incidents in the year included a 15-minute JLLM outage in May after a bad deploy, a dead GPU host the same month, an August spike in JLLM's error rate from about 0.3% to around 8% for roughly 12 hours, a moderation pipeline failure, a memory bug that hid saved memories on about 15,000 chats, and a CDN hiccup in October. The pattern is instructive: most Janitor incidents are narrow — one model, one pipeline, one feature — and the November event is different in kind, not just in degree. Every earlier incident left the site reachable and the status page readable. The domain hold removed the service, the status page and the help centre in a single stroke, which is why it produced a search spike several times larger than any of the earlier incidents. It is also why the usual advice — "check the status page" — failed precisely when it was most needed.
That history also puts the platform's reliability in perspective. Janitor's status page showed 45-day uptime of 99.962% for the Backend and 99.987% for the Frontend when checked on 25 October 2026, and HostTracker records 90-day uptime of 99.364% with a single incident — the November one — accounting for 823.6 minutes of downtime. Those are respectable figures for a fast-growing service, and they are consistent with a platform that is usually up but has a real, recurring tail of partial failures. The November outage does not make Janitor unreliable in general. It makes it a service with a demonstrated single point of failure at the domain level — a failure mode that, once it has happened, is worth asking whether it has been fixed.
What the outage means for users — and for the platform
For users, the practical lesson is about where to put trust. Conversations and characters live on Janitor's platform, and when the domain goes down there is no way to reach them — no export on demand, no offline copy, no alternative route except the backup domain the team announces. Users who care about long-running characters should assume a domain-level failure can happen again, and keep their own notes or exports where the platform allows it.
There is also a narrower lesson about proxies. Many Janitor users bring their own API keys, and an outage that takes down the Janitor site does not take down their provider — OpenRouter, DeepSeek and other APIs were unaffected throughout the November event. But that independence cuts both ways: users whose chats live on Janitor cannot reach them through their provider alone. The provider is fine; the front door is gone.
For the platform, the questions are structural. A status page on the same domain as the service it monitors is a known anti-pattern, and this outage demonstrated why: the page was unreachable at the exact moment users needed it. A backup domain, announced quickly, mitigated the damage — jani.com gave users a route back within hours — but the reporting found at the time of writing does not confirm whether jani.com is now a permanent, documented fallback or a temporary measure. Similarly, the cause — a hold apparently triggered by an abuse report the team considers erroneous — raises questions about registrar-level safeguards that this article cannot answer from the available sources: how the report was filed, what review took place, and what would prevent a repeat.
The broader context matters too. Janitor AI is not the only AI service to have gone down in recent days — our earlier piece on the AI outage wave affecting Microsoft, the FAA and Chase documented a cluster of unrelated outages hitting major services in the same period, and our guide to the ChatGPT "error in message stream" problem covers the kind of partial failure users most often mistake for a full outage.
The bottom line
So, is Janitor AI down? The accurate answer depends on when you ask. On 19 November 2026, yes — unambiguously, for everyone, for about fifteen hours, because its domain was placed on hold at the registry. On 20 November, it was coming back in stages, with the app working again and a backup domain announced, while the primary domain was still disrupted. On the morning of 21 November 2026, independent monitors show the service reachable and report no ongoing incident — but no fresh official statement confirming a full restoration was found in the sources checked for this article, so that verdict rests on the monitors rather than on the team.
That three-part answer is less satisfying than a single yes or no, but it is the honest one, and it is the reason the search spike happened the way it did. People searched "janitor ai down" because the site was gone and the status page was gone with it. They kept searching because the recovery was gradual and unevenly communicated. And the spike is now fading because, by the monitors' account, the service is back.
What We Could Not Verify
- Whether janitorai.com is fully restored for all users as of 21 November 2026. UptimeRobot's check at 01:58 GMT on 21 November found no issues, HostTracker recorded no incidents in the previous 24 hours, and Entireweb showed no reports in the last hour as of 20 November. However, no fresh, timestamped statement from the Janitor team confirming full restoration of the primary domain was found in the sources checked. This article therefore reports the service as monitor-indicated up, not team-confirmed restored.
- The exact minute the outage ended. Sources agree the disruption lasted roughly 15 hours and that the app was working again in the early hours of 20 November WIB, but they do not converge on a single restoration time for the main domain, and HostTracker's 823.6-minute incident figure and the "approximately 15 hours" figure are not identical. This article reports both figures with their sources rather than merging them.
- How the abuse report that triggered the hold was filed, and by whom. The Janitor team described the report as erroneous, and a service banner described it as a false abuse report. No source found names the reporter, the content of the report, or what review Namecheap conducted before applying the hold.
- Whether jani.com is a permanent fallback or a temporary measure. The backup domain was announced during the outage. No source found states whether it will remain available, or whether Janitor has published it as a documented fallback.
- Whether any user data or conversations were affected. No source found reports data loss from the November outage, and the outage was a domain-level block rather than a data incident. But no source found explicitly confirms that no data was affected either, so this article does not claim that.
- User-report totals as a measure of severity. StatusGator recorded 9,074 reports in 24 hours and Entireweb 3,299 in 24 hours, but the two services count reports differently and cover different windows. The figures are reported as each service published them, not as a single combined total.
Sources: Companion Scout, "Is Janitor AI Down? Status Checks and Fixes (2026)" (companionscoutai.com); Bittime, "What Is Janitor AI? Downtime Causes and Latest Links as of November 2026" (bittime.com); UptimeRobot, "Janitor AI Status" (uptimerobot.com); HostTracker, "Janitor AI Status: Is Janitor AI Down Right Now?" (host-tracker.com); StatusGator, "Janitor AI Status" (statusgator.com); Entireweb, "Is Janitor AI down? Live Status" (entireweb.com); CompanionHunt video update on the 19 November outage, posted 20 November 2026; Google Trends, Explore, past 7 days worldwide, "ai" rising queries, as recorded in our research note of 21 November 2026.