Gmail status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Gmail reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Gmail users through our website.

• Website Down (40%)
• Errors (32%)
• Sign in (28%)

Live Outage Map

The most recent Gmail outage reports came from the following cities:

City	Problem Type	Report Time
Montpellier	Errors	15 hours ago
Paris	Website Down	15 hours ago
Évry	Sign in	1 day ago
Lubbock	Website Down	1 day ago
Cardiff	Errors	2 days ago
Grièges	Errors	2 days ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Gmail Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Steven Pu (@reedvoid) reported 38 seconds ago

  @pitdesi Summer of 2004 I built an app that used gmail attachments as cloud file storage, I called it Goggles. I showed it to a few classmates who worked at Google, they told me to shut it down immediately b/c I'll get sued. Keep wondering what've happened if I didn't listen.

• 
  Mo Syed (@msyed_) reported 2 minutes ago

  Google Drive is getting smarter. 1/11 AI’s biggest problem is not intelligence. It is context. Your files are in Drive. Decisions are buried in emails. Meetings sit in Calendar. And the real process lives partly in people’s heads. Google’s new Projects feature points to what AI work is supposed to look like next 🧵👇 2/11 Google Drive has been a digital dumping ground for years. Docs. PDFs. Invoices. Spreadsheets. Research. Random files you swear you will organise later. The issue was never storage. It was finding the right context at the exact moment you needed it. 3/11 Projects lets you create a dedicated workspace around one specific outcome. Instead of asking Gemini to trawl your entire Drive, you add the files that matter to one project. Then return to that same context whenever you need. No re-explaining the brief. No pasting the same documents into every new chat. 4/11 Say you are building an e-commerce site. Meeting notes are in one document. Product information is somewhere else. Website copy is in another file. With a project, you can ask: “What are the main design themes across these collections?” Or: “Pull every necklace description into a structured list.” That is a much better workflow than hunting across tabs. 5/11 It gets more useful when Gmail and Calendar join the picture. You can ask Gemini to find a relevant quote in an email or check whether a project meeting is coming up. Drive, Gmail, and Calendar stop being separate places you have to manually search. They become one connected context layer. 6/11 There are limits. For example, Projects may be able to find an upcoming Calendar event without being able to create one for you. But that is still the direction of travel. AI is moving from a blank chat window into the tools where work already happens. 7/11 The winning question is no longer: “What can this model do?” It is: “What information does it need, what can it access, and what should it be allowed to change?” More access makes AI more useful. It also makes permission design, privacy, and verification much more important. 8/11 That need for verification is now spilling into public policy. AI-generated misinformation is being used in fights over solar and wind projects. False claims about toxic smoke, chemical leaks, property values, wildlife deaths, and even “mini tornadoes” are being generated, shared, and repeated in real-world planning debates. 9/11 This is how the misinformation pipeline works: Generate convincing nonsense cheaply Spread it through social media Real people repeat it It starts sounding familiar Decision-makers face public pressure By the time someone checks the claim, it may already have shaped the outcome. 10/11 Sam Altman says AI still has not had its iPhone moment. Meta reportedly paused a plan to reshape teams around AI agents after internal pushback. Both point to the same reality: Getting AI into real workflows is much harder than a polished demo makes it look. The bottleneck is not only the model. It is the messy human system around it. 11/11 That is why Forward Deployed Engineers are becoming important. They work inside an organisation, learn the exceptions, the broken processes, and the unwritten rules, then adapt AI around the real work. AI adoption will not come from replacing whole teams overnight. It will come from people who can connect powerful models to trusted context, clear permissions, and the way work actually gets done.

• 
  Luck (@LuckLedger) reported 2 minutes ago

  @blessedby__ @GoogleWorkspace @Em0517561932234 imo this would make the implication even worse. Gmail forces ‘Gemini Flows’ and automation down my throat constantly, yet they don’t use their own tech? Seems dumb for one of the biggest AI companies in the world to have a standardized reply that’s completely irrelevant, instead of y’know, using their technology to automate a relevant reply.

• 
  Shripal Gandhi (@ishripalgandhi) reported 6 minutes ago

  This has become a recurring problem. This has happened earlier too (maybe 2 months ago?), and its happening again. Entire businesses run on Whatsapp (I am NOT a proponent of that!) and this kind of an advance hoc review/hold can bring businesses to a standstill! Solution: 1. Move back to good ol' email. 2. For IM and VCs, use Meet/Teams/Zoom. 3. For channels, try Slack. 4. For OTPs, use SMS 5. Want to save on call charges then use the voice call feature on Meet/Teams. There is nothing unique that Whatsapp offers that your life cannot get by without. It's just that we are way too deep in a comfort zone and Zuck is exploring that to the fullest! If you have a damn smartphone then you sure as hell have a Gmail account or you wouldn't have been able to download Whatsapp. So move your arse out from WhatsApp and explore other options.

• 
  Prashant kumar (@Prashan88148198) reported 6 minutes ago

  @Google Please help me my gmail account is comprised by parent control hyjacking. Give any solution of this compromise service break down

• 
  Nate Tucker (@natetucker) reported 6 minutes ago

  Gmail search is broken. We're building an AI assistant which lets people talk to it. It tries to look up contacts by the name it hears. 60% of the time it failed. The reason it failed was because we were learning on Gmail search. Gmail's search doesn't understand that words can sound alike. It only matches from the start of a word, caps how many results it returns, and doesn't sort by how relevant a result actually is. Somehow it gets worse, because the old matching would grab any word containing the letters you typed. Searching for "Dr. Stevens" could match the "dr" hiding inside "Sandra" or "Pedro," and one of those wrong matches would often win. We got fed up and built our own directory instead.

• 
  sadecatgirl ~ she/they/them (@iAshleyxoxo) reported 7 minutes ago

  @Mlfg_AK @TutaPrivacy yeah :( i rather keep it organized but still have the smart features off cuz of ai scanning…my main email is apparently getting shut down so i have to move to gmail

• 
  Patrick Lu (@lightninglu10) reported 10 minutes ago

  @jjacky Yes, very easy. Use Google cloud, add Gmail api, add Gmail mcp or equivalent, may require you to add the secret json somewhere, oauth and login, but now ai has your inbox. If you want to give the agent its own inbox, just do the same thing

• 
  Gabriel Mican (@youwishmusic) reported 12 minutes ago

  @cartierfamilyZ Why are people so pressed about data centers all of a sudden? All of your online information from X account, to gmail, any online storage, website etc is stored in a data center. We've had data centers for decades, all of a sudden they're an issue? Take a deep breath, and tune out the idiotic noise.

• 
  anshuman (@athletic_coder) reported 19 minutes ago

  okay, so i was diving into the rollouts generated by qwen3.5 4B, here are some findings (given model has access to only api_search and api_fetch tool at any given point of time) > model is looking to gather information it needs to address query but ends up burning tokens and exhausts upper cap of 50 turns in getting a query that actually returns functions that can help model clear ambiguity. > in the partially succeeded rollout (score this way was 17), where the model did discover the correct tool, it would eventually fail to solve that particular task, eg. - (sales subset) example id 504: Qwen found and used the correct Salesforce update endpoint. It updated the wrong Marcus Wong. - (sales subset) example id 503: Qwen found and used the CampaignMember endpoint. It enrolled prohibited people because it skipped the policy. - (sales subset) example id 1204: even knowing the Facebook photo endpoint does not tell it which image is eligible; it must retrieve Gmail policy + Sheet rows first. so my takeaway is that entire problem breaks down to two subproblems: 1. tool discovery (has to be token efficient) 2. the core policy i.e. Prompt > infer what information is missing > search for a capability to retrieve it (basically 1st subproblem) > call that capability > interpret returned observation > identify what remains uncertain > retrieve more evidence > choose target(s) > execute minimal writes > verify end state > finish the moment we break the problem, it is clear that it is a variant of computer use problem. the model needs to learn the ReACT loop. that's it. that's the optimization objective.

• 
  MostlyHarmless (@harmless_human) reported 20 minutes ago

  @rallynate @kbean511 There has probably been a data center near most of them for years and they didnt know. Where do they think netflix and youtube video is stored and all the photos on grandma facebook and gmail account they never delete anything from? The real issue is the mega computer centers running AI programming that shouldnt even be called data centers.

• 
  JovanHuttonPulitzer™ אני לא סובל טיפשים! (@JovanHPulitzer) reported 25 minutes ago

  @stelzner_n1150 This has to be the stupidest comment ever sorry. 100% of cell phones, smart phones, cable boxes, electric meters, computer usage and email GOES THROUGH DATA centers and always has! Here are the facts: That claim is backwards. The internet and smartphones have always run on data centers (or the same thing under older names: server rooms, colo, hosting farms). What changed is the size, location, and branding — not whether the machines exist. The internet never ran without servers Commercial internet use took off in the early 1990s. From day one it needed: ISP modem banks and backbone routers Web/email/DNS servers Co-location and hosting facilities Those are data centers. Exodus, AboveNet, Equinix, and telco “carrier hotels” were already packing raised-floor rooms with racks in the mid-to-late 1990s. Google’s first big custom plants and Amazon’s early AWS rooms were 2000s expansions of that model, not a new species of building. You could run a static homepage off a PC in a closet. You could not run search, streaming, maps, payments, or “the internet” as people actually used it. Smartphones made the dependence worse, not optional The iPhone shipped in 2007. A phone is a thin client. Almost everything people treat as “the phone” lives off-device: iCloud / Google account sync Maps and traffic App Store / Play Store Gmail, Photos, Drive Instagram, TikTok, YouTube, Spotify Siri / Assistant Banking and payments That traffic hits someone’s data center. If those rooms went dark, the glass slab in your pocket would still make calls on a good day and not much else. Twenty years of smartphones is twenty years of more backend compute, not twenty years without it.

• 
  Shanica North (@KickAssShanica) reported 28 minutes ago

  @ArcyloOfficial Get comfy! For me, my Gmail is a connector. This is OAuth into my inbox. Grok can: • search and read mail (body, headers, attachments) • draft replies • send / reply / forward if you grant write/send • label, trash, organize Base hook is often read-only. Send is an extra permission you click on purpose. If you connect it, the bot is sitting in the same box as bank alerts and 2FA codes. That is the whole risk. You can revoke anytime. Grok Bot can also skip my inbox and get its own address (AgentMail / similar plugins). Then it sends and receives from something@….agentmail.to, not from you. I use that if I want an agent that emails people without reading my personal mail. My GitHub OAuth into the GitHub user I sign in as. With the scopes I approve it can: • read public and private repos that account can see • search code, list branches, summarize PRs • open/update issues • create branches, push files, open/review/merge PRs • delete files if write is on Private repos work only if I granted repo (or equivalent) at connect time. Safer pattern: tell it to branch + PR, not push straight to main. Same revoke page. What it cannot do by default • It does not get your password. • It does not stay logged in if you disconnect the connector. • It does not magically see my GitHub orgs I never authorized. • Connecting email does not connect GitHub, and the other way around. Practical rule for me Do not hook personal Gmail if that inbox has 2FA and money mail unless you want an assistant reading it. GitHub is useful if I chose to still keep repos, ask it to show the diff before any write. If you only wanted “what does this button do,” that is the button: it is not a viewer badge. It is a key you can take back. This is what I’m experiencing with learning to use it. It’s different and I’m starting to like it.

• 
  AncienCarnelian (@AncienCarnelian) reported 28 minutes ago

  @krainboltgreene @Speedkicks @slime_machine Unfortunately, it's basically impossible to create your own email server. The software and hardware is easy enough to get setup, but the major email service providers (Microsoft Outlook, Google Gmail, contracted emails through those 2) block all but a shortlist of providers.

• 
  Pers (@0xPerser) reported 33 minutes ago

  Gave Grok Bot a real assignment: run parts of my business like an actual employee, not a chatbot answering questions. First mistake, immediately corrected. I almost had it spin up its own Gmail account — Google's already catching and disabling Gmail accounts created by agents like this. Switched to AgentMail instead, since it's already a listed plugin in Grok Bot's browser. OAuth, three minutes, 24 email tools live and working. For payments, Stripe has a live MCP server — the agent can create customers, issue invoices, run transactions, all through plain language, no custom integration built by hand. Before any of it touched something real, I wrote the rules first. What it's allowed to read. What it's allowed to send. What it's allowed to spend, and the exact caps. What needs my approval before it happens. One command that shuts the whole thing down if I need it gone. Passport first. Rules second. Never the other order. The mistake almost everyone makes isn't giving an agent too much access. It's giving it access before the rules exist to contain it.

• 
  Joewest (@masterverse2) reported 36 minutes ago

  @TheOriks @maxvayshia It's much better if you sign in Just use "signin with gmail" That way, you get music suggestions tailored for you

• 
  Tony Tong | Founder | Ancient Systems x AI (@tonytonggg) reported 37 minutes ago

  @wolframs91 Watching an AI dig in for two hours over an SSH setup is its own kind of unglamorous. My Nanobot instance sits on top of MiniMax 2.5, and wiring in a Composio server purely for Gmail access ate its own stubborn afternoon. Tools cooperating is never the easy part.

• 
  SPtheToker (@spthetoker) reported 37 minutes ago

  @Google @googlemaps I think we need another company thats not US based to start running maps. Whole company thinks they own the maps. You own jackshit. Bring back Askjeeves or ******* yahoo. **** @Google AND STOP GIVING ME THIS STORAGE ******* ERROR ON GMAIL. ****.

• 
  Altr (@altrhorizon) reported 37 minutes ago

  @XFreeze the recovery piece stands out to me most. for longer tasks across multiple PDFs + MCPs like Gmail, Notion or Google Drive, losing progress to a token or network issue would be rough. that kind of resilience matters a lot more as agent workflows get longer!

• 
  berg@robert weisberg (@robertw477) reported 39 minutes ago

  @argingerigorian I merely created an email bot to help reclassify, label and sort Gmail into folders. In 90 mins I was down. Really weird considering what I was doing.

• 
  SYNTHLEX (@SYNTHLEX_) reported 39 minutes ago

  xAI just handed Grok Bots the right to read, write, and act inside your Microsoft accounts. The timeline is fighting about whether that is safe. The real story is a single permission scope that almost nobody clicked through to read. Grok's Outlook connector asks for two things at sign-in: Mail.ReadWrite and Mail.Send. Full send authority from the first Accept. Grok's Gmail connector, by comparison, starts read-only and only earns send access when you turn it on. Read that again. The Microsoft account got the more permissive default. That is not a footnote. That is the exact line where an assistant becomes an employee. Now look at the screenshot everyone scrolled past. A left sidebar of named workers. Schedule Manager. Inbox digest. Bug triage. Expense digest. Launch post. That is not a chat window. That is a staff directory. First. The unit of AI work stopped being the prompt. You are not typing questions into a box, you are assigning roles to a roster. The interface is an org chart wearing the skin of a messaging app. Second. The permission is the product. For two years the bottleneck was never capability, it was trust. A model that drafts a flawless email but cannot send it is a brilliant intern who still needs you to press the button. Mail.Send deletes the button. The value was never in the draft. It was always in the authority to act without you standing there. Third. Each bot runs on its own cloud computer that keeps working after you log off. Persistence changes the verb. A tool waits for you. An employee works while you sleep. The 9:41 AM digest in that screenshot is a report on labor done in the hours the user was gone. Fourth. OneDrive and SharePoint access is gated to Business and Enterprise tiers, and setup routes through an Azure admin consent. xAI is not selling you a chatbot. It is selling your IT department a workforce. The Copilot war will not be won on model quality. It will be won at the admin approval screen. Fifth. The skill this rewards is not prompting. It is delegation. Deciding which bot gets Mail.Send and which stays read-only. Knowing what "make it 45 min instead" means when a human types it to a manager instead of a machine. The person in that screenshot is not a user. He is running a standup. The model can already do the work. The only decision left is what you sign it the authority to do without asking you first. That was never a prompting problem. It is a management one.

• 
  Corrupt.sui (@Corruptlabss) reported 41 minutes ago

  @conso_xyz Yup I have it But it is asking to login with Gmail and not with X

• 
  komocode (@komocode) reported 41 minutes ago

  @Pirat_Nation Or maybe just login every now and then. Gmail deletes your email after 2 years of inactivity. Standard practice.

• 
  A Mithra Is Fine Too (@Mithra_Is_Fine) reported 44 minutes ago

  @x_PuraVida_x @IntCyberDigest The age of spoofed email is basically over with near universal DKIM and SPF. You can’t even get delivered to gmail without those, and if you have them setup no real mail server will deliver a spoofed email from your domain.

• 
  JC Christopher (@JCChristopher) reported 45 minutes ago

  @RealMarkHe @bot I'm a Grok Heavy user, so honestly I have not had any issues. But I will acknowledge that I am not using it for certain things that take a lot of tokens: I am not programming a website or application, building a database, or cleaning up my Gmail (I'm retired, so I don't give a damn about email, one of the perks of retirement). I am using Bot the way 90% of the public would likely use it, as employees and confidants.

• 
  Tony Tong | Founder | Ancient Systems x AI (@tonytonggg) reported 45 minutes ago

  @Mikadzyki_NFT Handing off a shift to a stack of workers only holds up if the plumbing actually connects. My own Nanobot instance runs on MiniMax 2.5, and getting a Composio MCP server wired in just for Gmail access ate a stubborn chunk of my week.

• 
  Sherif Oyedeji (@sherif_oyedeji) reported 49 minutes ago

  Your business can be doing real work with real clients and still lose a prospect tonight because they searched your name and found an old number, an empty Instagram, and a personal Gmail address. They won't tell you. They'll just go quiet. We fix that at ORAC Partners.

• 
  Austin (@IamAroke) reported 51 minutes ago

  @Harshsharma836 The issue isn't really about Gmail specifically being too broad, it's the leading % in the LIKE pattern. Even a rare domain would still force the same full scan since the wildcard sits at the start

• 
  Mindy (@Mindy43087085) reported 57 minutes ago

  @paulickreport Why are you now requiring a gmail account to comment, to even read comments? Please go back to just any email, or make a login/sign in for your specific site.

• 
  Josephoba-Omega Network N-3 Ambassador (@Earl011289) reported 58 minutes ago

  @landoan2003 @ScaleNetworkAI You don't need code. Use your Gmail address to login and register okay