How to set up a ChatGPT Dot morning brief: schedule, delivery checks and 20 real daily-brief examples (OpenAI Dots)
Set up a ChatGPT Dot (OpenAI Dots) morning brief in seven steps: name every calendar, ask for a fixed time with time zone and end date, deliver to chat first, then check the task under Scheduled. Eight places briefs get stuck, with fixes, and 20 real daily-brief setups.
The short version: what a Dot can put in your morning brief, how to set it up, where it gets stuck
Is a morning brief worth handing to a Dot?
Yes. A morning brief is one of the easiest jobs to hand a Dot, because you can set it up in one sentence in the conversation. It does four things well in the setups below. It turns your day into a short list: uehara's brief read all 16 of his calendars, merged duplicate events and showed private ones by time only, and Palpa's caught a meeting that had moved an hour earlier. It reads mail overnight so you get one brief instead of a stream of notifications, as 단테랩스 does with important mail from the last three days at 7 every morning. It collects news from sources you name, such as up to five items a day, a weekday brief under 200 words with official sources, or a blog's RSS feed split by plan. And it delivers in the format you ask for, down to まさやん's calendar turned into an image at 7. The weak spots are the clock, the facts and the proof: briefs arrive a few minutes late, one detail can be wrong, and a task marked completed doesn't prove the brief reached you. Start read-only, delivered to the Dot chat, and add more once a few briefs have checked out. Longer stories, from five decisions a day to overnight work reports, are on the morning brief use case page.
How to set it up in seven steps
- Create the Dot in the ChatGPT mobile app, the desktop app or a desktop browser (getting started; Get started with your dot). Connect only what the brief needs: Google Calendar for the day, Gmail for mail that needs you, Google Drive if documents belong in it. Each app is a separate connection, and adding Slack as a way to reach your Dot doesn't give it your inbox (Connect computers and apps; Integrations).
- Try it small, then once in full. Ask for your next meeting and one named email thread, then ask for one complete brief by hand, open every source link and tighten the format. A good result lists tomorrow's account review once even when it appears in both mail and calendar, links the event and the customer thread, and asks you to choose the proposal version without sending mail or editing the event.
- Write the ask in four parts, the way わっしー does: what to keep doing, where to look (name every calendar), what to report, and what not to do yet. He repeats the "no sending, no posting, no editing events" line in the ask even though his custom rules already cover it.
- Set custom rules so any send or calendar change is set to Ask before taking action or Hand off to you (Getting started with your dot; Dots privacy FAQs).
- Ask for a fixed schedule. Your Dot can decide on its own when to follow up, but a morning brief depends on timing, so request a fixed schedule and include the time zone and how long it should run (Get started with your dot). A recurring task needs four things: what to check, when (time zone and any end date), which changes deserve a notification, and where to deliver. Then ask the Dot to confirm what it saved (Tasks and memory).
- Deliver to the Dot chat first. Add a Slack channel or email once a few briefs have come through, and keep sends behind approval. If the brief includes weather or news, name one official page or feed in the ask.
- Check the saved task yourself. Select it under Recent activity in your Dot's profile, or open the Scheduled section in ChatGPT, where you can change its repeat schedule, time or completion notifications (Getting started with your dot). You can also ask the Dot to list, change or cancel its scheduled tasks, including your ChatGPT Work automations (Tasks and memory; what's new; announcement). Stopping work in progress and cancelling a schedule are separate actions, so say which one you mean.
Where it gets stuck, and what worked
- It arrives late. uehara's 8:30 brief arrived at 8:44, opening with "Sorry for the delay", and the Dot had told him up front that it can't guarantee an exact time. What worked: schedule it a little before you need it, and ask every brief to start with the time the run began.
- It misses a calendar. When uehara asked for "events" without the word "all", the Dot read only his main calendar and missed 2 events in another one. Palpa's brief said which calendars showed busy times only. What worked: list every calendar by name and ask the brief to end with what it read and what it couldn't see.
- The task says Completed, but you can't find the brief. Future Stack Reviews couldn't tell whether a 12:46 "No change" message came from its saved 12:40 test, and Takumi Shibata couldn't tell whether his hourly checks kept running. OpenAI's advice is to review the result even when a run is marked completed (Getting started with your dot). What worked: for the first three mornings, note when each brief arrives and compare it with the task's last run under Scheduled.
- One detail is wrong. The AI Frontier found one story out of three that overstated who gets access, わっしー's to-do count was right but one detail number was wrong, and Tina Saul spotted dropped items within two days. What worked: a source link on every item, and a weekly check of two or three items against the original mail or event, counting how long the check took.
- A scheduled send or write is blocked. In early October, ChatGPT Work scheduled tasks that sent Gmail or wrote to Google Sheets ran on time and then had the write stopped by safety checks while reading kept working (tmoiv, gaelto). In gaelto's case no approval prompt appeared. What worked: keep the brief read-only, ask it to report any blocked send with the exact error, or let the AI write a structured record and have a tool you control, such as an Apps Script trigger, do the send (Abiram).
- You end up with two briefs. If the Dot can't see or change a task and you ask for a new one, the old one keeps running. What worked: ask the Dot to list every scheduled task first, and edit the existing one under Scheduled yourself if the Dot's change is refused.
- The cloud browser hits a check. Lolipop's editors watched a weather search land on a reCAPTCHA page and then a Cloudflare "Verify you are human" page. OpenAI notes that some sites block the cloud browser or ask for extra checks (Getting started with your dot). What worked: name one official forecast or news page in the ask.
- It feels like more work than it saves. Desmond Dixon scored his 48 hours 6 out of 10, citing a computer that had to stay on, manual approvals and slowdowns. What worked: build the brief from connected apps and the Dot's own cloud computer, which work without your device, and keep approvals for sends and calendar changes only. Connect your computer only if the brief needs local files, and keep it online with the ChatGPT app open while the Dot uses it (Get started with your dot).
What it can do, and what it can't yet
It can build a daily brief from your calendars, Gmail, Drive, news sites, RSS feeds and public Telegram channels; deliver it in the Dot chat or a Slack channel; turn your calendar into an image; be set up during a phone call (단테랩스, Evgeniya Odud); hold non-urgent items until the next brief because it knows your fixed commitments (@theaiimpact); come back with news between scheduled checks (@lena_podzorova); and list, change or cancel its own scheduled tasks when you ask.
It can't yet promise an exact delivery time, prove from a Completed label that a brief reached you, or count on scheduled sends and sheet writes going through without you. A brief that runs every day also uses your plan's usage every day, as わっしー points out.
One message to start with
Every weekday at 7:30 [IANA time zone] until [end date], send me a morning brief here in this chat. Read [list every calendar by name], Gmail from the last 24 hours, and [one named news page or feed]. Start with one line: the time this run started, the sources you read, and anything you couldn't open or could only see as busy time. Then give today's events in time order, each listed once even if it appears in mail and calendar, the emails that need my reply, and up to three news items, with a source link on every item. Keep it under 200 words. Only read and summarize: don't send, post, archive or change events. If a send or write is blocked, give me the exact error. After you save the task, read back its time, time zone, days, end date and where results go.
Quick answers
How do I set up a morning brief with my ChatGPT Dot?
Ask for it in the Dot conversation, in one sentence: what to check, the time, your time zone, an end date and where to deliver it. uehara did exactly that ("Every weekday at 8:30 a.m., send me the events from all calendars and the emails that need a reply, in 5 lines or fewer"), with no settings screen, and stopping it works the same way. OpenAI Help's own example of a recurring task is a Dot that reviews your calendar each morning; the saved task then shows under Scheduled, where you can change its time and repeat (Getting started with your dot; Tasks and memory). Run one full brief by hand first.
Which apps should it read?
Only what the brief needs: Google Calendar for the day, Gmail for mail that needs you, Google Drive if documents belong in it, and Slack if you want it delivered to a channel. Slack alone doesn't read your inbox (Connect computers and apps). Name every calendar: when uehara asked for "events" without the word "all", the Dot read only the main calendar and missed 2 events in another one.
Will it arrive on time every morning?
Not to the minute. uehara's 8:30 brief arrived at 8:44, opening with "Sorry for the delay", and the Dot had told him up front that it can't guarantee an exact time. Future Stack Reviews couldn't establish whether a 12:46 "No change" message came from its saved 12:40 test, and Takumi Shibata couldn't tell whether his hourly checks kept running. A run marked completed doesn't by itself mean the brief was delivered, so for the first three mornings note when each one arrives and compare it with Scheduled.
Can I trust what's in it?
Check it for the first week. The AI Frontier found two stories right and one that overstated who gets access; わっしー's to-do count was right but one detail number was wrong; Tina Saul saved 30 to 45 minutes of gathering but spotted dropped items within two days. Palpa's brief ended by saying what it hadn't done and which calendars showed busy times only. Ask for a source link on every item and that same closing line.
Can it post the brief to Slack or email it to me?
Posting to chat and Slack worked: 단테랩스 gets important mail from the last three days at 7 every morning in both, and Seowoo Han's #news channel got a five-item briefing at 09:56 KST. Writes are where things went wrong in early October. In ChatGPT Work scheduled-task reports (not Dot reports), tmoiv's Gmail sends and gaelto's Google Sheets writes were blocked by safety checks while reading still worked. Start read-only with delivery to chat, set Ask before taking action for sends and calendar changes, and add email once a few briefs have come through. Your Dot can also read and edit those scheduled tasks for you (what's new).
Which plan do I need?
A Dot comes with Pro ($100, $200 or $500 tiers) outside the EEA, UK and Switzerland, with Business Premium in any supported region, and with Enterprise once an admin switches it on (How to get ChatGPT Dots; in Europe, see Dots in the EU and UK). The full list of what a Dot can do is on the features page. If you'd rather ask for an update by phone or Slack during the day, see AI secretary catch-ups; if the Dot or its brief has gone quiet, start at Dots not showing or not working.
Real daily briefs, grouped by what they pull
Each case links to the person's own post or video, or to the Made with Dots summary of it. Names and handles are as the source gives them.
Calendar
-
まさやん (@masayan-ai-hack) asked his Dot to turn the day's calendar into an image delivered every morning at 7, alongside a Slack channel watch every 10 minutes. He notes that sends and deletes still stop for confirmation, and he added a custom rule of his own: don't send email (YouTube, Japanese).
-
@argofowl chats with his Dot, FowyDot, from a macOS menu bar. Asked "What do I have on my agenda today?", it listed a few emails still waiting for replies, his 1:00 PM meeting, and an open afternoon, then offered to draft the replies (X post, screenshot). Calendar setup: Dots + Google Calendar.
-
uehara (DEV Community, Oct 10) set up a weekday brief in one sentence, in the conversation itself: "Every weekday at 8:30 a.m., send me the events from all calendars and the emails that need a reply, in 5 lines or fewer." There was no settings screen, and stopping it also works by saying so ("Stop the morning summary"). On Oct 8 it arrived at 8:44, opening with "Sorry for the delay." The content checked out: it read all 16 calendars, removed duplicate events, and showed private events by time only, though it ran to 6 lines instead of 5. The word "all" mattered. An earlier on-demand summary, asked for "events" without it, read only the main calendar and missed 2 events in another one. The Dot had also told uehara up front that finishing at an exact time is something it can't guarantee (DEV, English).
-
わっしー (@WASSY_sun, Oct 4) gets one screen each morning with the day's events and to-dos from Google Calendar and his to-do list. His ask has four parts: what to keep doing, where to look, what to report (events in time order and the top three to-dos), and what not to do yet (no sending or posting, no editing events or sheets, no purchases). He repeats that last part in the ask even though his custom rules already cover it, because rules can be misapplied. The to-do count in the summary was right, but one detail number was wrong, so he checks the numbers himself. He also notes that a role that runs every day draws on his Codex usage every day (note, Japanese).
Email and overnight triage
- 단테랩스 set up a scheduled task during a 30-minute phone call with his Dot: every morning at 7, summarize only the important mail from the last three days and send it to chat and Slack. The same video keeps the misses in: the Dot once misreported progress before correcting itself (YouTube, Korean).
- Sam Altman told Dan Shipper on The Every Podcast that his early mornings used to go to whatever fire needed putting out. Now his Dot triages those issues and flags only what's truly urgent (YouTube, Every).
- @theaiimpact says their Dot read every connected app overnight and gave them one morning brief instead of forty notifications. At 4pm it held everything non-urgent, because it knows school pickup is at 3:30 (TikTok).
- @VraserX describes a Dot that briefs him every day, checks his mail and hands him ready-to-send drafts, and keeps an eye on his calendar to tell him what's next (X post). For the inbox side, see email triage cases and Dots + Gmail.
News and research
- IA Para TODOS asked his Dot for up to five important news items every morning, plus monitoring of concert announcements in Brazil and recurring reminders (YouTube, Portuguese).
- Andrew Kononov and his Dot, AltMan, are building a routine: five tech, AI, science, and startup stories every morning, a topic bank for the ideas he picks, and scripts when he's ready to record. His goal is three reels a week (Instagram).
- A2Z AI asked a Dot named Norm for a weekday morning brief under 200 words: three AI updates, official sources, and why they matter. The video shows the request and a real sample, and he says plainly that recurring delivery wasn't verified yet (Threads; YouTube).
- The AI Frontier had a Dot build a morning briefing, then checked every story himself. Two were right, and one overstated who gets access (YouTube Short). That is the reason to ask for source links on every item.
- Sultan AlFifi (سلطان الفيفي) set a daily automation on Saudi time for a news summary. It arrived with a table, images, and links, and he added it to a ChatGPT Space. Calling the Dot didn't work for him yet (YouTube, Arabic).
- Evgeniya Odud set her briefs up by phone from a café: a morning to-do list digest and twice-daily news from her favorite Telegram channels. She used public channels only and didn't connect her own Telegram chats (Instagram, Russian).
- George Liu gave his Dot, Blot, the Cloudflare blog RSS feed and asked for daily Birthday Week summaries split by Free, Business, and Enterprise plans (Threads). For source-watching jobs that run longer than a week, see the research watch workflow.
Weather, and what the cloud browser ran into
- Lolipop's editors (Japan, Pro 100, Sep 30) asked their new Dot to look up tomorrow's weather in Tokyo, the kind of line many briefs open with. Watching its cloud browser, they saw a Google search for "東京 天気 明日" (Tokyo weather tomorrow) land on a reCAPTCHA "I'm not a robot" page, then a weather site show Cloudflare's "Verify you are human". The Dot ended on the Japan Meteorological Agency's Tokyo forecast table. From the screen they couldn't tell whether it had solved the checks or moved around them (lolipop.jp, Japanese). OpenAI notes that some sites block the cloud browser or ask for extra checks (Help), so for weather it helps to name one official forecast page in the ask.
Slack and team channels
- Seowoo Han (@swhan0329_) opened one Slack channel per job and told the Dot, "research the market every day and post a report to this channel." The #news channel got a briefing at 09:56 KST on Oct 1 with five items, each with details and links (Threads, Korean). Setup: Dots + Slack.
- Desmond Dixon ran his business through a Dot for 48 hours, including proactive morning reviews across Slack and email. He scored it 6 out of 10 and canceled for now, citing a computer that has to stay on, manual approvals, and slowdowns with heavy use (YouTube).
Wake-up and timing
- Tiffany Kyazze (@techtiff.ai) put her Dot in charge of waking her up: it now clocks in five minutes before her alarm goes off (Threads). Dots can't place calls to you at launch, so this kind of check-in arrives as a message (Help).
- @lena_podzorova already had a 9am Morning Brief in regular ChatGPT that checks mail and calendar. What she likes about her Dot, Busya, is that it comes back with news between those checks without being asked (Threads, Russian).
When a brief is late, missing or wrong
If a brief doesn't show up, open the task under Scheduled, check the saved instructions and timing, and review the result even when a run is marked completed (Getting started with your dot). A completed run doesn't by itself mean the result was achieved or delivered (Tasks and memory). These four early reports show what that looks like in practice.
- Palpa (Japan, Sep 30) asked its new Dot, named Giorno, for today's and tomorrow's schedule plus recent important mail, cut down to about three things that needed preparing, with no sending and no calendar changes. The answer came in about a minute and a half. It listed the events in Japan time, linked each prep item to the original mail or event, and caught that a meeting tomorrow had moved an hour earlier. It ended with what it hadn't done (no mail sent, no event changed) and what it couldn't see: some calendars only showed busy times (Palpa's write-up, Japanese). That closing line is worth asking for in every brief.
- Tina Saul (KINTAL, Oct 1) asked the Dot, KORA, to go through the calendar, tasks and email during a shower and breakfast. The pack was ready by the time Tina came back. It replaced 30 to 45 minutes of pulling things together by hand and separated commitments that were still open from ones already answered or done. Within two days Tina had also spotted quite a few dropped items. The advice from the post is to count your review time when you decide whether the brief helped (KINTAL).
- Takumi Shibata (DevAIse, Oct 8, Windows app) asked a Dot for an hourly watch on Google Trends for AI terms, first for Japan and then also for the US. One notification with links arrived in the conversation. From the screen, Takumi couldn't tell whether it came from the first check right after setup or from a scheduled run, or whether the hourly checks kept going. A separate watch for important mail got a reply saying it was set up, and nothing after that could be confirmed (AI Agent Lab, Japanese).
- Future Stack Reviews (FSR, Oct 6) asked for a one-time test 20 minutes later and a weekday 09:00 check that would stop after three runs. The Dot confirmed both. The test was due at 12:40, and a "No change" message stamped 12:46 came from the Dot. FSR couldn't establish whether the saved task had sent it. By Oct 9 both tasks were listed as Completed, with nothing upcoming (FSR, section 06). FSR's pause test, where a one-time task was listed as Completed without ever running, is on getting started in the section on stopping your Dot.
When a scheduled send or write is blocked
A brief that only reads and posts to chat is the easy case. Briefs that send mail or write to a sheet ran into a different problem in early October: the task ran on time, and the write was stopped. The five reports below are ChatGPT Work scheduled-task reports, not Dot reports. None of these people says a Dot was involved. They matter here because your Dot can open and edit these same Work tasks.
- tmoiv (Oct 7) had a Gmail workflow that had run for weeks. The tasks fired on time, then failed with "OpenAI safety checks blocked both the verified result batch and required failure-telemetry Gmail send before invocation; requests left pending." They removed and reinstalled the Gmail connector, confirmed that sending worked in a normal chat, recreated every production task from scratch, and tried a clean task with no other connectors. It still didn't send (community post).
- gaelto (Oct 10) has a task that reads connected services and writes the result into Google Sheets. Reading and processing finished. Two writes were blocked, including a smaller fallback of five cells, with "This tool call was blocked by OpenAI's safety checks." No approval prompt appeared, and a read afterwards showed nothing had been saved (community post).
- Daareek (Oct 8) found that ChatGPT's settings showed Allow all actions, while the assistant's own permission tool reported Gmail and GitHub as Use my default, which it read as Allow low-risk actions. Trying to set Allow all actions from there failed with "HTTP 404: Action not found". On the same account, scheduled Gmail reading and organizing worked and only the scheduled send failed (community post).
- drevendev (Oct 10) runs about 15 hourly ChatGPT Work scheduled tasks on the web with GitHub and Google Drive both set to Allow all actions. The last-run time only showed that a task had started, not that its work was saved. Some GitHub writes were blocked by safety checks while reads worked, and one Google Docs insert reported success but damaged the text next to it (GitHub issue #52872).
- Abiram (Oct 7), who builds automation for companies, suggested a way to keep runs unattended: let the AI step write a structured record of what should happen (recipient, template, row, status "approved"), and let a tool you control, such as an Apps Script trigger, do the actual send or write (community post).
More asks you can copy
Replace the parts in brackets.
Every weekday at 7:00 [IANA time zone], turn today's Google Calendar into one image: time, title, and location for each event. Send it to me in ChatGPT. Don't create, move, or delete events. Confirm the schedule. (from まさやん's setup)
Every morning at 7:00 [IANA], summarize only the important mail from the last three days. For each item, give the sender, one line on why it matters, and whether I need to reply. Post it here and in [Slack channel]. Don't send or archive anything. (from 단테랩스's setup)
Every weekday at 8:00 [IANA], give me up to five news items on [topics] from [named sources]. Keep it under 200 words, link the original source for every item, and mark any claim about pricing or access that you couldn't confirm on the official page. (from A2Z AI's and The AI Frontier's briefs)
Read my connected apps overnight and give me one morning brief instead of separate notifications. After [time], hold anything non-urgent until tomorrow's brief unless it changes today's plan. My fixed commitments: [e.g. school pickup at 3:30]. (from @theaiimpact's setup)
Every weekday at 8:30 [IANA time zone], send me today's events from all my calendars ([list them]) in time order, then the top three to-dos and the emails that need a reply. Keep it to 5 lines. Don't send or post anything, and don't change events, sheets or tasks. End with one line: which calendars you read and the time you sent this. Confirm the saved schedule. (from uehara's and わっしー's setups)
Start every morning brief with one line: the time this run started, the time window it covers, and every source you read. If you're sending this outside a scheduled run, say so. (from the Future Stack Reviews and Takumi Shibata reports)
List every brief you were scheduled to send me in the last three days, with the time each one was due. For each, tell me whether you can find it in our conversation and when it arrived. Don't run any of them again, and don't change the schedule. (from the Future Stack Reviews and Takumi Shibata reports)
List every scheduled task you can see: its name, time and time zone, where it delivers, whether it's on or paused, and when it last ran. Tell me if any two look like duplicates. If you know a task exists but can't open or change it, say so. Don't change anything. (from the stuck points above)
Change my morning brief to 7:30 [IANA time zone] on weekdays only, and keep the instructions and destination as they are. Then read the saved task back to me. If the change is refused, tell me the exact message and stop. Don't create a new task in its place. (from the stuck points above)
At the end of every brief, add one line on what you couldn't do this time: any source you couldn't read, and any send or write that was blocked, with the exact error. If everything worked, write "all sent". (from tmoiv's and gaelto's reports)
Related
- Morning brief: long-form real cases · Email triage cases · Meeting prep cases · Meeting prep workflow
- Dots + Gmail · Dots + Google Calendar · Dots + Slack · Research watch workflow
- All workflows · Getting started · Integrations hub
- How to get ChatGPT Dots · Dots in the EU & UK · Features · Google Drive · AI secretary · Dots not showing
Copy this task brief
Responsibility: prepare my weekday morning brief. Sources: [mail labels], [calendar names], [project documents]. Scope: changes since the previous completed brief; on the first run, use the last 48 hours. Schedule: [time] [IANA time zone], Monday–Friday, until [end date]. Confirm the schedule. Output: 1. Today’s meetings, with preparation needed. 2. Deadlines in the next seven days, with owner and source link. 3. Unanswered questions that need my input. 4. My next three actions, ordered by impact and urgency. Keep the brief under 500 words. Deduplicate the same event across sources. Label missing dates as unknown. Put the finished brief in [destination]. If nothing material changed, say so in one sentence. Start by producing one brief now, then incorporate my feedback.Download Markdown ↓