Most people treat AI like a vending machine — ask a question, get an answer, close the tab. ChatGPT Dots flips that model. Instead of you going to ChatGPT, ChatGPT goes out and works on your behalf, on a schedule, while you do something else entirely.
What ChatGPT Dots Actually Does
Dots is an interface layer inside ChatGPT that lets you trigger its computer-use and browsing capabilities, then schedule those actions to repeat at whatever interval you set. Think of it less like a chatbot and more like a background agent you’ve briefed once and then left running.
The key mechanic: you describe a task, ChatGPT executes it immediately using its browser, saves the task as a scheduled job, and then re-runs it automatically. You get notified when something worth your attention shows up.
That’s a genuine shift in how useful an AI assistant can be day-to-day.
A Practical Setup: Job Listing Tracker
Let’s say you’re passively job hunting in a specific niche — UX writing roles at Series B startups in Austin. Manually refreshing LinkedIn or Wellfound twice a day is tedious and easy to forget.
Instead, you’d open Dots and prompt it something like:
“Scan job boards including LinkedIn, Wellfound, and Indeed for UX writer roles at Series B or Series C startups based in Austin, Texas. Look for postings with salary ranges above $90,000. Run this search now, then repeat every 45 minutes. Notify me whenever a new listing appears that wasn’t in the previous scan.”
ChatGPT uses its computer to browse those sites, compiles what it finds, and saves the task. From that point on, it’s checking on your behalf — you’re not involved until something new surfaces.
You can refine it mid-stream too. After reviewing the first batch of results, if you notice most of the roles are at agencies rather than product companies, just tell it: “Filter out agency roles going forward.” The scanner updates.
Other High-Value Use Cases
Once you understand the pattern — describe a browsing task, set a cadence, get alerts — a lot of previously annoying monitoring problems become trivially easy.
Price tracking. Point it at a specific product page on a retailer’s site and ask it to flag you if the price drops below a threshold. More flexible than most built-in price alert tools because you can describe the logic in plain language.
Competitor content monitoring. Ask it to check a competitor’s blog or press page every few hours and notify you when new content appears. Useful for staying current without subscribing to a dozen newsletters.
Local event discovery. If you’re looking for a specific type of community event — say, small jazz performances or indie film screenings — you can have it scan local event sites on a weekly basis rather than remembering to check yourself.
Research monitoring. Academic or industry professionals can have it watch preprint servers or publication pages for new papers matching a topic, author, or keyword.
How to Write a Good Dots Prompt
The quality of your scanner depends entirely on how clearly you brief it upfront. A few things that make a real difference:
- Be specific about sources. Name the websites rather than saying “look online.” Generic instructions produce generic results.
- Define what “new” means. If you want only listings posted in the last 24 hours, say so explicitly. Otherwise it might resurface older results.
- Set a sensible interval. Every 30–45 minutes makes sense for fast-moving things like rentals or flash sales. Every few hours is fine for job boards or news. Don’t over-poll — it wastes compute and floods you with noise.
- Include a filter condition. “Notify me only if X” is more useful than “notify me about everything.” Less noise, more signal.
Checking and Adjusting Scheduled Tasks
Once a scanner is running, Dots keeps it listed in your scheduled tasks panel. You can open any task and edit the interval, adjust the prompt, or pause it entirely. This is where the setup pays off: you’re managing a small fleet of persistent agents rather than asking one-off questions.
If your needs change — you found an apartment, you accepted a job offer — you can kill the task in seconds. Nothing lingers.
The Bigger Shift
The real value here isn’t any single scanner you build. It’s the habit of thinking about what you’re currently monitoring manually and asking whether ChatGPT could just… do that for you. A lot of the low-grade mental overhead in daily life — checking, refreshing, remembering to look — is exactly what a scheduled AI agent is good at absorbing.
Start with one thing you check repeatedly out of habit. Turn it into a Dots task. See how much mental space that frees up. Then go from there.