How I got ChatGPT to write an AI automated newsletter for me.

And a LinkedIn post and a SEO post. Seriously. It took me 6 hours, but I made it work.
I found this Insta-post by theaiavalanche and was intrigued and curious – could I make it work?
I copied the prompt, modified it to my needs, and fed it into my ChatGPT and set to work.
While I know a little html, I am absolutely no developer, so if I can do this, you can too.

The prompt:

I need you to set up a weekly task. Every Monday and Thursday at 9 a.m., you will use your web search feature to find the most interesting news stories in the field of AI since the last task run was performed. Here's what you'll do with the results:
Provide me with a 100-word summary of the top three news articles that I can copy and paste into a newsletter for my subscribers. Each summary should have a source link to the original article. Create a 200-word thought-provoking post for LinkedIn on the news story that is most likely to engage users. It will be written to fit the platform, but do not abandon my writing style. Avoid AI cliches and hyperbolic phrases. Write a 500-word SEO-optimized article based on the news story. The article should be written in my style, using easy-to-read language, and offer an interesting take on the subject. Important! Turn email notifications on and alert me via email every time the task has been completed.

I also added "Have you understood the task? What do you need from me to perform this?" like I always do, to make sure ChatGPT understand what I need.

ChatGPT summed up as follows, so I could get started:
Yes – here’s how I’d tackle it every Monday and Thursday at 09:00 Europe/Copenhagen time:

1. Web search for AI‐related news since the last run.

2. Top 3 newsletter summaries (100 words each, with source links).

3. LinkedIn post (200 words, thought-provoking, in your style, avoiding clichés).

4. SEO article (500 words, easy-read, your voice, an interesting angle).

5. Email notification to you as soon as it’s done.
   
   

The stack:

Here’s a quick rundown of everything I leaned on to get the hands-free AI newsletter up and running:

• A Gmail account for sending the emails via the Gmail API

• Google Cloud Shell (Python runtime) as my “always-on” environment

• A Chromebook (with Cloud Shell in the browser) to develop and test everything

• Python 3 plus these libraries:

  • google-api-python-client, google-auth-httplib2, google-auth-oauthlib (Gmail API)

  • feedparser (to pull Google News RSS)

  • openai v0.28.0 (to generate summaries, posts, and articles)

• OAuth 2.0 Client ID in Google Cloud for secure Gmail access

• Cron to schedule the script every Monday & Thursday at 09:00 CET

• OpenAI API tokens (your sk-… key) to power the AI writing

• A small local file (last_run.txt) to remember when we last fetched news

Together, these tools transform raw headlines into polished AI-written content—and deliver it on autopilot.

1. Using Gmail to send from

I started by teaching the script to “log in” to Gmail and send mail for us.
At first, Google kept blocking our attempts — Cloud Shell’s headless setup refused the usual browser callback. After a few rounds of trial and error, I switched to an “out-of-band” login: the script prints a link, I visit it in my browser, paste back the code, and voilà—the test email landed in my inbox 🎉

2. Scheduling the automated email

Next up: making sure this all runs by itself. I added a simple cron job that says, “Hey computer, fire off this script every Monday and Thursday at 09:00 Copenhagen time.”
With that in place, even if I'm sipping coffee on a beach, I can still get my AI news fix on schedule.

3. Gathering & writing the news for the AI newsletter

I tapped into Google News’s RSS feed to pull in fresh “artificial intelligence” stories.
From there, the script uses OpenAI to whip up:

• Three crisp, 100-word summaries for the newsletter

• A thought-provoking 200-word LinkedIn post

• A 500-word SEO-friendly article
  

In one fell swoop, we turned raw headlines into polished content ready to publish.

4. Navigating API Roadblocks

Of course, no journey is without detours. I hit two big snags:

1. Library Changes: The OpenAI Python library updated and broke our calls. Rather than rewrite everything, I pinned the project to the previous version that still worked.

2. Quota Limits: My free token bucket ran dry mid-run. I topped up billing and added simple error handling so if OpenAI ever chokes again, the script still sends whatever it managed to generate.

5. End results

This is what was built:

• Every Monday & Thursday at 09:00 CET, cron runs my script.

• It fetches new AI news via Google News RSS,

• Summarizes the top three stories,

• Crafts a 200-word LinkedIn post and a 500-word SEO article via OpenAI,

• Sends everything in a neatly formatted email to myself (with further tweaking I might get it to send to a newsletter base)

• And updates its timestamp so it only pulls fresh items next time.
  

6. Conclusion

While I might not actually use AI to write my LinkedIn posts, I love the idea of getting ChatGPT to automate tasks for me, and this was a great way of testing it.
It took much longer than anticipated (approximately 6 hours) and someone with more coding skills than I could probably do it in an hour 😉
I had to top up my OpenAI account with some $ to replenish tokens, but I did fine with a small amount.
I also had to ask step-by-step simpler instructions some of the way - I got a bit lost in the tech of it, but this is where ChatGPT is great - just ask for simpler instructions.

This post is written by ChatGPT and me.