It's been a fun weekend. I am on vacation in India visiting family but have not been able to pull myself away from redesigning my website. I've been following a lot of SEO experts and they talk about AEO and how we need to show up in LLM prompts, as that is where people are going to get information. Here is a post from Yamini Rangan about this topic (excerpt below).
The average number of words in a traditional search query: 5. The average number of words in an LLM prompt: 23. That simple stat signals a major shift in how buyers find information, discover businesses, and make decisions.They’re now asking much more specific questions in LLMs. That means marketers need to create much more specific content.
I was feeling pretty good about having "specific content" as I have been posting about GenAI in GTM on my newsletter for almost 2 years. I have 90+ articles and over 4000 subscribers. Imagine my surprise when ChatGPT said that it does not have LinkedIn newsletters in its training data.
I don’t have access to LinkedIn newsletters, LinkedIn data, or any other private/paid-access content unless it was publicly available on the open web before training and allowed for use.
Here is what ChatGPT had to say about my content being on LinkedIn.
- Terrible external discoverability:
- Content is effectively siloed inside LinkedIn
- If someone Googles “GenAI GTM playbook” or “GenAI for sales,” your LinkedIn newsletter rarely shows up
Translation: Two years of content. Completely invisible to the AI systems
So, my next project was obvious. I needed to find a way to get all my 90+ newsletters onto my website. This is not as obvious as it seems as the process involved is -
LinkedIn does not provide an API to get a list of articles that I have published. So a real time automation was not possible. Perplexity's comet browser ran into challenges with doing this for 90+ articles. Luckily I was able to get an archive from LinkedIn. The archive was missing the newsletter images, and so I need to find a different way to get the images.
I needn't have worried about the images as those are for humans and add no value to the LLM, but we are still in a world where humans will read the articles, so I decided to figure out a way to automate this step in the process. Claude recommended running javascript on Chrome. It generated the javascript, told me how to paste it into the console, helped me debug errors and finally I had all the images downloaded.
There were 2 more steps in the process. The first was to organize the 93 newsletters into topics. I took all the HTML and passed it to Claude and had it categorize the articles, identiy topics and generate a mapping file or categories and articles.
Finally, I had everything I needed to add my newsletters to my website. I passed the articles, images and mapping file to Claude Code and asked it to add it to the blog section of my website and now I have my newsletter articles on my website.
Sometimes you don't need to automate a process. A batch download can meet the same need. Sometimes you have to automate parts of the process and connect them together manually. You don't have to automate it end to end.
The irony isn't lost on me. I spent two years writing about how AI is changing GTM, while my own content strategy made me invisible to AI-powered discovery. Now when someone asks an LLM about GenAI for sales or GTM automation, my work has a chance of actually being in the conversation.
The process taught me something else: I kept waiting for the "right way" to automate this - some clean API-driven solution. But LinkedIn doesn't have an API for this. The right answer was batch downloads, targeted scripts, and manual assembly. Not elegant. But done.
If you're building expertise in public, make sure it's actually public. And if you're waiting for the perfect automation before solving a problem - sometimes good enough automation is what gets you to finished.