Your old social posts contain clues about what you should publish next. Here’s how to get AI to find them, and build an AI coach.
Ask ChatGPT, “What should I post on LinkedIn?” and you’ll probably get perfectly reasonable ideas.
There’s just one problem: AI doesn’t know what works for you.
A smarter approach, inspired by a system Buffer’s Hailley Griffis built, is to turn AI into a personal social-media coach by giving it something much more valuable than another prompt: Your actual performance data.
Step 1: Give AI Your Track Record
Export the last three to six months of data from your primary social platform. Include whatever you have available: the post itself, date, topic, format, impressions or reach, likes, comments, shares, saves, clicks and followers gained.
Then ask AI to analyze the entire dataset for patterns.
Don’t stop at “Which posts performed best?” That’s obvious.
Ask better questions:
- Which topics consistently beat my average?
- Which posts generated followers rather than merely likes?
- What do my five biggest failures have in common?
- Which opening hooks appear most often among my winners?
- Which successful subjects haven’t I revisited lately?
- Which posts received modest reach but unusually high clicks or follows?
That last question can uncover hidden winners—posts the algorithm didn’t distribute widely but that worked extremely well on the people who actually saw them.
Step 2: Make AI Find Your Content Gaps
Now ask: “Based on what my audience responds to, what topics am I surprisingly under-publishing?”
You might discover that your pricing posts consistently outperform everything else, yet pricing represents only 5% of your content.
That’s not merely an interesting statistic.
That’s your next content series.
The Killer Trick: Make It Predict Before You Publish
Before posting something new, give your AI coach the draft and ask it to compare the post with your historical winners.
Ask: “Based only on my previous performance, what characteristics of this post increase or decrease its chances of succeeding?”
Then publish it anyway and record what actually happens.
Over time, you create a feedback loop:
Post → Measure → Analyze → Predict → Post Again.
AI isn’t magically predicting the algorithm. It’s helping you notice patterns in your own audience that are easy to miss.
And that’s the advantage.
Millions of marketers can ask AI what they should post tomorrow.
Only you can give it the data showing what your audience loved yesterday.
So Why Might I Want to Build an AI Coach
-
Decode My Hidden “Zero-Fail” Patterns: Instead of just telling me what went viral, I’d want it to isolate the exact micro-elements shared by my top 5% of posts—like the specific sentence structure in my hooks or the psychological trigger in my calls-to-action—so I can intentionally replicate them.
-
Catch My Blind Spots and Creative Ruts: I’d have it regularly audit my content mix to flag topics I’m quietly neglecting. If my audience loves a certain problem-solving angle but I haven’t touched it in two months, I want the coach to call me out and hand me a content gap analysis.
-
Run Pre-Flight Stress Tests: Before hitting publish, I’d drop my drafts into the coach and ask: “How does this stack up against my historical flops, and what’s going to make people scroll past this?” Getting a predictive critique before the post goes live would save a lot of wasted effort.
-
Translate Raw Metrics into Revenue Drivers: Vanity metrics like likes are nice, but I’d want the coach to track which specific content themes actually correlate with deeper engagement, clicks, and community growth—cutting through the noise of the algorithm.