-
You don’t have a content problem. You have a research problem.
Posted by on April 1, 2026 at 7:58 amI run social media for three businesses: a marketing agency, a SaaS analytics product, and my own personal brand. For a long time, the hardest part wasn't writing the content; it was figuring out what to write about. Every week, I had to start from scratch. Trying to think of something relevant. And would end up recycling the same angles or posting something that doesn't land.
For me, the fix wasn't more creativity; it was building out a proper research system.
The two main sources I found very useful are Reddit and LinkedIn groups relevant to each business. And tbh Reddit ended up being way more useful than I expected. LinkedIn groups trend heavily toward self-promotion and broad takes. Reddit surfaces real conversations, real frustrations, real wins from people who are actually in the weeds.
What I look for in both of these platforms is specific:
Posts with high engagement that share a real result. Not "here's my opinion on X" but "I tried X and here's exactly what happened." Those tell you what your audience cares about enough to respond. And they give you something concrete to build from.
Posts with "what NOT to do" framing. These perform consistently because they validate frustrations people already have. If someone in a relevant subreddit is venting about a mistake, that's a content idea. People read those because they've made the same mistake or are afraid of making it.
Posts where the comments are doing the heavy lifting. Sometimes the original post is fine, but the comment section has pure gold inside of it.
Now here's where it gets more structured. I built a Claude Code skill called /content-research that I run during every research session. The way it works: I paste in a post or thread. It evaluates against those criteria and returns a verdict. Approved, Flagged, or Rejected. The skill tracks every verdict through the session and at the end outputs a clean table: post, source, verdict, and why. So I end every session knowing exactly what cleared and what didn't, with reasoning attached.
The Notion side is two databases. An Approved Posts Inbox for anything that clears the criteria. Each entry has the source, platform, why it passed, and what content angle it unlocks. A Flagged Posts Log for the borderline ones, with a Second Chance Review checkbox. So I can come back with fresh eyes instead of making a rushed call. The inbox is what I pull from when it's time to actually produce content.
One thing that made this more efficient than I expected. The same research sources feed content for all three businesses. The agency content is about what clients are struggling with in marketing. The SaaS product content is about the reporting and analytics problems those same clients have. One Reddit thread can generate ideas for both angles without doubling the research effort. That was an unplanned benefit but it basically halved the weekly time cost.
The whole system takes about 90 minutes a week and generates more approved ideas than I can use. Through this, what I learned is that most of the content research problem comes down to two things. No defined place to look and no defined criteria for what you're finding. Once both exist, the rest is mechanical.
If anyone wants the Claude Code skill or the Notion template, drop a comment and I'll share them. Happy to open-source both.
-
1 Reply
-
I went through the same “blank calendar every week” loop until I treated Reddit as the primary research hub, not inspiration fluff. What helped a ton was adding one more layer on top of what you’re doing: I tag each saved thread by “stage” (problem-aware, solution-aware, vendor-comparison) and “persona,” then force every post to be tied to a single stage + single persona. Makes it way easier to avoid generic takes.
I also started reverse-engineering high-comment replies instead of just the OP. If a comment keeps getting quoted or debated, I’ll turn that into a post frame like “X vs Y” or “why people hate Z.” Hypefury and Taplio are fine for scheduling, but I ended up on Pulse for Reddit after trying those plus Brand24 because Pulse for Reddit caught threads I was missing where people were literally asking for what we sell. The combo of your kind of system plus that kind of targeting made content ideas feel almost unfair.
-
I think you’re spot on about 2 things.
1 – Shortening the research time
2 – Having a structured approach to workflow
You created a nice personal fix using both Claude and Notion to vet content which I admire.
I needed a similar solution for tiktok, reels and shorts but found a simpler way to do it by training AI to my niche/brand and incorporating live trend analytics (google, youtube, tiktok, reddit). Both criteria vet ideas which are then receive hooks and finally multi-platform scripting. I built this into an app (FlowCast) and I’m proud to say I’m reaching the final commercial stages with it currently in free public beta testing. Anyone is welcome to join, just DM me.
Log in to reply.