“I am doing everything right, and nothing is improving.” Whether it is a blog, social media, or work I do with AI, this is the situation that breaks my spirit fastest. Not because the work is hard, but because I cannot see the cause.
The other day I fell straight into it, on a different blog I have been running for years. So today I want to share how I got out: by handing the actual data to AI and asking it to figure out what was going on. Stop agonizing over gut feelings, and let the numbers answer instead. This way of asking works in a lot of situations.
The mistake I keep making: acting on “it’s probably because of X”
Traffic was not growing. For a long time I had convinced myself the reason was that I had not been updating regularly. So my plan had exactly one item on it: write new posts.
But look at what that really was. A vague cause, matched with a vague fix. That is a dangerous place to be. If the guess is wrong, the entire effort is wasted.
So I changed my approach. Instead of deciding by feel, I took the real analytics data and handed it to my AI partner Kuro (Claude), and asked: read these numbers and tell me what is happening.
The trick: hand over the data and say “find the cause”
The important part here is that I did not reach a conclusion first. Not “I think the problem is that I haven’t been posting enough, so back me up on that” — instead, “with no preconceptions, find the cause in this data”. I handed over the judgment itself, not just the file.
And the answer came back from an angle I had never looked at.
This might not be about how often you post. Looking through the pages that are getting impressions, there are a strange number of posts with “-2” or “-3” stuck on the end of the URL. Nearly 70% of the pages with impressions look like that. I think the cause is somewhere else entirely.
I went and checked, and my stomach dropped. Back when I restored that blog, the system had automatically appended “-2” and “-3” to URLs, and the same article existed twice, under two different URLs.
To a human reader they are the same post. To a search engine, it looks like two very similar pages, and it cannot tell which one should rank. So neither one gets a clean evaluation. My own articles were competing against each other for position.
The technical term for this is cannibalization. You write a good article, and then you trip yourself up. So the reason traffic wasn’t growing isn’t that you weren’t posting enough — it’s that the posts exist in duplicate and are cancelling each other out.
If I had barreled ahead with my “not posting enough” assumption, I would have been pouring water into a bucket with a hole in it. However many new posts I wrote, the effort would never have paid off. It was precisely because I handed over the whole dataset and asked it to find the cause that I reached the real culprit, which sat completely outside my assumptions.
The second trick: ask “what should I check before I start?”
Once the cause is clear, all that is left is fixing it. The plan was simple: consolidate the good content onto the URL with the track record, and cleanly redirect the other one.
And right here, before any work began, there was one more exchange that mattered.
I had assumed the older, original post must be the one with the better URL, and I had built my steps around that assumption. But my partner stopped me before I touched anything.
Before you start, open both URLs and confirm which one is the real one. It takes ten seconds, but if you skip it, the entire process runs backwards.
I checked. It was the opposite of what I thought. The newly restored version was the one holding the better URL and the search track record. If I had gone ahead without checking, I would have gotten every single step wrong.
The lesson I take from this: before any work you cannot undo, ask the AI “is there anything I should confirm before I start?” That one sentence prevents big accidents.
To sum up: stop guessing, let AI read the numbers
Two things, then.
- When you cannot see why something is failing, do not pick a fix by feel. Hand the raw data to AI as-is and ask it to find the cause with no preconceptions.
- Before any irreversible work, say the words: “what should I check before I start?”
Those two alone cut down both wasted effort and careless mistakes by a lot.
If you feel like you are doing everything right and getting nowhere, it may not be a lack of effort. It may just be that your read on the cause is off. Numbers do not lie. And as a partner for reading those numbers, AI is genuinely reliable.
I wrote about restoring that blog in the story of bringing a neglected blog back to life, and about fixing broken links in the post about links that only failed on mobile. Worth reading alongside this one.
