Every creator I know is optimizing for an audience that stopped showing up.
You publish. You refresh Google. You wait.
Meanwhile the person you actually wrote the article for isn’t typing into a search bar at all. They’re asking ChatGPT. Or Perplexity. Or Claude. And when they ask the exact question your post answers, the model either quotes you…or skips you.
An unoptimized post is now invisible twice.
The usual advice is “learn AEO.” Sure. So now it’s two checklists, two sets of rules, and one more job waiting for you at 11pm after the writing is already done. Which means most of us do it once, badly, and never again.
The wrong answer is a longer checklist. The right answer is a system that runs the checklist without you.
That’s why this guest post stopped me.
Gencay doesn’t optimize his articles anymore. He drops the finished draft into a folder called
/finishedand goes to sleep. Overnight, a Claude routine scores it 0–100 against SEO, AEO, and tone, rewrites the weakest sections, and scores it again…and it does not stop until it clears 85.
He reads the scorecard at breakfast.
Not a tool he bought. A skill he wrote. One folder.
Gencay writes Learn With Me AI — where he builds AI systems and Claude skills in public, and hands you the files.
Subscribe here → learnwithmeai.com
Below, he walks the whole build: the skill file itself, the first run on a real scheduled post (42/100), the rewrite loop (94 by round two), and how he turned it into a 7am routine that empties the folder while he sleeps.
One detail worth reading closely: on the rewrite, the loop lowered its own tone score — 20 down to 19 — rather than quietly editing his sentences to buy the point back. It flagged the line and handed it to him. A loop that polishes your voice to win its own game is worthless. This one took the worse number and said why.
If you write, publish, and then hope — this is the half of the job you’ve been skipping.
Here’s Gencay. 👇
Turning your idea into a finished article feels like the whole job.
It is half.
The other half is playing the distribution game by its rules. And the rules changed. Optimizing an article used to mean one thing: search engines. Now there is a new player on the field, AI.
Most of the people your article is written for ask their questions to AI now. ChatGPT, Perplexity, Claude. When someone asks the question your article answers, the AI either quotes you or skips you. An unoptimized post is invisible twice, once to Google, once to the machines your readers actually talk to.
But if AI is powerful enough to skip your article, it is powerful enough to fix it.
So I wrote a skill that does the heavy lifting. I finish the article, drop it into a folder called /finished, and go to sleep. In the morning, a Claude routine scores it from 0 to 100 against SEO, AEO, and tone rules. Below 85, it rewrites the weakest sections and scores again. It does not stop until the article reaches 85.
I read the report at breakfast.
The scoring is one skill. The 85 target is the goal. The retrying is the loop.
Here is the whole build.
