We paid writers and agencies to produce SEO content. Months went by and almost nothing got published. When something did go out, the topic had been picked by someone's gut.
The strange part is that the answer was already sitting in our own account. Search Console told us every query we showed up for and every one we did not. Nobody was reading it.
- Clicks a month
- 150→813
- Impressions a month
- 29k→202k
Google Search Console, Nov 2025 to Aug 2026.
The bottleneck was never the writing
Writing a post is a day. Getting a post published took weeks, because it waited on a topic, then a writer, then a review, then a cover image, which was its own separate wait.
So I did not set out to build "an AI that writes blogs". I set out to remove the waiting between the steps.
What the tool does
I built Rankwright. It runs the whole loop, with a person in the middle.
- Pulls 90 days of Search Console data and clusters the queries by intent.
- Drops anything an existing page already ranks well for, so we never compete with ourselves.
- Blocks topics we have covered before under different wording.
- Proposes around twenty titles. I tick the ones worth writing.
- Drafts the post with internal links, an FAQ block and FAQPage schema.
- Makes the cover art in the house style and redoes it from notes.
- Publishes to WordPress and tracks every keyword daily against its baseline.
It suggests, I pick. It writes, I approve. Nothing goes out unattended.
The human stays in the loop on purpose
The tool is fast, but speed was never what made content good or bad. The two decisions that matter, which topics and whether a post is good enough, stay with a person. Everything around those two decisions is automated.
What changed
Clicks from Google went from 150 a month to 813. Impressions went from 29k to 202k. September is on pace for about 1,020 clicks, six times where it started when I took over SEO.
The bigger change is quieter. Publishing stopped depending on anyone being free. Nobody chases a writer any more.