I built a tool to rank higher in AI answers

TL;DR: Built an SEO tool that actually correlates with what ranks #1 and AI answers. Started because I was tired of optimizing for metrics that didn't matter. Tested it on real client campaigns for 6 months.

The problem that wouldn't leave me alone

I run an SEO consulting business, and something was off.

Every tool gave me scores that didn't match actual search results. I'd optimize a page to 95/100, and it would rank on page 3. Meanwhile, a competitor's "72/100" page sat at #1.

Worse, none of these tools tracked what Google actually tells us matters: relevance, quality, and search intent. They were measuring keyword density from 2010, word count, readability scores that may or may not matter.

I kept thinking: what if I could build something that actually predicts what ranks?

The false start

Like every developer who's watched too many YouTube videos, I thought: "I'll build this properly. Next.js, React, all the cool stuff."

Authentication took me a week. One entire week of configuration hell.

Then I had a moment of clarity: why am I doing this to myself? I went back to what I actually know — LAMP/PHP. Two hours later, auth was working.

Sometimes boring is beautiful

Here's what nobody tells you: the best tech stack is the one you can actually ship with.

PHP isn't sexy. But the server hasn't crashed once. No build errors. No deployment nightmares. No days being stuck on dependency hell. It just works.

Making it actually work

I first started experimenting with search engine embedding models to see if I could measure topical relevance the way search engines actually do it. And surprisingly it worked.

Then I pulled data from 96 different #1 ranking pages across various niches and built a custom scoring model around those patterns.

The correlation was there. When my tool said a page was well-optimized, it tended to rank.

Six months of reality testing

I didn't launch immediately. I used it on my own consulting clients for six months. Real campaigns, real money on the line.

The tool guided my optimization recommendations. Multiple campaigns hit the top spot. Not because of magic, but because it pointed toward what actually mattered—relevance, quality, intent alignment.
After beta testing with other SEOs who found bugs I missed, it felt ready.

What I learned

Ship with what you know. That week fighting with auth on React taught me the "right" technology is the one that lets you build the thing.
Measure what matters. If your metrics don't correlate with results, they're vanity metrics.

Test in production. Using this on real client work forced me to make it actually good. Client work keeps you honest.

Where it's at now

The tool is live. It analyzes content for topical relevance and gives optimization recommendations based on what's actually ranking at #1.

Is it perfect? No. Will it magically get you to rank #1? Also no. But it solves the problem that drove me insane: knowing whether your optimization efforts are pointing in the right direction.

You can check it out here:
https://www.keywordgrid.com

If you've felt like SEO tools were grading you on the wrong things, I'd love to hear if this resonates. And if you try it, let me know what breaks—I'm still in "ship and iterate" mode.