Neither does anyone selling you one.
You typed your own questions into ChatGPT to check. Sometimes your company came up. Sometimes a competitor did. Sometimes it invented something about you that isn't quite right. You closed the tab knowing a little more and feeling no more certain.
Here's the part nobody leads with: you don't have a visibility problem you can fix by rewriting everything. You have pages AI can read and pages it can't, pages worth being found for and pages that don't matter, and no way yet to tell which is which. That's diagnosable.
It'll arrive as a checklist, a retainer, or a tool that scores your whole site red and offers to fix it. You've heard this pitch before in a different costume. It was "clean all your data first."
Your content isn't the problem. Your content is the symptom. Most of what's "wrong" with a site in AI search isn't writing. It's structure:
You don't rewrite those. You find them. And you only fix the ones sitting between you and a page worth being found for.
Every page in scope gets pulled, read the way an AI crawler reads it, and checked against four things:
Then findings join to your real traffic, so what comes back isn't a pile of problems. It's a sequence: this page carries weight and it's broken, fix it first. This one scored badly and nobody visits it, leave it alone.
Every finding carries a confidence level.
Most scanners return a confident score for every field, because a dashboard full of green numbers demos beautifully. It also lies to you in exactly the places you can least afford it. You've sat across from four vendors who answered your data-quality question with something that sounded technical and meant nothing. You wrote "vague on data quality" in your notes every time.
An audit that says "I don't know, and here's why" is worth more than one that's confidently wrong. You knew that already. This is a system that agrees with you.
I'd rather say that directly than pretend otherwise.
Some things we can verify. Some we can't yet.
Verifiable and fixable:
Still an informed guess:
The fixable fundamentals are most of the work. The deeper mechanics are guesses right now, and the honest version of this labels them as guesses. My rubric is versioned, because it's going to change as the ground does.
If someone's selling you certainty about how ChatGPT decides what to quote, that's the tell. They're guessing too. They're just not telling you.