I use AI every day. It researches keywords for me, drafts outlines, checks pages
for missing schema, and writes first drafts of blog posts. My own site runs on a
content engine I built with it.
So this isn’t an argument against AI. It’s an argument against handing your
website to AI with nobody in the loop who knows your business, knows search, and
knows when the tool is wrong. Those are different things, and the difference
shows up in your rankings, in what AI assistants say about you, and in whether
the person reading your page picks up the phone.
A language model knows what most cleaning companies, gyms and brokerages say
about themselves. It doesn’t know what you charge for a move-out clean in Round
Rock, which of your classes fills up at 6am, or what the HOA in a specific condo
building actually covers.
That first-hand detail is exactly what makes a page worth ranking. Without it,
you get a page that is accurate in general and useless in particular.
Ask a model to write a page about a topic and it produces something close to the
middle of everything already published on it. That’s fine for a first draft. As
a finished page, it gives Google and AI assistants no reason to choose you over
the twenty pages it was averaged from.
I see this in my own testing. In October 2026 I asked ChatGPT what an Austin
real estate agent’s bio and about page should say. It answered almost entirely
from Google’s and OpenAI’s own documentation, and the bio it offered was a
fill-in-the-blank template with no Austin in it. The only real agent sites it
cited were cited for their neighborhood content, the kind of local knowledge a
model can’t produce on its own.
In the same round of testing, I asked whether Google Local Services Ads are
worth it for a small Austin house cleaning company. The answer was built on an
example campaign the model made up, which it labeled as not reflecting real
Austin rates. That label was honest. Most AI-written pages don’t carry one, and
nobody reading them can tell which numbers are real.
On your own website, an invented price, statistic or claim is your invented
claim. Customers hold you to it.
If you and three competitors all ask the same tool for “a services page for a
house cleaning company in Austin,” you’ll get four versions of the same page.
Interchangeable pages don’t build trust with people, and they don’t give search
engines a reason to rank one over another.
Google’s public guidance is that it cares whether content is helpful, not how it
was produced. Its spam policies do target pages produced at scale mainly to
manipulate rankings, which it calls scaled content abuse, whether a person or a
program made them.
In practice, that means AI isn’t the risk. Publishing what AI produces without
adding anything is. A page needs something only your business could have
written: real prices, real service areas, real examples, real answers to the
questions your customers ask you on the phone.
| Step | AI on its own | A practitioner using AI |
|---|---|---|
| Choosing topics | Writes about whatever you type in | Uses search and AI-assistant data to pick topics your customers actually look for, and skips the ones you can’t win |
| Research | Summarizes what’s already out there | Checks each claim against its original source, and gets the details from you |
| Drafting | Produces a finished-looking page in seconds | Uses the draft as a starting point and rewrites it around your business |
| Facts and numbers | States them confidently, sourced or not | Cuts any number that can’t be traced to your data or a named source |
| Local detail | Generic city names | Your actual suburbs, buildings, neighborhoods and customers |
| Compliance | Doesn’t know your industry’s rules | Applies them: fair housing language for real estate, no medical claims for wellness, honest pricing |
| Publishing | Whatever it wrote goes live | Nothing goes live until someone who knows the business has read it |
The AI makes the work faster and cheaper. The person makes it right.
My content engine does most of the mechanical work. Once a week it researches
topics for cleaning companies, real estate agents and gyms in Austin, scores
them on search demand and how often people ask AI assistants about them, and
queues them up. It writes drafts with a brief for each one that lists which
claims have to be checked against their original source before anything is
written.
Then it stops. Every draft lands in WordPress as unpublished, and nothing goes
live until I’ve read it. On one of the first posts it wrote, I rewrote the
title and URL before it went live. It also checks every week whether AI assistants are
citing my site when people ask about these topics, so I can see what’s working
instead of guessing.
That’s the setup I offer clients: the engine handles research, drafting
and technical checks on a schedule, and I make sure everything that goes out
under your name is accurate and sounds like you.
I’ll review your site and send you three to five specific findings: what’s
holding your pages back and what I’d fix first. Some of them you’ll be able to
fix yourself.
Most advice on cleaning contracts is about chasing bids. This covers the other half: the…
Most gym marketing advice comes from software vendors. This is the short version for an…
Most advice on cleaning contracts is about chasing bids. This covers the other half: the…
Most real estate marketing advice is a long list of ideas in no particular order.…
Once fundamental SEO is in place, the process of ongoing content updates will help you…