Why raw ChatGPT service pages don’t rank (and what AI-assisted content actually needs).

You typed "write me a service page for a plumber in Fort Worth" into ChatGPT, pasted the output onto your site, and three months later it's ranking nowhere. Maybe you did this for ten pages. Maybe fifty. The traffic didn't come, and now you're wondering whether Google is punishing you for using AI.

Here's the short answer to "does AI content rank on Google": yes, it can — and no, yours probably won't. Not because Google detected a robot. Because the page you published contains nothing that a thousand identical pages don't.

Google's actual position (it's not what you think)

Google has been unusually direct about this. In its February 2023 guidance on AI-generated content, Google Search Central stated that its focus is "rewarding high-quality content, however it is produced." Using automation, including AI, is not against its guidelines. Using automation to generate content primarily for ranking in search is — and always has been, since long before ChatGPT existed.

Then in March 2024, Google rolled out a core update plus new spam policies explicitly targeting "scaled content abuse" — publishing many pages that add little value regardless of how they were made. Google said the combined changes were expected to reduce low-quality, unoriginal content in results by about 40% (it later reported the actual reduction came in around 45%).

Read those two policies together and the picture is clear. Google isn't running an AI detector and flagging your pages. It's asking one question: does this page contain anything that doesn't already exist a thousand times?

Why a raw generated page fails that test by construction

A language model writes the statistical average of everything it has read. Ask it for a "plumber in Fort Worth" page and you get: "fast, reliable service," "licensed and insured technicians," "emergency repairs available 24/7," a bulleted list of services, and a paragraph about customer satisfaction.

That output isn't wrong. It's average — mathematically. It's the same page the model writes for a plumber in Fort Worth, Fresno, or Fargo, with the city name swapped. And every one of your competitors with a ChatGPT login can produce the identical page in eleven seconds.

Google's systems are built to surface content that demonstrates first-hand experience and original information — that's the "E" for Experience that Google added to its E-E-A-T quality framework in December 2022. A raw generated page has zero experience in it. There's nothing to demonstrate, because nothing in the page happened. No specific job, no specific price, no specific street, no specific customer question. It is unhelpful by construction, and the March 2024 policies exist precisely to keep it out of results.

There's a second-order problem too: when fifty local businesses publish near-identical generated pages, Google has no signal to rank one above another — so it ranks the ones with the strongest everything-else (reviews, links, real content elsewhere on the site). The generated page didn't just fail to help. It cost you the slot a real page could have earned.

What the model is missing: the stuff only you have

The fix is not "write it by hand." The fix is realizing that AI is a drafting tool, and a drafting tool is only as good as the raw material you hand it. What you have — and what no model, and no competitor, has — is:

• Your actual jobs. The slab leak on a 1962 pier-and-beam in Arlington Heights. The tankless install in a Keller new build. What you found, what it took, what surprised you.
• Your actual prices. Even ranges. "Water heater replacement in our area typically runs $1,400–$2,600 installed, depending on tank vs. tankless" beats "affordable pricing" every single time — for the reader and for the query.
• Your actual geography. Which neighborhoods you serve, which you don't, how the housing stock differs. Post-war homes off Camp Bowie have different plumbing problems than 2019 builds in Walsh.
• Your actual customer questions. The five things people ask on the phone before booking. Those are search queries, verbatim, and nobody's answering them with real answers.

None of that requires writing skill. It requires ten minutes of talking into your phone or answering a questionnaire.

The workflow that works: specifics in, draft out, expert pass

Here's a process that produces AI-assisted pages that actually earn their rankings:

1. Dump the raw material first. Before any prompt, collect: 3–5 real jobs (anonymized — no client names without permission), your price ranges, your service area with actual neighborhood names, your genuine differentiators (response time, warranty terms, who answers the phone), and the questions customers really ask.
2. Prompt with the material, not the topic. "Write a drain cleaning page using these four jobs, these prices, and these three customer questions" produces a categorically different page than "write a drain cleaning page."
3. Do the expert pass. Read the draft as the practitioner. Cut anything you wouldn't say out loud. Fix anything that's technically off — models get trade details wrong confidently. Add the one detail the model couldn't know.
4. Check the uniqueness test. Could a competitor publish this page unchanged? If yes, it's not done. Every ranking-worthy page fails to transfer.
5. One page per real intent. A "plumber in Fort Worth" page and a "water heater replacement Fort Worth" page target different searches. Ten thin generated pages targeting the same phrase with different synonyms is exactly the scaled-content pattern the March 2024 policy names.

Illustratively: a typical Keller plumber who publishes six pages built this way — each with real jobs, real prices, real neighborhoods — will beat a competitor with sixty generated pages, because Google has forty-five-percent fewer slots for the generated kind and a stated preference for the first kind.

How West Fork approaches this

We use AI in our content work and say so. What we don't do is publish a draft the model wrote alone. Every service page starts with a structured intake — your jobs, pricing reality, service area, and the questions your customers actually ask — and ends with an expert edit and a uniqueness check: if a competitor could run the page unchanged, it goes back for another pass. The tool accelerates the writing. The business supplies the substance.