How to write location pages with ChatGPT

Why do ChatGPT location pages often fail?

Because the usual prompt is "Write a landing page for [service] in [city]", run once per city. The model fills the gaps with general claims, and every page ends up saying the same thing in slightly different words. Google's spam policies call that pattern doorway abuse when the pages target "specific regions or cities that funnel users to one page", and scaled content abuse when many pages are generated "without adding value for users".

The second failure is invented detail. Asked to make a page "local", a model will name neighbourhoods, landmarks or travel times it has guessed, and may present them confidently. A page that says your clinic is near a metro station that does not exist costs more trust than a plain page.

The tool is not the problem. Google's guidance on generative AI content says AI can be useful for research and adding structure, and asks site owners to make sure the result is accurate and adds value, including the titles, descriptions and structured data the AI produces.

What is the process for writing location pages with ChatGPT?

1. List only the places the business serves, and decide which have a branch and which are service areas without one. Our service area page template explains the difference.
2. Build one fact sheet per place in a spreadsheet: localities and pin codes covered, branch address and hours (if any), visit or travel times, local prices, staff or doctors there, local questions customers ask, and any real reviews from that area.
3. Fix the page outline: the same sections in the same order for every page, so the model never decides the structure.
4. Write one system prompt with the rules (below) and reuse it for every page.
5. Generate one page per conversation or API call, pasting only that place's fact sheet, so details from one city cannot leak into the next.
6. Check each draft against the fact sheet and with the swap test before it goes anywhere near the site.
7. Write the title, meta description and schema yourself, or check the model's versions line by line.
8. Publish in batches, link each page from a locations hub and its service page, and watch Search Console before adding more.

What prompt should you use for a location page?

Put the rules in the system prompt (or the first message) and keep them the same for every page. A version that works:

You write location pages for [business], a [type of business]. Write only from the facts provided for this place. Do not add place names, landmarks, distances, prices, statistics or reviews that are not in the facts. If a section has no facts, write "NO FACTS" for that section instead of filling it. Use this outline: H1 with the service and place; a two-sentence answer saying how we serve this place; areas covered; how the service works here; prices and timings; local questions and answers; how to book. Plain, specific sentences. No claims about being the best, no guarantees, no medical promises.

Then send the fact sheet for one place: "Facts for Kothrud, Pune: [paste the row]". The "NO FACTS" instruction matters most. It turns a gap in your data into something you can see and fix, instead of a paragraph of filler you might miss.

Where do the local facts come from?

Mostly from the business, not the internet. The most useful facts are ones only the client knows: which localities actually send customers, what a visit costs there, which doctor works at which branch, what people ask on the phone. Ask for them in a short form per branch or area.

For public facts, such as localities in a city, nearby areas and the local economy, use sources you can check and keep the link next to the fact in the sheet. If you let ChatGPT research with browsing turned on, treat what it returns as leads to verify, not as facts.

How do you check AI-written location pages before publishing?

1. Compare against the fact sheet. Every place name, number and claim on the page should appear in the sheet. Delete anything that does not.
2. Run the swap test. Replace the place name with another city. Any sentence that is still true is generic; cut it or replace it with a fact.
3. Compare two sibling pages side by side. If paragraphs line up almost word for word, the pages are near-copies.
4. Read the claims as a patient or customer would. Remove superlatives, guarantees and, for healthcare, anything that promises an outcome.
5. Check titles, meta descriptions and schema for the right service, place and details. Google's guidance names these as AI output to review too.

If many published pages already sit unindexed, our guide to location pages that are not indexed covers the fixes.

How many location pages can you write this way?

By hand in a chat window, a few dozen before the copying and pasting becomes the job. Through the API with a script, many more, but the fact sheets and the checks still scale with every page. That is the point where agencies usually stop doing it by hand.

Scaleo SEO runs this process as a pipeline. It researches each city's localities, landmarks and local economy, writes every service and city page on its own from the client's business facts, has a second AI model score each page with checks that include copy that would read the same in another city and banned filler words, rewrites pages that fail, and publishes to WordPress, Elementor, headless ACF or static HTML. See how Scaleo SEO builds location pages.