If you plan to use Claude to write location pages, the first thing to accept is that speed is not your real problem. Any model can produce fifty city pages in an afternoon. Most location pages fail because they are the same page wearing a different city name, and Google has been filtering that trick since long before AI made it cheap to pull.
The first batch we ran for a home services client proved this fast. The drafts were clean, quick, and nearly identical. Search Console filed most of them under “Crawled, currently not indexed” and left them there.
The method is less about prompting and more about inputs. Get the inputs right and Claude handles the writing without much fuss.
To use Claude to write location pages that rank, build one master prompt with your service details and brand voice, then feed it a separate data sheet for every location: landmarks, service area quirks, real customer reviews, local pricing notes. Claude drafts each page from its own sheet, so no two pages share the same substance.
Why Most Location Pages Fail Before Claude Gets Involved
Google treats thin, templated city pages as doorway pages. The city name changes, nothing else does, and the page exists only to catch a query.
Claude cannot fix that on its own. Give it “same service, new city, go” and it will produce fluent doorway pages. Fluent is worse, honestly, because they look finished and nobody checks them.
The real fix happens before you open Claude at all. You need something true and specific about each location, and that is research work, not writing work.
Build The Data Sheet Before You Build The Prompt
For every location, we fill a short sheet first. Fifteen minutes per city, sometimes less once you find a rhythm.
What goes on it:
- Neighborhoods and landmarks you actually serve
- One or two real customer reviews from that specific area
- Pricing or travel fee differences
- Anything odd about the place, like hard water, old wiring, or strict permit rules
- Drive time from the nearest office or technician base
Old housing stock in one suburb means different plumbing problems than new builds ten minutes away. That single detail gives a page a reason to exist.
If you cannot fill the sheet for a city, that tells you something too. Maybe you do not need a page for it yet.
How Do You Prompt Claude To Write Location Pages?
Use one master prompt and never rewrite it per city. The master prompt holds your service description, tone rules, page structure, and target length. The data sheet is the only thing that changes between pages.
A Claude Project works well for this. Load the master prompt and two or three finished example pages into the project knowledge, then paste one data sheet per chat.
In our own setup, the prompt tells Claude to lead with the local detail, not the service pitch. Readers in that city already know what a plumber does. They want to know you work on their street.
We also make Claude write the H1 and title tag last, from the finished page, instead of first. Titles come out far less templated that way.
One warning from experience. Ask for one page per chat. Batch five cities into a single conversation and Claude starts reusing phrasing across them, which defeats the entire point of the data sheets.
Keep the drafts around 500 to 800 words. Location pages do not need to be long. They need to be specific.
Checklist for a Location Page That is Hard To Outrank
Run every page against this before it goes live. If a draft misses more than two, it goes back for a rewrite.
On the page itself:
- A local detail in the first two sentences, not the service pitch
- At least one real customer review from that exact area
- Neighborhoods or landmarks named, the ones you actually serve
- Pricing, travel fees, or response times specific to that location
- One local problem only that area has, like old pipes or permit rules
- LocalBusiness schema with areaServed filled in and tested
- A unique title tag and H1, written after the page, not before
- FAQ answers pulled from real customer questions, not invented ones
In the writing:
- No em dashes anywhere
- Openings that differ from your other location pages, checked side by side
- Short paragraphs, two lines maximum
- Mixed sentence lengths so the rhythm does not read machine-flat
- No AI filler words. If the draft says “seamless” or “elevate”, it is not done
- The same thing called by the same name throughout, no synonym rotation
The last group matters as much as the first. Google’s spam systems and AI detectors both key on uniform style across pages, so a set of pages that reads slightly different from each other is harder to flag and harder to copy.
If a competitor wants to outrank a page like this, they have to do the local research themselves. Most will not.
How I Use Claude To Write location Pages at Scale
Sameness creeps back in around page fifteen or twenty. Every intro starts to sound like a cousin of the last one, even with different facts underneath.
A few checks catch most of it:
- Read the first sentence of every page in one sitting
- A site: search for phrases you suspect are repeating
- Side by side comparison of two random pages, once a week
When two intros feel like siblings, rewrite one by hand. Two minutes of manual editing per page keeps the whole set honest.
We also rotate structure across the set. Some pages open with a review, some with a neighborhood detail, some with a pricing note. The master prompt offers those openings and Claude picks based on what the sheet contains.
The rotation matters more than it sounds. Google’s site-level quality signals look at patterns across pages, not just each page alone.
Claude for SEO After The Writing is Done
The draft is the visible half. Claude for SEO work on location pages also covers the pieces around it.
We use it to build FAQ sections from real customer questions pulled from call notes and Google Business Profile Q&A. Those short answers get quoted by AI assistants far more often than body copy does.
Schema is next. Paste your NAP details and service list, ask for LocalBusiness markup with areaServed filled in, then check the output in Google’s Rich Results Test before it ships. Claude gets the structure right almost every time, and almost is why you test.
Internal linking is the last piece. Hand Claude the list of live location URLs and it will map which nearby city pages should reference each other, which saves a tedious afternoon.
What Separates a Location Page That Works From One That Doesn’t
The pages that performed for us shared a single trait. Someone who lives there could read them and nod.
That standard is easy to check. Send a draft to the client’s technician in that city and ask what is wrong with it. Whatever they correct becomes the best line on the page.
Claude speeds up everything around that standard. It does not replace it. Treat the model as a fast writer with zero local knowledge, hand it the knowledge, and your location pages stop being a duplicate content risk and start earning their keep.
FAQ’s
Can Claude write location pages that rank in Google?
Yes, when each page is built from unique local inputs. Pages generated from one prompt with only the city name swapped usually stay unindexed.
How many location pages can I create with Claude at once?
Draft one page per chat to avoid repeated phrasing. With data sheets prepared in advance, ten to fifteen finished pages in a day is realistic.
Do I need a separate prompt for every city?
No. Keep one master prompt for structure and voice. The only thing that changes per page is the location data sheet you paste in with it.
SEO and digital marketing specialist. I also write about jobs abroad, relocating overseas, and learning languages. MBA in Marketing.