Do City Landing Pages Still Work for AI Search?
By the alphaa team — we run AI-visibility scans across thousands of businesses and look at a lot of city-page templates. Last updated 21 August 2026.
Short answer: yes, city landing pages still work for AI search — but only the ones that contain something true about that specific city. An AI assistant retrieves passages, not pages, so a page earns a citation by containing a passage worth lifting. Fifty pages generated from one template with the place name swapped contain one passage repeated fifty times, which gives the engine nothing to choose and gives your site a duplication problem. Three genuinely local pages beat fifty templated ones, reliably.
Why the honest answer is "it depends on the page," not "yes" or "no"
The advice you will find is split, and both camps are half right. "City pages are dead" comes from people who watched Google devalue doorway pages. "City pages are essential" comes from agencies whose product is producing them at volume.
The mechanism explains the split. When an AI assistant answers "best [service] in [city]," roughly this happens:
- The query is expanded into several related searches — the service, the city, sometimes the neighbourhood or a modifier the user did not type.
- Those searches run against a retrieval layer — a search index, the engine's own crawl, a maps or business dataset, or some combination.
- Candidate passages come back and are ranked mostly on how well they answer the specific question, not on how well the page is optimised.
- The model composes an answer from the passages that survive, and cites the ones it leaned on.
Every step of that rewards distinctive, specific text and punishes near-duplicate text. A templated city page is near-duplicate by construction. That is the whole story: the format is not the problem, the sameness is.
The one-sentence test for a city page
Before you publish, apply this: if you swapped the city name for a different city, would any sentence on the page become false?
If the answer is no, the page has no local content — it is a template with a variable in it, and it will not earn a citation, because there is nothing on it that a model could only have learned from that page. If the answer is yes, and yes several times over, you have a real page.
This is a stricter test than it sounds. "We proudly serve the Tucson area with reliable, honest service" survives no swap. "Our Tucson crews carry permits for the historic Barrio Viejo district, where trucks over 24 feet cannot access most streets" survives it emphatically.
What actually goes on a city page that earns citations
Facts that are true of that place and nowhere else. In practice, a page that works usually contains four or five of these:
- Local operating facts. Which crew or office covers it, response times from where you are actually based, days or seasons you run there, the neighbourhoods and adjacent towns included.
- Local regulatory or logistical detail. Permits, HOA rules, building certificate-of-insurance requirements, parking and access restrictions, county inspection quirks, state licensing that differs. This is the highest-value category, because almost nobody publishes it and it is exactly what a person asking an assistant wants to know.
- Local pricing reality. What a job in that market typically runs and why it differs from the next market over. Ranges with the variables named, not invented precision.
- Named local proof. Reviews from customers in that area, projects with the neighbourhood named, a local partner, a local trade association or chamber membership.
- Local contact reality. If you have a real staffed address there, say so with hours. If you do not, say you serve it from your main location — that is a trust signal, not a weakness, and inventing a virtual office is the fastest way to create the contradictory-address problem below.
How templated pages actively hurt
Three concrete failure modes, all of which we see in scans:
Entity dilution
Engines build a picture of your business as an entity: one name, one identity, a defined service area. Eighty pages each claiming a different city as home blur that picture rather than extend it, and a blurred entity is harder to retrieve for any query, including the one city where you are genuinely strong. The mechanics are in entity SEO and how AI identifies your business.
Contradictory addresses and phone numbers
Templated city pages very often carry a per-city phone number or a virtual-office address. Those details then propagate into directories, and now the engines have several conflicting records for one business. Conflicting facts do not average — they suppress, because the safest thing for a model to do with a contradiction is not to assert either version. See directory listings and NAP consistency.
Crowding out the pages that would have worked
Retrieval is competitive within your own domain as well as across the web. Fifty thin pages sit between an engine and your three good ones. This is the most common version of the problem: the good page exists, and it is buried in a template farm.
How many city pages should you actually have?
A rule of thumb that holds up: one page per place where you could write 400 words of true, specific detail without straining, and no more. For most single-location businesses that is one to five pages. For a business with real crews or offices in multiple markets, it is one per market — which is a different job, closer to the one described in multi-location AI visibility.
If you serve thirty towns but only have things to say about four of them, the right structure is four real pages plus one honest service-area page that lists the other twenty-six as places you cover. The list still helps — it gives an engine a literal string match for those town names — but it does so without thirty near-duplicate documents attached.
Building one: a worked structure
- H1 with the plain query. "[Service] in [City], [State]". Not clever, not a slogan.
- A 40–60 word answer-first paragraph. Who you are, what you do in that city, the area covered, and one qualifying fact. This paragraph is the passage most likely to be lifted, so it should stand alone and be true out of context.
- Coverage. Neighbourhoods, ZIP codes, adjacent towns — as text, in a list.
- The local specifics. Two or three sections from the fact list above, each under a question-shaped heading a customer would actually type.
- Local proof. Two or three reviews or project notes from that area, with the area named.
- A short FAQ. Three or four questions specific to that market, answered in two to four sentences each — the extractable format described in how to write content AI engines will quote.
- Schema. LocalBusiness (or the right subtype) with areaServed naming the city. Use one canonical address — your real one — across every city page rather than a per-page address. The full pattern is in schema markup for AI search.
And make sure the text is in the HTML the crawler receives. City pages are frequently built as client-rendered components fed by a data file, which can leave AI crawlers with an empty shell — see why AI crawlers cannot read your JavaScript site.
Common questions
Will AI engines penalise my city pages?
There is no penalty in the sense classic SEO uses the word. Retrieval simply does not surface passages that are duplicative and unspecific, and a diluted entity is harder to match. The effect is exclusion by irrelevance rather than punishment — which is worth understanding, because it means there is no penalty to "lift" either. Improving the pages is the only lever.
Should I delete the templated city pages I already have?
Consolidate rather than delete outright. Keep the handful you can make genuinely specific, improve those, and redirect the rest into a single service-area page. Preserve any page that has real backlinks or real traffic. Then re-check your visibility after a few weeks — the re-crawl and re-indexing lag is real, as covered in how long AEO takes to work.
Can I use AI to write the city pages?
For structure and drafting, yes. For the local facts, no — a model does not know your permit rules, your crew locations or your market pricing, and if it invents them you have published errors under your own name. The part that earns the citation is precisely the part a generator cannot supply. Use it to write around facts you provide.
Do city pages help if I have no reviews in that city yet?
Less than you would like. Off-property corroboration is usually the deciding signal in local AI answers, and a page asserting presence with nothing outside your own domain confirming it is a weak input. Build the page, then get two or three reviews that mention the area — see why Google reviews decide your AI visibility.
Are city pages better than a Google Business Profile service area?
They are complements, and the profile usually matters more for near-me style queries. Set the service area properly first — it feeds the map and business datasets engines lean on — then use pages to answer the questions a profile cannot hold. The profile side is in Google Business Profile and AI answers.
The bottom line
City pages are not dead and they are not magic. They are a container, and AI search only rewards containers with something specific inside them. The test is the swap test: change the city name, and if nothing on the page becomes false, the page was never about that city. Write the four pages you can genuinely fill, list the rest honestly, keep one canonical address, and put the local knowledge you already have in your head onto the page — that knowledge is the part no competitor and no generator can copy.