VisiblyRank
All posts
By VisiblyRankChatGPTLocal AI searchGuides

How to get recommended by ChatGPT in your city

A practical order of work for local businesses: find out what ChatGPT says today for your city prompts, see what it cites, check that OpenAI's crawlers can reach your pages, then close the source gaps.

When someone asks ChatGPT for "a good family lawyer in Manchester", the answer usually names a few firms and links to some sources. If your firm isn't named, those sources are the first place to look. When ChatGPT searches, it builds the answer from the pages it found, and for a single local prompt that's a fairly short list.

So the job is less mysterious than it sounds. Find out what ChatGPT says now, find out what it's reading, make sure it can read you, and then fix the gaps one source at a time. OpenAI's own help page is clear that ChatGPT ranks search results on multiple factors and placement is not guaranteed. Nobody can promise you a spot. You can stop losing it for reasons you could have fixed.

Step 1: write down your city prompts

Start with the questions a local buyer actually asks. For a multi-location business that means a set per city. Mix the explicit forms ("best physiotherapist in Leeds", "physio near Headingley") with the implicit ones that don't name a place ("physio who treats running injuries", "physio open late"). Add the comparison and decision prompts too: "is [your brand] or [competitor] better for back pain?"

Keep the list short at first. Twenty prompts per city you care about is plenty to learn from. Our guide on choosing prompts to track covers how to build and group them.

Step 2: see what ChatGPT says today

Run each prompt with search on and record the businesses named, the order, and how each is described. A one-line "known for short waiting times" is the kind of phrase that decides a click.

Where you run the test from matters. OpenAI's help page says ChatGPT may use an approximate location based on your IP address, may rewrite a "near me" question into a search that includes the city, and can use saved memories when rewriting. If you test from your desk, you're seeing your desk's answer. Use a clean session, and if you're checking a city you're not in, run the prompt with the city named and accept that it's an approximation. We cover the difference in how AI engines decide an answer is local.

Step 3: list what it cites

For every answer, open the sources and sort them by type. In local answers you'll typically see some mix of business websites, local news and city guides, review and listing sites, and community threads.

Then ask two questions. Do the businesses ChatGPT names turn up in several of the cited sources, or just one? And is your own location page among the sources at all, given that it exists? Whatever the answers, they tell you where to work.

One note on listings. For restaurants, OpenAI documents that ChatGPT can show reservation availability from OpenTable, Resy, and Yelp, and it says ChatGPT may share location with third-party providers for things like nearby business listings. Listings and reviews clearly feed local answers. Keeping them accurate is real work, and it belongs to whoever runs your listings or local SEO. We stay on the AI side: the pages and sources ChatGPT reads and cites.

Step 4: check that OpenAI's crawlers can reach your pages

OpenAI's crawler documentation draws a clear line. OAI-SearchBot is the crawler for ChatGPT search, and sites that opt out of it won't be shown in ChatGPT search answers. OpenAI recommends allowing it in robots.txt and allowing requests from its published IP ranges. ChatGPT-User is different: it fetches a page when a user's conversation calls for it, and it isn't used to decide whether content appears in search.

So check three places for your location pages specifically: robots.txt, your CDN or firewall rules, and whether the address, hours, and services are in the HTML or only appear after JavaScript runs. Store locators are a common blind spot. We go through the location-page version of these checks in can AI crawlers read your location pages?, and the general version in can AI crawlers reach your site?. OpenAI notes that a robots.txt change can take around 24 hours to register for search.

Step 5: fix the cited-source gaps

Now compare. For each city prompt you have a list of cited sources and a list of businesses named. A gap is a source that ChatGPT trusts for that prompt where you're absent or described badly.

Gaps on your own site are the fastest to close. If the cited pages for "physio near Headingley" are competitors' location pages that state treatments, prices, opening hours, and the streets they serve, and yours is a template with an address and a map, the fix is obvious. Write the page a local buyer would want: what you do at that branch, who it's for, when it's open, how to book, what's different about it. Plain facts in text, near the top.

Gaps on other domains take longer. If a local paper's "best physios in Leeds" piece keeps getting cited, the question is whether you deserve to be on it and who to ask. If a city subreddit thread is cited, see our notes on Reddit and local forums in local AI answers. Don't chase every source. Prioritize the ones cited under several of your prompts.

Step 6: measure with city or coordinate monitors

Answers change by run and by week, so a single check doesn't tell you whether a fix worked. You need the same prompts tracked on a schedule, from the right place.

For local clients we track this in Promptwatch. Monitors run the prompt panel daily per model with a set location. State and city tracking starts on the Business plan ($579/mo) and is on every agency plan. Where the radius matters, as with a clinic competing street by street, local GEO/SEO targeting pins the monitor to exact coordinates between 20 m and 5 km, so ChatGPT answers as if asked from that spot. That's on Business, Agency Scale, and Enterprise.

Citation analytics then shows which pages ChatGPT cites for each city, and Page Tracker follows a specific URL, yours or a competitor's location page. Agent Analytics shows whether OAI-SearchBot and ChatGPT-User actually fetched your location pages, using IP-verified logs, and the crawl-to-citation path shows whether a fetched page went on to be cited. When a page gets crawled and still isn't cited, the problem is content or competition. When it never gets crawled, no rewrite will help.

If you want us to run this for your locations, write to hello@visiblyrank.com.