Every week a local business owner tells me some version of the same thing: “My Google ranking is fine, but people keep telling me they found my competitor through ChatGPT.”
When somebody types “best chiropractor near me for sciatica” into ChatGPT, they get one short answer with two or three businesses named in it. There is no page two. You are either in that answer or you do not exist for that customer.
This is the process I run for every client to get them into those answers. It is the same six steps whether the business is a chiropractor, a med spa, or a roofer.
Build the prompt list your customers actually type
Keywords and prompts are not the same thing. Nobody types “chiropractor Tampa” into ChatGPT. They type “I threw my back out and I need someone who can see me this week in South Tampa, who should I call.”
So the first thing I do is write out 30 to 50 real prompts in four buckets:
- Discovery: best [service] in [city], who should I go to for [problem]
- Comparison: [client] vs [competitor], is [client] any good
- Problem first: the symptom or job described in plain English, no business type mentioned
- Brand: tell me about [client name], is [client] legit
I get these from three places: the questions the front desk answers on the phone all day, the review text on their Google profile, and the “people also ask” style follow ups ChatGPT itself throws out when you run a first prompt. That list becomes the scoreboard for everything after this.
Step 2: Get a real baseline before you touch anything
You cannot improve a number you never wrote down. Before I change a single page I run that whole prompt list through the models and record where the client stands.
Two numbers matter at this stage. The first is share of voice in AI answers, which is how often the client gets named compared to everyone else in the market. The second is position, meaning whether they show up first in the answer or buried in a sentence at the bottom.
I use Cairrot’s ChatGPT rank tracker for this because running 50 prompts by hand once a month is not a business, it is a chore, and hand running also gives you garbage data. ChatGPT personalizes to your account and your history, so if you check your own client from your own logged in browser you will see them more often than a stranger would. The tracking needs to be clean and repeatable or the whole baseline is fiction.
Write the starting share of voice down. That single number is what you show the client in 90 days.
Find out who actually owns the citations in your market
This is the step most people skip, and it is the one that decides whether the next 90 days work.
When ChatGPT answers a local question, it is pulling from sources. Almost never the business’s own website at first. It pulls from directories, “top 10” listicles, local news, Reddit threads, and review aggregators. If you do not know which sources it is pulling, you are guessing at what to fix.
So I pull the full citation list for the prompt set and sort it by who owns each source: the client, a competitor, or a neutral third party site. Citation and mention tracking splits it out that way, and there is a filter for neutral domains that strips out you and your competitors so you are only looking at the sites the model trusts to decide the market.
That neutral list is your target list. If the same three directories and one listicle keep showing up across 50 prompts in a market, getting your client onto those four sources is worth more than 20 blog posts.
The other half of this step is reading the actual language. Mentions and citations are different signals. A mention is the AI naming the business in the answer. A citation is the AI linking the page as a source. You want both, and you want to read the exact wording the model uses about your client, because “affordable but hard to get an appointment” is a review problem, not an SEO problem.
Confirm the AI crawlers can actually read the site
Now the technical layer. It is boring and it is where about a third of local business sites are quietly broken.
GPTBot, ClaudeBot, PerplexityBot and the rest have to reach your pages and parse them. A lot of small business sites are built so the content only appears after JavaScript runs, or the services live inside an image, or a plugin is blocking the bots in robots.txt without anyone realizing it. If the crawlers never make it in, everything in step 5 is wasted effort.
I check three things: whether AI crawlers are hitting the site at all and which pages they hit, whether the important pages render server side, and whether the site has clean schema and an llms.txt file telling the models what is here.
The AI crawl monitor is what I use to see the crawl side, because it separates the bots by name and shows frequency and which URLs they are pulling. When you publish something new, crawl data is also how you know the models have picked it up, usually a week or two before you see any movement in the answers.
Build the pages and the placements the model will pull from
Everything up to here is diagnosis. This is the part that actually moves the number, and it runs in two lanes at once.
On the site: one page per service and one page per problem, written to answer the prompt directly in the first two sentences. Not a hero image and a paragraph about “our commitment to excellence.” A direct answer, then the detail underneath it. Location pages for every area they serve. An FAQ block on each one using the real phrasing from step 1. Schema on all of it so the entity is unambiguous.
Off the site: the neutral sources from step 3. Directory listings claimed and filled out completely, the Google Business Profile categories and services fully built out, review volume and recency pushed up, and placement on whichever local listicles and publications keep getting cited in that market. For most local businesses this off site half is heavier than the on site half, because the models trust third party consensus more than they trust your own about page.
Report the three numbers that matter, every month
I report the same three things every month so progress is not a vibe:
- Share of voice across the prompt set, against the same competitors, month over month
- Citations earned, meaning which of the client’s pages and placements are now being pulled as sources
- AI traffic and leads, the people who actually landed on the site from an AI answer and then called or booked
That third one is the one owners care about, and it is the one most agencies cannot show. It takes GA4 wired up properly so AI referrals are separated out instead of getting dumped into direct traffic. AEO reporting that pulls all three into one dashboard is what I send clients, because a screenshot of a ChatGPT answer is a nice thing to text somebody and a terrible way to prove a retainer is working. If you want the longer breakdown of which metrics are worth tracking, Cairrot has a good rundown of the AEO metrics that replaced old SEO KPIs.
How long this takes
First movement usually shows up somewhere between week 4 and week 8, and it almost always shows up in the long tail problem prompts first. The short competitive ones like “best [service] in [city]” are the last to move because those are the ones every competitor is fighting for.
Ninety days is a fair window to judge it. Anyone promising you the top spot in ChatGPT in 30 days is guessing, and anyone telling you it is impossible to influence has not tracked it.
FAQ
Can you actually rank in ChatGPT, or is it random?
You can influence it. It is not a ranked list like Google, it is a generated answer, so the goal is being named and cited often enough across a set of prompts that you show up in most of them. That share is measurable and it moves when you change the sources the model pulls from.
Do I need to be number one on Google first?
No. Strong SEO helps because the models lean on search results, but I have moved clients into AI answers who sat on page two of Google. Directory presence, reviews, and third party mentions do a lot of the work.
Why does ChatGPT show my business to me but not to my customers?
Because it remembers you. Logged in, with your own chat history, ChatGPT is far more likely to surface your own business back to you. That is why baseline tracking has to be done with clean prompts, not from your own account.
How much of this can a business owner do themselves?
The prompt list and the Google Business Profile cleanup, easily. The tracking, the citation analysis, and the placement work are where it turns into a real time commitment, which is usually the point where people call us.
Want to see where you stand right now?
We run this exact process for local service businesses. If you want to know how often ChatGPT is naming you versus the shop down the street, get in touch and we will pull your current AI visibility and walk you through it.