AI answer-engine visibility · Birmingham, AL
In your category. In the Birmingham metro. Probably this week.
There was no click to track and no ranking that would have warned you.
Entity records, review language, and the sources these models actually read.
Forty prompts. Six assistants. Seven days. One number you can act on.
What was in the chest
Pick a category. This is the shape of what comes back — who the assistants name, how often, and where you land against them.
TIMES NAMED — ACROSS 40 CUSTOMER PROMPTS
Illustrative run. The company names above are invented for this demo — no real business is shown as present or absent. Your audit uses the real ones, pulled from live assistant responses.
What changed
You probably felt this before you could measure it: rankings hold, impressions hold, the phone rings less. Here is what the data says, from sources with nothing to sell you.
None of those numbers tell you whether an assistant said your name. That is a different measurement, and almost nobody is taking it.
The gap
Rank tracking answers "where does our page sit in the results?" That question still matters. It is no longer the only question, and it is not the one your customer is asking an assistant.
An assistant does not return a list. It returns one answer with two to five businesses in it. You are in that answer or you are invisible. There is no page two to climb from.
Most local agency stacks — the rank tracker, the call tracker, the Google Business Profile dashboard — never send a single query to an assistant. They cannot tell you what ChatGPT said about your category yesterday, because they never asked it. That is not incompetence. It is a tooling gap that opened faster than the tools did.
Chapter break
In most categories, one business gets named far more than everyone else. Not always the biggest. Not always the best at the actual work. Usually the one whose name, address, reviews, and citations all agree with each other everywhere the models look.
The audit tells you who your dragon is, how far ahead it is, and which signals it holds that you do not. Then you go get them. No sword required.
THE INCUMBENT — NAMED IN 29 OF 40 PROMPTS
The offer
One category. One metro. A number you can act on, and a list of what to fix in the order it should be fixed.
The seven days
I build the 40 prompts for your category and lock the competitor set with you.
Every prompt, every assistant, multiple passes. Raw responses captured and stored.
What the models know about you, where they learned it, and where the record disagrees with itself.
Share of voice scored, fix list written and ranked, call booked.
Flat. Paid up front. Delivered in 7 days.
No retainer, no contract, no upsell required.
The honest part
Someone asks ChatGPT for a crawl space company. It says your name. They type your name into their phone, or just call you. There is no referrer, no UTM, no click to count.
Your intake form asks how they heard about you. They say "online," or "somebody recommended you." Your CRM files it under direct or word of mouth. So the channel growing fastest in your market is the one your reporting is worst at seeing.
If someone promises you clean AI attribution, they are selling something. I would rather hand you a leading indicator that is honestly measured than a revenue number that is quietly invented.
The mechanism
There is no secret incantation. There are four signal families, and most local businesses are quietly failing at two of them.
Name, address, and phone that match everywhere. Structured data that resolves. Clean separation from the three other businesses in the metro with a similar name. If the record disagrees with itself, the model hedges — and hedging means it names somebody else.
Models read the text, not just the average. Four hundred reviews that all say "great service" carry less weight than ninety that name the specific job, the specific neighborhood, and the specific problem you solved.
Every category has a source set the assistants lean on. It is rarely the one your SEO report tracks. Sometimes it is a trade directory. Sometimes it is a forum thread from 2023. The audit shows you the actual list.
Not keyword pages. Answer pages — written in the shape of the question, carrying the specifics a model can quote: pricing ranges, process, service area, and what disqualifies a job.
Qualification
Who runs this
I am Connor Hand. For six-plus years I have run paid media and lifecycle analytics — at a Birmingham ad agency and at a national tech company. My job was building the reporting that told executives which channels were real and which were rounding errors. I am good at telling the difference.
Tiny Mage runs enterprise-grade AI-visibility tracking that local agencies do not have access to. That is the edge, and it is a boring one: I can measure this, and most people selling it cannot.
I am not an AI company. I am an analyst who found a channel nobody was counting, in a city I live in, in categories where one job is worth more than most agencies charge for a month.
Birmingham, Alabama. One person. No account managers, no offshore team, no dashboard you will never open.
What comes after
The audit is a complete product and it ends cleanly. Most owners who buy it want the work done and the score watched. That is a separate decision, made after the report is in your hands.
No annual contract. The initial term is 90 days because nothing in this channel moves in 30 and I am not going to pretend otherwise. After that, cancel any month.
You are not required to take it. Plenty of people will hand the fix list to the agency they already pay. That is fine. It is your list — you bought it.
Four a month. I would rather be honest than busy.
Capacity
Each audit is roughly twenty hours of work I do myself. There is no team behind me and I am not going to pretend there is. When a month fills up, you go on the list for the next one and I tell you the date before you pay anything.
Currently booking: September 2026
Updated by hand, not by a countdown script. If it is wrong, email me and I will fix it.
Already spending $20k a month on ads? There is a separate engagement for that — acquisition, lifecycle, and the measurement layer that tells you which of your channels actually produced booked jobs rather than claimed conversions.
Tiny Mage Growth →Objections
No. SEO optimizes for position in a ranked list of links. This measures whether a model names you inside a generated answer. Some inputs overlap — a clean entity record helps both — but the measurement is different and so are most of the fixes. Plenty of businesses ranking first organically never get named by an assistant. That gap is exactly what is worth finding.
No, and be suspicious of anyone who does. These systems are probabilistic and the companies running them change them without telling us. What I guarantee is the measurement: after seven days you will know your exact share of voice, who is beating you, which sources they hold that you do not, and what to do in what order.
The models change. The inputs mostly have not. Entity consistency, substantive reviews, and presence in the sources a category actually cites have mattered to every version so far, and they mattered to search before that. That is why the remediation work stays useful even when the scoreboard moves.
Then ask them for three things: your share of voice across forty prompts and six assistants, the citation set Gemini pulled from in your category last month, and how those numbers moved since the month before. If they can produce all three, you do not need me. If they send you a blog post about AI, you have your answer.
Then you paid $1,500 to find that out, I will say so plainly in writing, and I will not pitch you a retainer. Knowing you are covered is a real result. It is also cheaper than a year of paying somebody to fix a problem you did not have.
Because in your category one job is worth $5,000 and up. If the audit surfaces a single fixable reason you are missing from answers your competitors are winning, and you recover one customer from it, the math is finished. If you are not in a category where that is true, this is the wrong purchase and I will say so.
The offer on this page is built for the Birmingham metro, because local answer sets are local and I know this market. If you are elsewhere with the same profile, email me. I will tell you honestly whether I can do it well.
Forty prompts. Six assistants. Seven days. One number that tells you where you actually stand, and a list of what to do about it.
The free version is three prompts and the raw output. No report, no call. It takes me ten minutes and it is the fastest way to find out whether this is worth $1,500 to you.
Free, ten minutes of my time
No report, no call, no sequence to unsubscribe from afterwards. You will see the actual answer the assistants give, with the company names in it. That is usually enough to tell you whether the $1,500 version is worth it.
REQUEST RECEIVED
I'll run the three prompts and email you the raw output, usually the same day. If you don't hear back within 24 hours, reply to that address and chase me.