AI visibility for a business in Chicago: every reading returned a source and no source led
All twelve Chicago readings on ChatGPT and Gemini on 15 and 16 October 2026 came back sourced, with four sources tied on three citations each. In a field that flat, agreement between your own records decides the citation.
Published by AI Knows Us · Updated 16 October 2026
Chicago was the only one of the four American cities in this run where every single reading came back with a source attached, and where nothing led. The Chicago reading was taken on 15 and 16 October 2026, on ChatGPT and Gemini, six buyer questions apiece. Twelve readings, twelve with sources, and 0 citations for us. Four sources finished level on 3 citations apiece: agencies.semrush.com 3, chicagoseoagency.co 3, paramdigitalmarketing.com 3 and goodfirms.co 3. A tie at the top of a fully sourced field means the answer is being reassembled from a different handful of sources each time it is asked, and that changes the job. In a city like this, what decides whether you get named is not the strength of one page. It is whether your facts agree with each other wherever an engine happens to find them.
Four sources, level on three citations each
| Source | Citations of 12 | Type of site |
|---|---|---|
| agencies.semrush.com | 3 | The same software company's agency listing that led in Los Angeles |
| chicagoseoagency.co | 3 | A small agency whose domain name is the city plus the service |
| paramdigitalmarketing.com | 3 | An agency site that does not name a place at all |
| goodfirms.co | 3 | A worldwide agency directory, also in New York's top four |
Twelve readings, twelve of them sourced, and the leaders separated by nothing. One global listing, one city named local shop, one unbranded agency and one more directory, each carrying the same weight. For comparison, in the other three American cities we read on the same two days, between one and three readings came back with no source at all. Chicago always had something the engines wanted to cite.
What a flat field does to the work
When one source dominates a city, the task is legible: there is a page in front of you and you work out what it has that you do not. A four way tie at 3 removes that target. Each time one of these questions is asked, a different mix of sources assembles the answer, and your business either appears consistently in that mix or it drops out of most of them.
The practical consequence is that contradictions cost more here than thin pages do. If two directory entries, a profile and your own footer disagree about a suite number, a trading name or which location is the main one, every source carrying the wrong version competes with the one carrying the right version. In a field with a clear leader you can out-publish that problem. In a flat field you have to settle it.
The Chicago part of the problem: the market crosses state lines
This is the one metro in the top American markets where "local" routinely means three states. The built up area runs from Kenosha in Wisconsin, through Illinois, into Lake, Porter and LaPorte counties in Indiana. A company that trades as Chicago can be incorporated in Indiana, licensed in Illinois, and listed at a Wisconsin address by a directory that picked up an old record. Different secretaries of state, different licence registers, different area codes on the phone numbers, and a genuine question about which address is the business.
Chicago businesses are better placed to fix this than most, because reconciling a record is what a great deal of this city's economy does for a living. The railroads interchange here, the derivatives exchanges clear here, and the freight and food distribution networks run on records that have to match at both ends. Treating your own name, address and licence data as a reconciliation exercise is the same discipline, applied to the thing the engines read.
The fields that have to match, and where they usually do not
| Field | Where the wrong version usually lives |
|---|---|
| Legal entity name | The state filing says one thing, the website footer says the trading name, and a directory has merged the two |
| Street address and suite | A suite or floor number present in one listing and missing in another, which reads as two addresses at one building |
| State | An Indiana or Wisconsin address on a business that presents itself as Chicago, with neither statement wrong and no sentence reconciling them |
| Phone number | A call tracking number on the website against the real line in every listing |
| Main location | Two or more sites with no page saying which is the head office and what each one does |
| Licence or registration | A number with no issuing state named, which a machine cannot check against anything |
| Price | Dollar figures on the website with no date, and a different set of figures in a listing nobody updated |
What to publish once the records agree
- One page of trading facts, written as plain text, that every other source is corrected against rather than the other way round.
- A sentence that reconciles the state question out loud: the suburb and state you are registered in, and the fact that you serve the Chicago market from there. Say it so nothing has to be inferred.
- A page per location that names what that location actually does, because two sites with identical copy give an engine no reason to distinguish them.
- The registration number and which state issued it, so somebody who is not you can confirm it.
- Prices in dollars with the date they were set, and the kind of work you turn down.
Two things to keep straight about the numbers on this page. The flat field described above is a finding about ChatGPT and Gemini on two days in October and about nothing else. Perplexity, Claude and Google AI Overviews sit in the product we sell and were not in this run, so no claim here reaches them. And we have a direct stake in it: we sell the measurement and we write the pages that fix it, which is why the Chicago row includes our own zero rather than somebody else's. The method and the full source mix sit in our record of what the two engines actually cited across 597 readings, where a vendor's or an agency's own site took 85.6% of every citation counted.
AI visibility in Chicago: questions that come with a reconciled record
Our head office is in the Loop and two branches are in Indiana. What do the pages say?
Give each site its own page with its own address, its own state and a line about what it does that the others do not. Then put one sentence on the head office page saying which is the head office. The failure case is three pages with the same copy and three different addresses, because that gives an engine no way to tell which one to name.
Four sources tied at three citations. Which one do I go after?
None of them. In a flat field there is no page to displace, so the gain comes from being correct in enough sources that whichever mix the engine assembles, your facts are in it and they agree. That is a different and cheaper piece of work than trying to beat a specific competitor.
Does a call tracking number hurt this?
It creates a contradiction if the tracked number is the only one on your site while every listing carries another. Publish the real line in text somewhere a machine will read it, and keep the tracking number for the places you need attribution from.
goodfirms.co was cited three times here. Should we buy a listing?
Keep your entry accurate if you have one, because a stale listing in a cited directory is one of the contradictions described above. As a channel the measured share is small: of the 10,652 citations we recorded across 2,708 distinct domains on 15 October 2026, review directories accounted for 0.8%.
Does being a local shop matter in these answers?
One of the four tied leaders carries the city in its domain and one does not, so proximity is neither the price of entry nor a disadvantage. If you are weighing a shop in the metro against one elsewhere, our piece on what proximity to your metro actually buys you sets out what changes and what does not.
How often should this be re-measured?
Every month, always on identical wording, or the comparison means nothing. Ask about the category and the place. The moment your own name appears in the question, the reading tells you only that you can type.
A city that always had something to cite
Twelve readings out of twelve came back with sources in Chicago, which says the engines are never short of material when somebody asks a local buying question here. Four sources level at 3 says none of that material has won yet. For a business with multiple sites, a state line running through its market and a record in three different registers, that combination is the most encouraging shape a city can take: the gap is not a competitor, it is a set of disagreements in your own data that you are allowed to go and settle this week. Settling it is billable work and we bill for it, at aiknowsus.com. The Chicago row above is what our own records have bought us so far, which is nothing, and we would rather publish that than a case study.