There is no Dallas number on this site, and here are the four American cities that have one

The city level readings on this site cover ten named places and no others. In the United States those are New York, Los Angeles, Chicago and Houston, at 12 readings each, taken on 15 and 16 October 2026 from six buyer questions per city put to ChatGPT and Gemini. The other six are London, Manchester, Birmingham, Mumbai, Delhi and Bangalore. There is no Dallas figure, no Miami figure, no Seattle figure, no Phoenix figure and no Atlanta figure here, and saying so is what makes the four we did read usable.

A reader looking for their own city deserves a straight answer rather than a market figure quietly presented as a local one. This is the straight answer.

What exists, exactly

City	Readings	Readings that returned sources
New York	12	9
Los Angeles	12	11
Chicago	12	12
Houston	12	11

Six questions per city, asked on two apps, which is twelve readings. That is a small reading and it is described as one. It shows how an answer about that city was assembled on those dates, and it is not a survey of any trade in that city.

Why a market figure is not a city figure

We hold a United States figure from a separate and much larger run: 210 questions per market on 16 October 2026, 1,260 readings in total, 20,143 citations across 4,947 distinct domains. That run holds no city breakdown at all, because none of its questions named a city. Using its numbers to describe a city would be presenting a national reading as a local one, which is the single most common way a local claim gets made without evidence.

The two readings also disagree in a way that proves the point. The national bank surfaced no United States only agency domain in the cut we published. The city scan surfaced small city specific domains in three of the four American cities. Same country, same category, two instruments, two answers, because the questions differed.

What to do if your city is not one of the four

• Take your own reading. Six questions on two apps is twelve readings and about an hour. Our whole city scan is 120 readings and it is not a large undertaking.
• Write the city into the question text. Do not rely on a location parameter. The engine surfaces we read accept one, return success and ignore it, which we measured before building either run.
• Keep every company name out, starting with your own. Put a brand in the wording and the answer comes back about that brand, so the reading describes your wording rather than the engine.
• Record whether each reading returned sources at all, in its own column, before recording whether you were in it. Three of New York's twelve readings came back with nothing behind them, and a reading like that cannot be won with a page.
• Carry the date and the denominator. Twelve readings on a named date is a reading you can compare to your own next one.

Other work on this site covers several American cities on the search and media side rather than the citation side, such as what clicks cost in Dallas and how to split that metro, and none of those pages carries an AI citation reading, because none of those cities was in the scan. SEO in Las Vegas, where two separate audiences share one valley, is another of them, and it is a search and pricing page rather than a reading.

The rest of what these four readings do not contain

No per industry figure, because the six questions per city were buyer questions in our own category. No figure for a suburb, a county or a metro definition wider than the city named. Two apps, ChatGPT and Gemini, with the three further surfaces our product covers absent from all of it. And each count is a domain appearing in a source list rather than a business named in the answer text.

We sell this measurement and the writing that follows it, so the result that does us no good is published with the rest: across all ten cities in the scan, our own domain was cited zero times. The Google Search Console side of our own record, two domains with their windows and the export date on the page, is our proof page.

Why ten cities and not fifty

Because a reading is only worth publishing if it was actually taken. Ten cities at six questions on two apps is 120 readings, and every one of them exists as a recorded answer with its sources written down. Fifty cities would have meant either five hundred more readings or an estimate dressed as a measurement, and the second option is what this site exists to argue against.

The list will grow the same way, one city at a time, each with its own date and its own twelve readings. Until a city is on the list there is no figure for it here, and that sentence is doing more work for a reader than a wider number with nothing behind it would.

Questions about which cities have a reading

Which American cities did you measure?

New York, Los Angeles, Chicago and Houston, at 12 readings each on 15 and 16 October 2026.

Is there a figure for my city?

There is one for New York, Los Angeles, Chicago, Houston, London, Manchester, Birmingham, Mumbai, Delhi and Bangalore. Anywhere else has no figure on this site, and dressing a market reading up as a local one would be a different claim in the same words.

Can I use the United States figure for my city instead?

You can, provided the label travels with it and says so. None of the national run's questions named a city, and our own two instruments surfaced different winners for the same country, which is the reason the label matters.

How do I measure my own city?

Six buyer questions with the city written into the wording, asked on two named apps, with no company name anywhere. Record whether sources came back, then who was in them, then the date.

Which engines were used?

ChatGPT and Gemini. The wider category run that sits behind the city scan, with its full method, is published as 597 readings across ChatGPT and Gemini in three markets.

The cities we did not read are the reason to trust the four we did

A dataset that answers for every city is a dataset that has stopped measuring and started estimating. Ours answers for ten places, four of them American, at twelve readings each, on two named dates. Everything outside that list is an open question, and an open question with a method attached is worth more to you than a number somebody extrapolated.