Industries and markets
Buyers ask by category, and by city.
Almost nobody asks an assistant a generic question. They ask for a category, in a place, with a condition attached. That shape is what the work has to match.
The pattern
Category, plus place,
plus a condition.
"A pediatric dentist in Chicago who takes Blue Cross." "An employment lawyer in Los Angeles for a wrongful termination case." "A SOC 2 compliance platform for a fifty person startup." Those are the questions that actually get typed, and each one is a filter the model has to resolve before it can name anybody.
A business gets named when it is easy to verify against all three filters. Category has to be unambiguous in your structured data. Place has to be consistent everywhere your name appears. The condition, whether that is insurance accepted, case type, company size or price band, has to be written down somewhere a model can read rather than implied by a photo.
The industries below are the ones where that gap is currently widest, which is another way of saying they are the ones where this work pays back fastest.
Industries
Where I do most of the work.
Healthcare and clinics
The highest stakes category for getting this wrong. Assistants are cautious about naming providers, so they lean hard on structured signals: verified NPI data, specialty taxonomy, accepted insurance, admitting privileges and hours that agree across every directory. Currently testing how much aligning those signals moves a practice from omitted to named, and how many crawl cycles it takes.
Request a healthcare prompt auditLegal
Practice area and jurisdiction decide everything. Attorney level entities get named more often than firm level ones.
Request a legal prompt auditHome services
Service area, license numbers and emergency availability are the three filters that decide who gets named.
Request a home services prompt auditB2B SaaS
Integration lists and honest comparison pages are what models quote when somebody asks for alternatives to a competitor.
Request a B2B SaaS prompt auditFinancial services
Credentials, registrations and a stated fee structure. Vague pricing is the fastest way to get skipped.
Request a financial services prompt auditHospitality and local retail
Menus, amenities and neighborhood context, written as text rather than baked into an image a model cannot read.
Request a hospitality prompt auditMarkets
Six metros I have tested.
Local context changes the answer more than most people expect. The same question asked in two cities returns different names, different sources, and sometimes a different level of caution from the model. These are the markets where I have run enough prompt testing to know what the assistants already believe.
These are working hypotheses from general practice, not published test results. Prompt testing for these markets is in progress and the findings will replace this section when they are ready.
- New York Currently testing whether assistants hedge in a market this dense, and whether the most clearly documented businesses are the ones named. Request a New York prompt audit
- Los Angeles Currently testing whether sprawling service areas confuse location signals, and whether neighborhood level clarity beats a single citywide claim. Request a Los Angeles prompt audit
- Chicago Currently testing how much weight earned mentions in strong regional publishers carry in this market. Request a Chicago prompt audit
- Houston Currently testing whether fast growing categories with thin coverage move from invisible to cited faster than established ones. Request a Houston prompt audit
- Miami Currently testing whether bilingual queries split the results, and whether businesses documented in both English and Spanish are named more often. Request a Miami prompt audit
- Seattle Currently testing whether a technical audience asks more comparison questions here, and whether structured comparison content answers them. Request a Seattle prompt audit
All of my work is remote. I work remotely from Karachi, Pakistan, with US clients. Market focus here means tested prompt history, not a local office in each one.
Putting it together
One category, one city,
one page.
The practical output is usually a small set of pages built to answer one intersection properly. Not forty near identical location pages with the city name swapped, which answer engines discount almost immediately, but a handful that genuinely say something different because the market genuinely is different.
A dental practice in Houston needs different proof than the same practice in New York, because the competing set is different and the conditions buyers attach are different. That is the level the work happens at, and it is why I cap the page count.
- Prompt inventory first.
- 25 questions across four engines in the audit, 40 in the monthly program, for your category and market. The four are ChatGPT, Perplexity, Gemini and Google AI Overviews.
- Gap map.
- Which questions name a competitor, which name nobody at all, and which return something factually wrong about you.
- Targeted pages.
- Built only for the intersections where a gap exists that is worth the cost of closing it.
- Re-test every thirty days.
- Same prompts, same engines, so the change is measurable rather than argued about.
Not sure where your
category stands?
Tell me your industry and your city. I will run the prompts and send back what the assistants say today.
Ask for a prompt gap check