AI visibility for Indian local service businesses: the aggregator profile is not your source of record

Most Indian local service businesses are recommended out of a profile they do not own. The aggregator listing, the maps pin and the review platform carry the facts, and the owner has a phone number and a WhatsApp catalogue. The measured answer says that is the wrong way round. Read ChatGPT and Gemini on 15 October 2026 and the page behind the answer belongs to the business itself, or to its agency, in 85.6% of 10,652 citations. Review directories account for 0.8%. India was 189 of the 597 usable readings in that run, and two of the 189 cited us. A business with no page of its own has nothing in the category that gets cited most, and in India that is the usual position rather than the exception.

What the city readings show, and what they do not

On 15 and 16 October 2026 we extended the run to cities: ten of them, six buyer questions each, both engines, 120 readings, all usable. We were not cited once in any of the ten. The three Indian cities came back like this:

City	Readings	Returned a source	Cited us	Local sources the engines used
Mumbai	12	11	0	searchmetric.in in 4 readings, goodfirms.co in 3
Delhi	12	12	0	nicodigital.com in 3, digiustad.com in 3
Bangalore	12	11	0	onecity.co.in in 4, aiseo.solutions in 4

Read that for the pattern rather than for the names. Every city produced its own small winner, and the winner was a domain that is specifically about that city. The six questions in each city were supplier questions in our own category, so this is evidence about how a city level answer gets assembled and not a measurement of any trade. There is no Pune figure and no Jaipur figure here, because we did not measure Pune or Jaipur. For a city we have not read, the honest number to carry is the market one.

The page an Indian owner can write in an evening

The thing the aggregator cannot hold is your terms, in your words, with a date. Put these on one page per service you actually sell:

• The price basis in rupees. Visiting charge, per unit, per hour, per sq ft, minimum order, and what makes it go up. A range with the month you last reviewed it answers the first question every caller asks. 102 of the 102 build spec conversations we ran asked for pricing in the reader's own currency.
• The service area as a list of localities and pincodes. Not a radius and not a city name. Andheri East, Powai, Chandivali, and the pincodes beside them. A radius is invisible to a reader and to a model.
• Hours that include the days you are shut. Festival closures, the weekly off, and the hour after which a call becomes an emergency callout at a different rate.
• The licence or registration you hold. Shop and Establishment registration, trade licence, FSSAI licence for anything involving food, GSTIN if you are registered. A number a caller can check is worth more than a line about quality.
• Who turns up. The named technician or the named person who supervises, and how many of them there are. Indian customers are choosing a person, and so is the answer that names you.
• How you take payment. UPI, cash on completion, advance, and whether you give a GST invoice.

Keep those facts identical everywhere they appear, because one wording repeated builds a single clear business while three variations build three weak ones. The source by source method for that part is in one set of facts, everywhere.

The aggregator profile is your second copy

Urban Company, JustDial, Sulekha, Practo and NoBroker are worth keeping complete and correct, because customers really do book from them and the profiles are structured and easy to read. They are a second copy of your facts rather than the record itself, and three limits are worth knowing:

• A profile cannot carry your price basis, only a headline rate the platform sets or approves.
• A profile cannot carry a locality list, so the platform decides which areas you are shown for.
• A profile goes stale in ways you will not notice, and a stale listing is the version that gets quoted back at you.

On the review side, the measured number settles an old argument: review directories were 0.8% of the citations recorded in that run. Collect honest reviews because customers read them and because IS 19000:2022 sets out how to publish them properly, and spend the quarter on your own pages. Which Indian directories are worth the afternoon is set out in reviews and directories for Indian businesses.

Both runs, and what neither of them measured

ChatGPT and Gemini, in both cases, and nothing else. Perplexity, Claude and Google AI Overviews were absent from the three market run and from the city run. Our product measures five surfaces; these datasets cover two, and the two never get folded into one claim. Nor is there a reading of any single trade here. The vertical questions in the three market run were worded generically, so there is no plumbing figure, no salon figure and no tuition figure, and every share above is a whole run, market or named city number. The method and the full three market breakdown are in what AI assistants actually cite across 597 readings.

Four things worth doing this week

• Write one page per service, each carrying the price basis in rupees, the locality list and the date.
• Ask an assistant for your trade in your locality, five ways, without naming your business, and save every source it cites.
• Check your aggregator profiles and your maps listing against the page you just wrote, and correct every difference.
• Put your licence number and the date you last reviewed the page at the bottom of it.

Questions Indian local service owners ask about AI answers

I get all my work from an aggregator. Why build pages?

Because the answer a customer now sees before they open any app is assembled from pages, and 85.6% of the citations recorded on 15 October 2026 went to somebody's own website. Keep the aggregator work and add the one asset it cannot give you, which is a page that states your terms in your words.

How much of a website do I actually need?

One page per service you sell, plus a contact page. Four or five pages of real facts do this job, and a thin page for every locality in the city does not, because there is nothing specific in it to repeat.

Should I publish rates when every job is different?

Publish the basis and the range, with the month beside it: visiting charge, per unit, minimum order, and the two or three things that move the figure. That is the question customers ask first, and a page that answers it gets used by both the caller and the assistant.

Do I need reviews on JustDial and Google to be named?

Reviews matter to the customer reading them and they are a small share of what these two engines cited, at 0.8% across 10,652 citations. Keep collecting honest ones, reply to them, and treat the pages on your own domain as the part of the work that moves the answer.

My locality is not in your city table. What number applies to me?

The India market figure: 189 readings on 15 October 2026, in which our own domain was cited twice. We measured Mumbai, Delhi and Bangalore directly and nothing else in India, so anything we said about another city would be a guess rather than a reading.

The locality list is the part nobody writes down

Ask an Indian service owner which areas they cover and you get six localities in ten seconds. Ask where that list is written and it is in their head, on a signboard, and nowhere a machine can read. Of everything on this page, writing those localities and their pincodes next to a rupee price basis is the single most specific thing an owner can publish, and specificity is the whole reason a small local domain beat much larger platforms in every city we read. This is also what we are paid to do, and 64 of the 102 build specs said to put that on the page instead of leaving it implied. AI Knows Us puts your customers' booking questions to five assistants and reports whose page answered. What a package of this work costs per location, line by line, is in local SEO services.