Twelve Delhi readings, twelve sources and no leader: what ChatGPT and Gemini used to answer buyers here
Every question we asked here came back with a source, the top three sit level at three readings each, and our own domain was in none of them.
Published by AI Knows Us · Updated 16 October 2026
Delhi is the most open city we have measured, and the reason is that nothing owns it. Twelve readings, six buyer questions on ChatGPT and the same six on Gemini, dated 15 and 16 October 2026: every single one came back with a source, our own domain was cited 0 times, and the sources used most were goodfirms.co in 3 readings, nicodigital.com in 3, digiustad.com in 3 and adexorb.com in 2. The top three are tied. Three of the four are small Indian agency sites rather than large platforms, so there is no entrenched leader here, only a field any business could out specify. The catch is the second finding: a buyer types Delhi and means any of three jurisdictions, and the page that gets quoted is the one that says which.
Twelve readings, twelve sources, nothing in front
Delhi returned a source in every reading, where Mumbai and Bangalore each returned one in 11 of their 12. This city gave the engines the most to work with and still produced no favourite. A tie at the top of a 12 reading set is not a league table, and that is the point: one reading separates the first source from the fourth.
The one global name in that list did not lead. goodfirms.co keeps a listing page for every city on earth, and here it was matched, not beaten, by two agency sites that cover this region only, which is a fair measure of what scale is worth on a question this specific. nicodigital.com is worth a second look as well: the same domain was cited in 44 of the 189 India readings in our 15 October 2026 run, and now in 3 of 12 Delhi readings. One small agency publishing steadily is holding position at national level and at city level at the same time.
Hold the figures at the size they are. Each city got six questions of the kind a buyer asks about a supplier in the category we trade in, which makes this a picture of how a Delhi answer is assembled rather than a verdict on any industry in the city. Two engines were read and they are named: ChatGPT and Gemini. Nobody read Perplexity, Claude or Google AI Overviews in this run or in the 597 reading run behind it, which is published with its question set at what AI assistants actually cite across 597 readings. Three of the five surfaces we sell measurement for therefore have no figure here, and stretching this one to cover them would be dishonest.
Your buyer types Delhi. Your paperwork says 07, 06 or 09.
This is where Delhi differs from every other market in India. The word covers three states and union territories, three tax jurisdictions and three labour departments, and your own registrations already declare which one you sit in, in codes a machine reads far better than prose.
| What the buyer says | Where the office actually is | First two digits of your GSTIN | State letters in your CIN | Shop registration under |
|---|---|---|---|---|
| Delhi | Delhi | 07 | DL | Delhi Shops and Establishments Act, 1954 |
| Delhi | Gurugram or Faridabad | 06 | HR | Haryana Shops and Commercial Establishments Act |
| Delhi | Noida or Ghaziabad | 09 | UP | Uttar Pradesh Shops and Commercial Establishments Act |
The first two digits of a GSTIN are the state code and anyone can look a GSTIN up on the GST portal. A company's identification number carries the state it was registered in, and the Ministry of Corporate Affairs publishes company master data for anybody to read. So if your website says Delhi, your invoice starts with 06 and your filings say HR, you have handed a machine three versions of where you are. That is the most common entity problem in this region and it is free to fix.
The fix is to pick the true one and let the rest be service area. Put the registered office in text on the contact page with the state named, then list the places you really work in: Connaught Place, Nehru Place, Okhla, the Gurugram and Noida sectors you cover. The phrase Delhi NCR is not an address and cannot be matched to anything.
What the capital's buyer mix changes about the answer
Delhi's economy is trade, distribution, institutional buying and professional services, and that shapes the questions. A large share of local demand runs through processes rather than impulse: central government and public sector purchasing through the Government e Marketplace, export business that needs an Importer Exporter Code from the DGFT, and wholesale buying across Okhla, Naraina, Mayapuri and the manufacturing belt in Noida. A buyer running one of those does not ask who is best, but who can supply, invoice and document the job.
So a Delhi page answers in that register. Name the registrations you hold, GSTIN, Udyam, IEC and a GeM seller account where you have one, and say what you can issue: a GST invoice, a performance bank guarantee, an e way bill, a signed service agreement. State the price basis in rupees with the month attached, because a quote with no basis gets compared to four others on price alone. What the same brief costs tier by tier is set out in what each tier of work costs per month in India, and publishing your own version of that table is the single most quotable thing on a supplier page.
The capital also hands you third party confirmation that most cities cannot. FICCI, ASSOCHAM, PHDCCI and the national traders' bodies are headquartered here, and a member page or a speaker listing is a statement about you that you did not write. In a field where the top three are level at three readings, one independent page naming your firm and your specialism counts for a lot.
The order that works in an open field
- Settle the address first: one registered office in text, the state named, the service area as a list of real localities and sectors.
- Reconcile the codes: GSTIN, CIN state letters, shop registration and maps listing all pointing at one place.
- Every service you sell gets its own page, and each page carries a rupee basis, the documents you can issue and the date both were last checked.
- Collect one independent mention a quarter from a chamber, an industry body or a trade paper, and link it from the page it supports.
- Ask both engines your buyer's question in its Delhi, Gurugram and Noida forms, and keep all three source lists. Those answers are rarely the same.
What NCR suppliers ask us once they see the Delhi numbers
Our office is in Gurugram. Can we still claim the Delhi question?
Yes, by being exact. Gurugram is the office, Haryana is the state, and the Delhi localities you work in are a named list further down the same page. A supplier that says this plainly reads as a real business in the region, and your invoices and filings agree with the page.
Does being tied at three readings mean the top sources are weak?
It means the position is unclaimed. A 12 reading set with a three way tie at the top is a field where a page with real specifics can take the slot, which is a very different starting point from a market where one domain holds most of the answers.
Do we need separate pages for Gurugram and Noida?
Only where the work is genuinely different, and in this region it often is: a Gurugram corporate buyer runs a procurement process, a Noida manufacturer buys on specification and documents, and a Delhi trader buys on terms and speed. One page per real buyer, each naming the place and what that buyer needs, beats three near identical pages with the city swapped.
Which engines and how many readings?
Two engines, named: ChatGPT and Gemini. Twelve readings for this city inside a 120 reading sweep of ten cities on 15 and 16 October 2026, which returned no citation to us in any of the ten. The three remaining surfaces our product reads were not part of it.
Do you have the same figures for Jaipur or Chandigarh?
No. The sweep covered ten cities and this was its only northern one, so nothing in it can be read across to Jaipur, Chandigarh or Lucknow. Those cities take the national reading instead, where 2 of 189 India readings on 15 October 2026 named us.
We sell to government buyers. Does any of this apply?
It applies to the research that happens before the tender. Procurement teams ask an assistant who does this kind of work before they shortlist, and the pages that answer name the registrations, the documents and the delivery terms in plain text.
An open field closes eventually
The Delhi result is the most encouraging number in this dataset, because a three way tie at three readings is what a market looks like before anybody has decided it. Somebody will publish the dated, rupee priced supplier pages for this region and hold the position for years, and on 15 and 16 October 2026 it was still nobody. Our commercial position is plain and belongs in writing: AI Knows Us sells the measurement that tells a firm whether its Gurugram address is quietly losing it the Delhi question, and then sells the pages that settle it. Nine citations in ten, 85.6% of the 10,652 we counted on 15 October 2026, landed on a page its own business or agency had published. That is why the writing is the product here rather than an add on.