SEO

How to Get Your Business Recommended by ChatGPT in 2026

Rakesh Ranjan Founder, Markital
13 min read
Markital article: how to get your business recommended by ChatGPT and AI search

India is now ChatGPT’s largest market, and your customers are asking it for recommendations. Here is how AI decides which businesses to name.

For about twenty years, being found meant one thing: ranking on Google. That is still mostly true, and anyone telling you it is finished is selling something. But a second front has opened, and it behaves differently enough that treating it as "SEO, but for robots" will get you nowhere.

This post is about what is actually known – from real studies rather than LinkedIn posts – about how AI assistants pick which businesses to name, and what a business in Ranchi should sensibly do about it.

Update, 12 November 2026: OpenAI now lets businesses claim a profile inside ChatGPT. Our step-by-step guide to claiming your business on ChatGPT covers the Business Hub process and what it means for Indian businesses.

Your customers are already doing this

In November 2026, OpenAI’s India managing director Prabhjeet Singh told the Economic Times that India has become ChatGPT’s largest market in the world by weekly active users, and that the Indian user base has roughly doubled since January. Back in February, OpenAI put the figure at more than 100 million weekly users in India.

Treat those as company numbers rather than audited ones – OpenAI has every reason to publish a big figure and no obligation to show its working. But the direction is not in dispute, and you can verify it without any data at all by asking the next five people who walk into your office what they used last time they needed a recommendation.

Here is why it matters commercially. Someone searching "best interior designer in Ranchi" on Google gets ten links and picks from them. Someone asking an AI the same question gets three names, each with a sentence explaining why. There is no page two, no map pack to scroll, and no second chance. You are either in the answer or you do not exist in that conversation.

What the data says about your traffic – including the part nobody quotes

The best evidence available on this comes from a field experiment by Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon. They recruited 1,065 Chrome users, randomly assigned some of them a browser extension that silently removed Google’s AI Overviews, and watched 68,089 real searches over January and February 2026.

Because the assignment was random and pre-registered, this measures cause rather than correlation – which almost nothing else published on this subject does.

The headline: on searches where an AI Overview appeared, outbound clicks to websites fell from 0.61 per search to 0.38 – a drop of roughly 38%. The share of searches ending with no click at all rose from 54% to 72%. AI Overviews appeared on about 42% of the searches in the study.

38%fewer clicks to websites when an AI Overview appears
72%of those searches end with no click at all, up from 54%
42%of all searches in the study showed an AI Overview

Source: Agarwal (Indian School of Business) and Sen (Carnegie Mellon), randomised field experiment, 68,089 searches, January-February 2026. Working paper, not peer reviewed.

That is the number everyone quotes. Here is the part that almost nobody does.

The effect was concentrated almost entirely in informational queries – the 71% of searches where someone wants to know a thing. For transactional and navigational queries, the kind where someone is looking to hire, buy, or reach a specific business, the measured effect was not statistically distinguishable from zero. The authors flag that these estimates are noisier, so this is not proof of safety. But it is a very different picture from the one in the headlines.

Clicks on ads, incidentally, did not move at all.

So the honest read for a local business is this: your blog traffic and your "how does X work" pages are genuinely exposed. Your "hire a plumber in Ranchi" traffic, so far, is much less so. That should change what you worry about, and in what order.

Two caveats worth stating. The paper is a working paper on SSRN and has not been peer reviewed. And it studied US desktop Chrome users, not Indian mobile users, who are most of this market.

How AI assistants actually pick which business to name

The largest public dataset on this comes from Yext, which analysed 155.5 million AI citations across 1,623 brands and four models – OpenAI, Gemini, Perplexity and Anthropic – in the first quarter of 2026, using questions with local intent asked the way a nearby customer would ask them.

The single most useful finding: around 80% of citations pointed at either the business’s own website or a structured listing – a directory entry with the business’s facts in it. Not clever content, not blog volume. Facts, in places machines can read them.

The models differ more than you might expect:

Share of citations pointing at the brand’s own website

Gemini52.0%
Perplexity50.9%
ChatGPT48.8%
Claude42.6%

Source: Yext Research, 155.5M AI citations across 1,623 brands, Q1 2026. Vendor research – Yext sells listings management software.

  • Gemini and Perplexity lean on your own website – 52.0% and 50.9% of their citations respectively.
  • ChatGPT leans harder on directories – 36.1% of its citations were listings, against 48.8% websites.
  • Claude pulls far more from reviews and social – 19.5%, roughly seven times ChatGPT’s share.

Worth knowing who is telling you this: Yext sells listings management software, and a study concluding that listings are 80% of the game is a study that sells listings management software. The category definitions are theirs. It is still the best public data available, and the broad pattern matches what anyone working on this sees.

Yext’s own explanation for the pattern – which they are careful to call a hypothesis rather than a finding – is that models prefer facts they can cross-check. A detail that appears identically on your site, in a directory and on a review platform is easier for a model to state confidently than one that appears in a single place.

That is the most useful idea in this entire post, so it is worth saying plainly.

There is no AI algorithm to game. There is a machine looking for facts that agree with each other.

Why most AI-search advice does not work in India

Look at the directories Yext found being cited most often: MapQuest with 6.0 million citations, TripAdvisor 3.8 million, Yelp 2.1 million, Healthgrades 2.0 million.

Now consider how much any of that matters to a business in Ranchi. Yelp has effectively no Indian presence. MapQuest is an American mapping service most Indians have never opened. Healthgrades lists American doctors.

Nearly every article you will read about getting recommended by AI was written for an American audience and tells you to go and fix your Yelp page. The pattern transfers. The list does not.

For an Indian business, the structured sources that actually carry weight are:

Most cited in the US data

  • MapQuest
  • TripAdvisor
  • Yelp
  • Healthgrades

Not one of these has meaningful coverage of Indian businesses.

What carries weight in India

  • Google Business Profile – more important than all the others combined, and free.
  • JustDial, Sulekha, IndiaMART – the broad directories with real Indian coverage.
  • Your category’s vertical – Practo, Zomato, 99acres, Urban Company, WedMeGood. One well-kept vertical listing beats ten generic ones.
  • Anywhere a human wrote about you – local news, an association page, a supplier’s partner list.

The job is not to be on all of them. The job is to make sure that wherever you do appear, the facts are identical.

Seven things that decide whether AI names your business

  1. Your name, address and phone number are identical everywhere

    Not similar – identical. "Shop 4, Main Road" in one place and "Shop No. 4, Main Rd" in another is a contradiction a machine has to resolve, and the cheapest way to resolve a contradiction is to cite somebody else instead.

  2. Your Google Business Profile is genuinely finished

    Correct primary category, every service listed, real hours including holidays, service areas, photographs that are not three years old, and the questions section answered by you rather than by strangers.

  3. You have reviews, recently, in words

    Volume matters, recency matters more, and the text matters most – a review saying "fixed my inverter the same evening" tells a model something a five-star rating with no text cannot.

  4. Your website states its facts in plain text

    Prices, service areas, timings and qualifications that live inside a JPEG, a PDF brochure or a slider that only renders after JavaScript are invisible to most of what crawls you. This is the most common technical failure we find, and it is usually a website build problem rather than a content one.

  5. Your pages answer questions in the shape they are asked

    A heading that reads "How much does an interior designer in Ranchi charge?" followed immediately by a direct answer is extractable. The same information buried in paragraph nine of a page headed "Our Philosophy" is not.

  6. Somebody other than you has mentioned you

    Self-description is weak evidence. A model weighing whether to recommend you is looking for corroboration from somewhere you do not control.

  7. Your markup says what your pages mean

    LocalBusiness, Service and FAQPage schema do not make you rank, but they remove ambiguity about what you are, where you are and what you charge for. This overlaps almost completely with ordinary technical SEO work, which is rather the point.

How to check what AI says about you today

This takes fifteen minutes and almost nobody does it.

Open ChatGPT, Gemini and Perplexity. Use a temporary or logged-out chat so your own history does not flatter the result. Ask each one the questions your customers would actually ask:

  • "Best [your category] in Ranchi"
  • "Who should I hire for [the thing you do] in Ranchi?"
  • "Is [your business name] any good?"
  • "What does [your business name] charge?"

Write down three things: whether you were named, which sources the answer cited, and – most importantly – whether anything it said about you was wrong.

That last one is the urgent category. An AI confidently telling people you close at 6pm when you close at 8pm, or that you do not offer a service you have offered for two years, is costing you money right now. And you cannot fix it by arguing with the chatbot. You fix it at the source it read, then wait.

Do this once a month. It is the only honest measurement available, because nobody can see inside these systems, and any agency claiming to track your "AI ranking" precisely is overstating what is knowable.

What has not changed

Quite a lot, actually, and it is worth saying because the hype around this is getting expensive for people.

Your Google Business Profile is still the highest-leverage thing you own, and it was already the highest-leverage thing you owned in 2023. Most of what makes you visible to AI is ordinary good SEO done properly – accurate facts, clean markup, pages that answer real questions, a site that loads. There is no separate discipline here that you need to buy as a line item.

Paid advertising still does the thing AI visibility cannot: produce enquiries on a schedule you choose. AI visibility, like SEO, is earned slowly and cannot be switched on in March because March was quiet. If you need customers this quarter, that is still a paid problem, and anyone redirecting your whole budget into "AI optimisation" is doing you harm.

FAQ

Is SEO dead because of AI?

No, and the available evidence actively says otherwise. The Agarwal and Sen experiment found the click loss concentrated in informational searches, with no statistically significant effect on transactional ones – the searches where someone is ready to hire. What has changed is that informational content now has to earn its traffic much harder, while the fundamentals that make you findable are the same fundamentals that make you citeable.

Longer than paid ads and roughly on the timescale of SEO – months, not weeks. Models are retrained periodically and the live-browsing ones depend on what is indexed about you right now. Fixing a wrong fact can propagate reasonably quickly; becoming a name a model reaches for in a competitive category does not. Anyone quoting you a fixed timeline is guessing.

Do I need a separate agency for AI search optimisation?

Almost certainly not. The work is accurate listings, a complete Google Business Profile, real reviews, a technically clean site and content that answers questions directly. That is a list any competent SEO should already be doing. If someone is quoting you a separate retainer for "GEO" or "AEO" on top of SEO, ask them precisely which tasks are in the second invoice that are not in the first.

Which matters more for an Indian business – my website or directory listings?

Both, and they are not really in competition. Yext’s data found brand websites to be the largest single citation source for Gemini and Perplexity, while ChatGPT drew more heavily on listings. Since you cannot know which assistant your customer is using, the realistic answer is to make the same facts true in both places. For an Indian business, start with Google Business Profile before any other directory.

Does posting more blog content help me get cited?

Only if the content answers questions people actually ask, in plain language, on a site that is technically readable. Volume by itself does nothing, and informational content is precisely the category the research shows losing the most clicks. A small number of genuinely useful pages, plus accurate facts everywhere your business is listed, beats forty thin posts – and costs less to produce.