Google Ads in India: The 2026 Operator's Guide

The honest summary

India is the market where paid search is easiest to start and hardest to run well. Demand is enormous, Google's share of tracked search referrals is close to total, and media costs are low. None of that is the problem. The problem is that everything which normally slows an account down — expensive clicks, tight budgets, a small keyword universe — is absent, so structural mistakes scale before anyone notices them.

This guide covers only what is genuinely different about India: language and script, the mobile weighting, the features and categories Google does not give Indian advertisers, the absence of published cost data, and the qualification problem that follows from cheap traffic. Everything else — the auction, the bidding strategies, the match types — works the way it does anywhere, so I have left it out.

One disclosure up front. I have managed Google and Meta budgets across seven or more countries, but I do not hold years of hands-on Indian account history and will not pretend otherwise. What follows rests on Google platform facts I retrieved and read, on market data with its source named, and on cross-market experience.

The benchmark data does not exist, and that is the important part

Search for Indian Google Ads costs and you will find dozens of pages with confident industry tables, often with city-tier multipliers attached. They look like data. They are not.

When I went looking for the underlying dataset — a publisher, a sample size, a date range, a stated method — there was nothing. The most-cited of these pages sits on a domain running one programmatic template that produces a country-by-industry benchmark page for a long list of countries, and its own recoverable description of its method says the figures are sourced from Google Keyword Planner and recalculated periodically. That is a forecast generator, not a measurement. Several of those pages return a 404 when fetched directly, so what you are reading is often a search-result summary of a page that no longer resolves.

I ran exactly this investigation for another market and documented the chain in full, because the mechanism is identical and worth seeing once. The Nepal cost investigation shows the same generator, the same 404s and the same circular citations. India is the more striking case: 1.4 billion people, near-total Google search dominance, one of the largest agency sectors in the world, and still nobody has published a real dataset.

Language: eleven you can target, and the ones you cannot

Google Ads publishes a targeting language list of about fifty languages. Eleven of them are relevant to India: English, Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Punjabi, Tamil, Telugu and Urdu. That is the deepest language support of any market in this series.

The absences matter more, because they are invisible until you hit them. Odia and Assamese are not on the list, and neither are several other scheduled languages. An advertiser trying to reach Odisha or Assam in-language has no supported targeting option, and Google states that ads created in an unsupported targeting language can be disapproved. That is a delivery risk sitting inside a creative decision, which is a bad place for one to hide.

The mechanic that trips people up more often is what language targeting actually does. It filters by the language settings on the user s Google account and content, not by the language of the query typed. Setting a campaign to Hindi does not make it match Hindi keywords and does not translate anything. It narrows the audience; your keywords still do all the matching work.

The transliteration problem

A very large share of Indian commercial search is typed in Roman letters but is not English. A user thinking in Hindi or Marathi types the phrase phonetically, often mixing an English product noun into an Indian-language sentence. That query belongs to neither keyword list: not the English one, because those are not English words, and not the Hindi one, because it is not in Devanagari.

This is where broad match becomes dangerous rather than merely loose. The system finds related demand across that boundary because it is designed to, and the search terms it brings back are spelled in ways your negative keywords have never seen. A negative list built from English spellings does not block a Roman-script Hindi variant of the same unwanted query.

The working approach is unglamorous. Build transliterated variants deliberately, as their own ad groups. Read the search terms report in both scripts, and add negatives in both, for months rather than weeks. My negative keyword approach for service businesses is the starting structure. In India you run it twice, once per script, and the second pass is the one nobody budgets time for.

Mobile-first is a structural fact here, not a slogan

StatCounter's June 2026 figures put mobile above two-thirds of tracked Indian page views, with desktop under a third — the only mobile-majority split among the large markets in this series. Read those as tracked referral shares from a pageview sample rather than as a census of people, which is what StatCounter itself says they are.

What that changes is everything downstream of the click. A form that works on a laptop and merely functions on a phone quietly halves your qualified volume when most traffic arrives on a phone, on a connection you did not test for, on a device several generations older than yours.

It also closes the channel conversation. Bing sits at roughly a percent of tracked Indian search referrals and effectively nothing on mobile, so the Microsoft Ads argument that has merit in Canada does not exist here. A second source of volume in India is Meta, YouTube or Demand Gen.

The features India does not get

Feature parity with the large Western markets is high, which is the opposite of what people assume. Merchant Center, Shopping, local inventory ads, call reporting and lead form assets are all available. The real exceptions:

The Shopping asymmetry catches competent teams. Ten Indian languages beyond English are targetable in Search; the Shopping feed language list is a different, shorter list on which Hindi is the only Indian entry. A retailer with a genuine Tamil or Bengali Search operation cannot extend it into Shopping in that language. That is a boundary, not a feed-hygiene problem.

Why volume-optimized Indian accounts fail

Here is the pattern, and I have watched versions of it in more than one market where clicks are cheap. An account launches on broad match with a conversion-volume bid strategy. Clicks arrive in quantity, cost per click falls, conversions climb, and cost per conversion looks excellent against anything the client has seen. Everybody is pleased for about six weeks.

Then the sales team stops answering the leads, because most are unworkable — wrong city, wrong budget band, wrong product, students doing research, competitors, and a steady drip of people who filled the form to see what was behind it. Nothing in the ad account changed. The account metrics are still improving. That is the point.

India is structurally biased toward this failure mode because low click costs remove the natural brake. In an expensive market a marginal keyword announces itself within days by burning budget. In India it runs for a quarter inside the noise. The account is not misconfigured in any way an audit checklist would catch; it is optimizing hard toward a conversion definition that stopped meaning anything.

The metric that lies

The lying metric in Indian accounts is cost per conversion. The second number that catches it is the ratio of conversions to sales-accepted leads, tracked weekly.

Cost per conversion falls for two different reasons and the report does not distinguish them. It falls when targeting and creative improve. It also falls when the system finds a cheaper, lower-intent slice of the auction and pours volume into it — which looks identical on the chart. The only way to tell them apart is to carry a downstream number back into the account.

So define a qualified lead with the sales team before launch, in writing, and get that status back into Google Ads as an offline conversion or at minimum into a weekly reconciliation. Watch the ratio, not the cost. If cost per conversion drops while the qualified share drops faster, the account is getting worse and every dashboard says it is getting better. The conversion tracking audit covers the plumbing — deduplication, counting settings, and the difference between a fired tag and a real event. Get that right first, because a qualification layer built on broken counting produces confident nonsense.

What Google will not let you advertise to India

This is Google's own advertising policy layer, not law, and it decides whether your ads run at all. Checked against Google's policy pages in July 2026:

• Real-money gaming is effectively closed. India has no entry in the country-specific gambling table, so there is no certification pathway. Google separately disallowed Rummy and Daily Fantasy Sports promotions targeting India from January 2026, citing local legal requirements. Social casino games with no real-money prizes remain allowed with certification.
• Cryptocurrency and related products cannot be advertised to India at all. India is absent from the approved-locations list and there is no certification route, whatever the agency content implying otherwise says.
• Financial services advertisers must be verified before serving to Indian users, a requirement in place since January 2023, showing they are licensed by the relevant Indian regulator or exempt. It catches advertisers who deliberately target users seeking financial services, not only those selling them.
• Alcohol is nearly closed. Sale advertising is permitted only for zero-alcohol products, and India is absent from the approved list for brand or informational alcohol advertising.
• Sexual content is blocked entirely for India-targeted serving.

If your category is on that list, no amount of account structure fixes it. Find that out before you build the landing pages, not after.

Privacy and consent, operationally only

India has a comprehensive personal data protection statute with rules made under it, administered by a national data protection authority, and the obligations are phased rather than all live today. A lot of commentary in circulation describes the regime as though every part of it already binds.

What it changes for an ad account, and only that: consent state becomes something your tags have to know about, so the tagging layer must hold and honor a signal rather than fire unconditionally. Audience size becomes a downstream consequence of consent design rather than a separate targeting decision. And a lead-generation funnel needs consent captured at the form, stored with the record, and visible to whoever calls the person.

One point worth flagging without my interpreting it: the statute defines a child by a higher age than the thresholds most ad platform policies are built around, and restricts advertising directed at children. Given the age profile of the Indian population, that is an audience question rather than a cookie-banner question, and one to put in front of a lawyer early.

And one correction, because it is sold hard. Google has no India-specific consent requirement. Its EU user consent policy covers the European Economic Area, the UK and Switzerland, and its ads data processing terms do not name India. Anyone telling you Google requires Consent Mode v2 on Indian sites is selling you something. Build consent architecture because your lawyer told you to, not because a vendor said Google demands it.

The prerequisite nobody wants to hear

Before any of this matters, someone has to define what a good lead is, and the person who defines it has to be the person who works it — the salesperson or the front desk who picks up the phone, not the founder describing an ideal customer. In a market with no benchmarks that definition is the only reference point you will ever have. Without it you are optimizing against form fills, and form fills in India are cheap enough to buy in effectively unlimited quantity.

Across the accounts I have run in other markets, a lead-to-customer close rate of around twelve percent of generated leads has been typical, lower for consultancies — a cross-market figure from my own book, not anything anyone should expect in India. It is worth mentioning only because it is a downstream number, and most agencies never see one. If your reporting stops at the conversion, no benchmark table is going to tell you whether the account works.

Honest scope: when Indian paid search is the wrong tool

Three situations where I would say no on a call.

• When nobody is searching for the category. India rewards demand capture and punishes demand creation on Search. If your product needs explaining before anyone would look for it, a Search budget buys a lesson rather than a pipeline.
• When sales capacity cannot absorb the volume. Cheap clicks move the constraint to the phone immediately. Launching a good campaign into a team that cannot call back within the hour wastes the only advantage the market gives you.
• When margin cannot survive an acquisition cost you have never measured. With no published benchmarks you find out what a customer costs by spending until you can observe it.

And a sequencing point: if measurement is not trustworthy, fix it before touching bidding. Most of what circulates about Quality Score is wrong anyway — I took it apart in four Quality Score myths, and none of it is India-specific because Quality Score is not a country-specific mechanic.

What this post is, and what to check yourself

The platform claims come from Google-published country lists and policy pages I retrieved and read: the Merchant Center countries and languages tables, the call reporting eligibility list and its India footnotes, the language targeting tables, and the Local Services country selector. Market shares come from StatCounter, June 2026. What I do not have is Indian account performance history, which is why there is not one cost figure in this post.

If you want the method rather than the country specifics, the Signal-to-Revenue Framework sets out how I sequence measurement, structure and bidding in any market, and everything above is best read as exceptions applied on top of it. I run that practice through Arcetis, and this guide leads with qualification rather than cost because in India cost is the thing that will not be your problem.

Related reading