AI marketing in India: the tools are free, the rules aren't

Short answer: For an Indian small business, AI marketing means using free or nearly free tools to draft posts, answer WhatsApp enquiries and translate copy into the languages customers actually use. The tool cost is no longer the obstacle. What catches owners out is everything around the output: platforms now have to label AI-made content under rules in force from 20 February 2026, ASCI requires a virtual influencer to say it is not human and has drafted wider AI labelling guidance, and pasting a customer’s phone number into a consumer AI plan is processing personal data under the DPDP Act. Start with one job, keep a human reading every regional-language draft, and never publish a claim the tool made up that you cannot prove.

Picture the owner of a small coaching institute in Lucknow, opening ChatGPT for the first time on a Sunday evening because a rival institute two streets over has been running slick Instagram reels for months and this one has none. The owner types a prompt, gets back a festival offer, a set of captions and a WhatsApp reply template in under a minute, all for free. It feels like the hard part is over.

It isn’t. The tool has no idea whether that offer copy needs a disclosure, whether the phone numbers being pasted in for a “personalise this” prompt need consent on file, or whether the confident claim it just wrote about exam results would survive a complaint. This guide is general information, not legal advice, and a specific claim or a regulated category is worth a call to a chartered accountant or a lawyer who knows your state and sector. If you want the plain definition first, start with what AI marketing is for an Indian small business.

Why the cost argument for AI marketing doesn’t hold up in India

Every global AI marketing guide leads with price, because in most countries price is still the barrier. In India it mostly isn’t. OpenAI made ChatGPT Go free for users in India for 12 months from 4 November 2025, a plan that previously cost roughly ₹399 a month, and several telecom operators have bundled paid AI plans into subscriber packages on top of that. Check what your current plan actually includes, since offers like this change without much notice.

What this means in practice is that the constraint has moved. It used to be “can we afford the tool.” It is now “who reviews what the tool produces before it goes out.” A shop that spends nothing on tools but nothing on review time either is not saving money, it is postponing a cost. Free consumer AI plans are also usually built for one person typing prompts, not a team with shared records of what was sent to which customer. For which plan to keep once a free year ends, see our guide to the best AI marketing tools in India.

Two sets of new rules now sit over Indian AI content

Two pieces of the picture changed between November 2025 and February 2026, and neither is optional reading if you post anything AI-made. MeitY released the India AI Governance Guidelines on 5 November 2025. They are advisory, meaning they set no new penalties and are not a binding law, but they are the clearest signal yet of where central government thinking is headed, and they explicitly leave sector rules to sector regulators: RBI for lending and payments content, SEBI for anything resembling investment advice, IRDAI for insurance. If you market in one of those sectors, read the relevant regulator’s own guidance, not just this article.

Separately, the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2026 were notified on 10 February 2026 and came into force on 20 February 2026. They bring what the rules call “synthetically generated information”, AI-made images, audio, video and text, into platform due diligence obligations for the first time. Neither set of rules is aimed mainly at a small business posting an AI-made ad. The obligation to check and label sits with the platform. The exposure for you is smaller but still real, covered next.

When a platform can pull your AI-made post down, and how fast

The IT Amendment Rules 2026 require platforms to label synthetically generated content clearly and prominently, and they allow unlawful content to be taken down quickly, in some cases within as little as three hours of a valid notice. That duty is on Instagram, YouTube and similar platforms, not on you directly. Where it reaches you is upstream: platforms are increasingly building in a step where you, the uploader, are asked to declare whether a post or an ad used AI generation.

An unlabelled AI-made ad, an AI voice-over on a reel, or a video with an AI-generated presenter is at higher risk of being flagged, labelled after the fact in a way you didn’t choose, or removed while you argue about it. Check what each platform you post to currently asks you to declare, since this is one of the areas moving fastest, and answer honestly rather than skip the question. It is a smaller task than reading a new law, and it is the one most businesses actually skip.

Disclosure rules for a synthetic voice, face or influencer

ASCI, the Advertising Standards Council of India, is a self-regulatory body, not a court or a regulator with statutory penalties, but its code is the industry’s working standard and most platforms and agencies follow it. ASCI’s existing influencer guidelines already require a virtual or AI-generated influencer to disclose that it is not a real person, in addition to the usual paid-partnership label a human influencer uses.

ASCI has gone further with draft guidelines specifically for labelling AI-generated advertising content, structured by risk: a synthetic product demonstration, an ad using someone’s AI-cloned likeness or voice, and similar categories each carry their own disclosure expectation. These are draft guidelines at the time of writing, so check whether they have been finalised before treating any specific clause as settled. Build the habit regardless: if an ad uses an AI voice, an AI-generated presenter, or a synthetic demonstration of your product, say so, plainly, near the content itself. Our guide to AI ad creative covers where AI earns its place in the ad itself.

Cloning a voice or a face is no longer a grey area

Some businesses assume “AI voice” and “impersonation” are different problems. Indian courts increasingly treat them as the same one. In July 2024 the Bombay High Court granted singer Arijit Singh an interim injunction against platforms and tools using AI to clone his voice and likeness without permission, a personality-rights case, not a copyright one. The Delhi High Court has made similar orders protecting other public figures since.

None of this touches a business using a hired voice artist or a generic AI voice. It matters the moment an ad, a reel or a WhatsApp voice note uses a voice or a face resembling a real, identifiable person, celebrity or otherwise, without their consent. A “celebrity style” voice preset is a prompt to check permissions, not a shortcut.

The ANI v OpenAI case and what it means for your drafts

Ownership of AI-written marketing copy is not settled in India, and pretending otherwise is the mistake. The Copyright Act 1957, at section 2(d)(vi), treats the person who causes a computer-generated work to be created as its author. What that means for a prompt typed into ChatGPT or Gemini has not been tested by an Indian court. A related and closely watched case, ANI v OpenAI, concerning AI models trained on news content, is currently before the Delhi High Court, and its outcome could shape how courts read authorship and fair use for AI-assisted content more broadly.

Until that settles, three habits protect you either way. Keep your own drafts, edits and the specific instructions you gave a tool, since your human contribution is what is most likely to be protected. Don’t claim an AI-written tagline or design is exclusively yours to license or sell if you haven’t checked. And don’t feed a competitor’s copy into a tool asking it to “rewrite this”, since that sits closer to a copyright problem than a shortcut.

What DPDP means the moment you paste in a customer’s number

A WhatsApp number, an order history, a customer’s name typed into a prompt to “personalise this message” is processing personal data, and the DPDP Act 2023 covers that regardless of which tool you used. Most of the Act’s substantive obligations commence on 13 May 2027, with consent manager rules following from 13 November 2026, so the framework is landing in stages rather than all at once, and the direction is clear enough to plan around now. Penalties for failing to put reasonable security safeguards around personal data can run up to ₹250 crore for serious failures, aimed at the business handling the data, not the AI vendor.

The practical rule for AI marketing specifically: don’t paste a customer list into a consumer AI chat window to “find patterns” or “write follow-ups” unless you know where that data goes afterwards, and most free consumer plans don’t publish a clear answer. For what a proper opt-in should say and how consent should be recorded before you build anything WhatsApp or SMS related around customer data, see marketing automation in India, which covers that ground in depth. This article only covers where AI content intersects with it.

Hindi, Hinglish and the scripts a fluent-looking reply gets wrong

AI-written Hindi, Tamil, Telugu, Kannada, Malayalam, Bengali or Marathi marketing copy often reads fluent at a glance and wrong in register up close: too formal for a WhatsApp reply, translated word for word from an English draft, or using a word a customer would never actually say.

A workable review routine costs almost nothing. Have someone who actually speaks the language of the city you serve, not just reads it, say the draft aloud before it goes out. Keep a short glossary of your own product and offer terms, since a generic AI draft will not know what you call your own thing. Pull two or three real enquiries from your WhatsApp history each month and compare their phrasing to what the AI produces, because that gap is the tell. Where it matters, test one ad in two versions, standard script and Roman-script Hindi or the equivalent, since customers often reply in one but read comfortably in the other. Bhashini, the central government’s Indian-language AI and translation platform, is worth testing alongside consumer tools, judged the same way, on what your own customers actually say back.

The GST question nobody budgets for

A GST-registered business buying a foreign-billed AI tool, ChatGPT, Canva, Midjourney and similar, generally has to self-account for 18 percent IGST under the reverse charge mechanism, because a registered business importing a service accounts for the tax itself. That IGST is usually claimable back as input tax credit, so the net cost to a registered business is close to zero once the paperwork is done correctly. An unregistered business, or one on the composition scheme, typically pays whatever GST the foreign supplier charges as a real cost, plus the usual forex mark-up most cards apply to a dollar-billed subscription. Confirm the current mechanics with your chartered accountant before assuming either applies to you.

Tier 1 competition, tier 2 opportunity

A Mumbai, Delhi NCR or Bengaluru business is not choosing whether to use AI in marketing, competitors already are, so AI-sped content there is closer to table stakes than an advantage. The same Google Ads keyword commonly costs two to three times more in a metro than in a smaller city, with national clicks typically running roughly ₹25 to ₹60 depending on category, which makes a slow or off-tone AI draft more expensive to absorb in a metro. See digital advertising in India for how that plays out across platforms.

A tier 2 business, an Indore, Coimbatore, Jaipur or Nagpur shop or clinic, tends to gain more from a narrower use of AI: handling the volume of Hindi, Marathi or regional-language WhatsApp replies that a single owner can’t keep up with alone, rather than competing on ad polish. That difference in where the value sits is worth deciding on before choosing what to build first, whether you set it up yourself or have it built for you.

Inside a Nagpur mithai shop’s Diwali campaign, mistakes included

A family-run mithai shop in Nagpur, hypothetical, sells sweets and snacks from one shop and takes phone and WhatsApp orders during festival season. Heading into Diwali, the owner sets up two AI tools: one for drafting Hindi and English social posts and WhatsApp broadcast copy, one for translating and checking tone on the Hindi drafts, plus the free ChatGPT Go plan for quick one-off captions. The two paid tools run ₹1,800 a month each, ₹3,600 total, plus 18 percent GST reverse-charged and reclaimed as input tax credit since the shop is GST-registered, so the real out-of-pocket cost stays close to ₹3,600 a month. A local freelancer sets up a simple WhatsApp auto-reply for order confirmations as a one-off build for ₹1,20,000, inside the typical ₹1,00,000 to ₹3,00,000 range for that kind of setup, paid once rather than monthly.

The mistake came in week one. An AI-drafted Diwali post described the shop’s ghee sweets as made with “pure desi ghee, guaranteed no adulteration,” a specific quality claim the shop had never tested or certified and could not actually stand behind if challenged. Nobody had asked the tool to make that claim, it generated the phrase unprompted because it sounded like typical festival sweet-shop marketing. The owner’s cousin, reading the draft before it posted, caught it and changed the line to “made fresh daily, ask us about our ingredients,” true, checkable and still warm. Nothing was published with the unsupported claim, but it would have taken one screenshot from a competitor or a consumer complaint to turn a festival post into a genuine problem. The fix cost five minutes. Finding it cost a habit: nobody posts an AI draft without one other person reading it first.

Forty-five days, in three stages

Days 1 to 15: fix the foundations. Confirm the Google Business Profile is claimed and current, check whether existing WhatsApp or SMS contacts have real consent on file rather than assumed consent, and pick the single AI job worth automating first, usually WhatsApp or Instagram DM replies, since that is where Indian customers already are. Don’t touch paid ads yet.

Days 16 to 30: run one job, reviewed daily. Turn on AI-drafted replies or posts for that one job only, and read every single output for the first two weeks, not spot-checks. This is where a wrong claim, a mistranslated regional post or a tone problem shows up, and it is far cheaper to catch here than after it has been running unreviewed for a month.

Days 31 to 45: add the second job, or stop. If the first job is stable and someone has a clear habit of reviewing it, add drafting or translation as job two. If review is still catching real mistakes weekly, fix the first job’s review process instead. See AI lead generation in India for what a WhatsApp-first setup looks like once you’re ready to build further.

Health, education, finance and property carry an extra check

Some categories cannot rely on AI-written copy the same way a sweet shop or a salon can, because the claims involved carry more weight and more scrutiny. A clinic or wellness business making a health claim, a coaching institute quoting a result or a success rate, a finance or investment business describing returns, and a real estate business describing a property all sit inside ASCI’s more closely watched categories, and several sit under a sector regulator too: RBI, SEBI or IRDAI territory for anything resembling financial advice, medical council rules for health claims. AI tools invent plausible-sounding specifics readily, a percentage, a guarantee, a result, and none of that checking happens automatically. In these categories, every AI-drafted claim needs a named person checking it against something real before it goes out, not just a read-through for tone.

What to fix before AI marketing becomes your priority

AI marketing is the wrong priority right now for a business missing more basic things. No claimed and current Google Business Profile means AI-polished posts reach people who then can’t find you or confirm you’re real. No consented contact list means there is nothing to automate a reply into, only a cold list to avoid touching with anything automated. And if nobody in the business has the time or the language skill to read a regional-language AI draft before it posts, adding AI adds risk faster than it adds reach. Fix those three things first, none of them cost anything close to an AI subscription, and all three make the eventual AI work land better. For where digital budget goes first in India, see what digital marketing costs in India and SEO, AEO and GEO in India. For the numbers behind several of the claims in this guide, see AI marketing statistics for India.

Frequently asked questions

Is there a dedicated AI law in India yet, in 2026? No. India does not have a standalone AI law. MeitY released the India AI Governance Guidelines on 5 November 2025, but they are advisory, not binding, and set no new penalties. Sector regulators such as RBI, SEBI and IRDAI handle AI rules within their own sectors, so a finance, insurance or investment business carries an extra layer that a sweet shop or salon does not.

Do I have to label an AI-made ad or video? The duty sits with platforms, not directly with you. The IT Amendment Rules 2026, in force from 20 February 2026, bring AI-made images, audio, video and text into platform due diligence, and platforms must label this content clearly and can remove unlawful content within hours. In practice, platforms increasingly ask uploaders to declare AI content, so check what each one currently asks before you post.

Does an AI-generated influencer or avatar need to say it isn’t human? Yes. ASCI’s existing influencer guidelines require a virtual or AI influencer to disclose that it is not a real person, on top of the usual paid-partnership disclosure. ASCI has also released draft guidelines for labelling AI-generated advertising content more broadly, covering things like synthetic product demonstrations and AI likeness or voice. Check whether that guidance has moved from draft to final before you rely on it.

Can I use an AI voice or face that sounds like a celebrity in an ad? No, and Indian courts are already enforcing this. In July 2024 the Bombay High Court granted singer Arijit Singh an interim injunction against platforms using AI to clone his voice and likeness, and the Delhi High Court has made similar orders for other public figures. Treat any AI voice or face that resembles a real, identifiable person as off limits without their consent.

Who owns marketing copy that an AI tool wrote for me? It is unsettled. The Copyright Act 1957 treats the person who causes a computer-generated work to be made as its author, but how that applies to a generative AI prompt has not been tested, and a case over AI training on news content, ANI v OpenAI, is before the Delhi High Court. Keep your own drafts and edits, and do not claim exclusivity you may not actually have.

Is it safe to paste a customer’s name and phone number into ChatGPT? Treat it as processing personal data, because it is. Most of the DPDP Act 2023’s substantive obligations commence on 13 May 2027, with consent manager rules from 13 November 2026, and penalties for failing security safeguards can run up to ₹250 crore. See marketing automation in India for how consent works in practice before you build anything around customer data.

Is AI good enough to write in Hindi or a regional language without a human check? Not yet, especially for Tamil, Telugu, Kannada, Malayalam, Bengali or Marathi, where AI output often reads fluent but stiff or wrong in register. Have a native speaker from the city you serve read it aloud before it goes out, keep a glossary of your own product terms, and test one ad in two scripts, standard script and Roman script, to see which one your customers actually respond to.

How much does AI marketing cost a small business in India each month? Tools alone are close to free: many plans are free or roughly ₹1,500 to ₹5,000 a month each, and ChatGPT Go has been free in India for 12 months from 4 November 2025, so check what your plan currently includes. An agency-run, AI-assisted programme typically runs ₹25,000 to ₹80,000 a month in fees plus media, more in competitive metro categories.

Want to know which AI job would pay off first for your business, and which of these rules actually applies to you? Get a free AI marketing audit. No jargon, no pressure.