Skip to content
Spilanth Studio
Guide · 7 min read

AI in advertising: what it does well in India, and where it goes wrong

Sriramsurya, Filmmaker & Creative Director

Sriramsurya

Filmmaker & Creative Director · Spilanth Studio, Chennai

Published

Short answer

AI in advertising means using AI to make ads, to decide who sees them and to learn which ones work. In India in 2026 the biggest change is in making them: a product film or a Tamil version that once needed a shoot day can now be made in days. The idea, the script and the judgement still decide whether an ad sells.

Ask ten marketers what “AI in advertising” means and you will get ten answers. One means a chatbot writing captions. One means Meta deciding who sees the ad. One means a Reel with no camera anywhere near it. They are all right, and they are very different jobs.

This guide separates them, shows two real campaigns that went in opposite directions, and ends with a way for a brand in India to start without wasting money. We make AI ads for a living, so we have a view. We will try to keep it honest.

Where does AI show up in an ad?

In three places, and you are probably already using two of them.

  1. Making the ad. Help with the script, then images, video, voice and music. This is the new part, and the part everyone is talking about.
  2. Delivering the ad. If you have run Meta's Advantage+ or Google's Performance Max, a machine has been choosing your audiences, placements and budget split for some time. That is AI in advertising too.
  3. Learning from the ad. Reading which version holds attention, which one is getting tired, and where the money is leaking.

The second and third have been around for years. The first is what changed recently, and it changes the economics for small brands most of all.

What changed in making ads

Many briefs no longer need a shoot day. A product on a table, a launch film, a story with characters: these can now be built without a crew, a location or a rain backup. A modest one-day shoot still runs into several lakh, and most of that pays for the day itself. Our cost breakdown goes line by line.

Versions became cheap. When one ad costs a shoot day, you make one and hope. When it doesn't, you can make five with different openings and let the audience tell you which idea works. That is the whole reason we sell ads in packs of five.

Regional languages stopped being an afterthought. A Tamil, Telugu or Hindi version used to mean dubbing the English one. Now each can be written in its own language from the first line, which is how it should sound. Here is what the difference sounds like.

A campaign that got it right: Cadbury and the neighbourhood shop

In 2021, Cadbury Celebrations, working with Ogilvy and Wavemaker, used machine learning to recreate Shah Rukh Khan's face and voice, with his participation, so that small neighbourhood shops could each get their own version of a Diwali ad naming their store. Thousands of local shops got an ad with a film star in it, something none of them could ever have paid for.

Notice why it worked. The technology was impressive, but the idea was generous: a big brand lending its biggest face to the small shops that sell its chocolate. People shared it because of what it did, not how it was made.

A campaign that went wrong: Coca-Cola's AI Christmas ads

Coca-Cola remade its much-loved “Holidays Are Coming” Christmas ad with generative AI in 2024, and again in 2025. Both times it was widely criticised online as soulless. The trucks and the snow were not really the problem. People objected to a brand built on warmth handing its most loved film to a machine, and the company talking about how few people it took to make it.

The lesson for a smaller brand is simple. How your audience feels about your brand decides how they feel about AI in your ads. A new D2C snack trying five fresh hooks is a very different thing from a heritage brand replacing a tradition.

Where AI ads still go wrong

Most failures fall into a few patterns, and all of them can be avoided:

  • Almost-real faces. Odd hands, dead eyes, lips that don't match the words. People spot these quickly and trust the ad less.
  • A product that drifts. The label changes between shots, the cap changes shape, the colour is slightly off. On a real product, that looks careless.
  • Pretending to be real. A “customer” who never existed, a review nobody gave, a result the product cannot deliver.
  • No idea behind it. Beautiful visuals with nothing to say. The technique becomes the message, and nobody remembers the brand.

The research on this is worth a read before you brief anyone: can people tell your ad was made with AI?

What are the rules in India?

ASCI, the Advertising Standards Council of India, has published guidelines on AI-made content in ads. Some things need a label, such as an AI presenter. Some are banned even with a label, such as fake testimonials or AI “after” photos that exaggerate results. Routine edits need nothing. We summarised them in do AI ads need a label in India?

Ownership is the other question brands ask. In short, the contract decides it, so get the rights in writing. Who owns an AI-generated ad explains what to agree first.

How to start without wasting money

  1. Pick one product and one buyer. Not the whole range, not “everyone”.
  2. Find the one feeling that sells it. Relief, craving, pride, proof or belonging. We call this Honey Badger Marketing.
  3. Make five versions with different openings and keep the product, the offer and the price the same in all five.
  4. Run them together on Meta with the same budget and audience. After enough views, compare how many people watch past the first seconds and what each result costs.
  5. Keep the winner and build the next five from it.

You don't need to buy an AI tool or hire an AI team for this. You need a clear brief and someone who can direct.

Should you make AI ads yourself or use a studio?

If you need a quick post for your own page, a do-it-yourself tool is fine and costs little. If the ad has to sell, has to show your real pack accurately and has to hold up when a customer looks closely, it is worth having a filmmaker write and direct it. That is what we do: see our services, or start with a product ad. Every ad starts with a free 30-second script, so you see the idea before you pay for anything.

Frequently asked questions

What is AI in advertising?

Using AI at any of three stages: making the ad (script help, images, video, voice), delivering it (Meta Advantage+ and Google Performance Max choose audiences and placements) and learning from it (reading which version holds attention and which one has gone stale).

Is it legal to use AI in ads in India?

Yes. ASCI's guidelines ask for a label in some cases, such as an AI presenter, and ban things like fake testimonials even with a label. Routine edits and obvious fantasy need no label.

Will AI replace ad agencies?

It replaces a lot of production work: the shoot day, the location, some of the crew. It does not replace knowing what to say, to whom, and in which order. That part is still the difference between an ad that sells and one that is scrolled past.

How much does an AI ad cost in India?

Do-it-yourself tools cost little but leave the script, accuracy and quality to you. At Spilanth Studio a filmmaker-made AI ad starts from ₹4,999 per 30-second ad, GST included, with a free script first.

Sriramsurya, Filmmaker & Creative Director

By Sriramsurya

Filmmaker & Creative Director · Spilanth Studio, Chennai

Filmmaker and co-founder of Spilanth Studio. Writes and directs every ad the studio makes, including the Nithya Stamp product ad and the Lamborghini Revuelto showcase film.