AI content for business: what's safe and what to avoid
Where AI content is safe for an Indian business, when a post or ad needs an AI label under the 2026 IT Rules and ASCI's new guidelines, and what stays off limits even with a label.

We do this for clients every week.
Using AI content for business is safe in India when a person checks every fact before it goes out, and when nothing realistic is passed off as real. Captions, ad copy, product descriptions, translations and routine photo clean-ups need no label, while realistic AI people, voices, places and unbuilt products often do. Fake endorsements, made-up results and look-alikes of real people are off limits even with a label. This guide sets out where those lines sit under the 2026 IT Rules, ASCI's new advertising guidelines and the platforms' own policies.
What counts as "AI content" under the new rules?
Two sets of rules now matter. The IT Rules amendment that came into force on 20 February 2026 created a category called "synthetically generated information": audio, images or video made or altered by a computer so that they appear real and could be taken for a real person or a real event. MeitY's FAQ on the amendment says it covers only audio, images and video, so text on its own falls outside it, as do routine edits such as brightness, noise reduction and compression, and good-faith presentations or training material.
The second is the Advertising Standards Council of India. On 29 September 2026 it published guidelines for labelling synthetically generated content in advertising, which take effect three months after publication. Its definition is almost the same: realistic audio, images or video, judged by their effect on the customer rather than by the tool that made them.
So the useful question is never "did we use AI?" It is "could a customer mistake this for something real, and would that change what they buy?"
Where AI is generally safe to use
Most day-to-day uses sit on the safe side, as long as someone who knows the business reads the result before it is published.
- Writing and editing. Caption drafts, ad copy, emails, FAQs and product descriptions. ASCI lists "generating or enhancing advertising copy" among uses that need no label. A Borivali sweet shop can draft Diwali captions with AI, but someone still checks that the box weights and the "pure ghee" line are true.
- Translation and subtitles. Marathi, Hindi or Gujarati versions of a post, and captions on a reel. Both ASCI and MeitY exclude accessibility uses, provided the meaning isn't changed.
- Routine photo and audio fixes. Colour correction, lighting, noise removal and minor blemish removal on real photos of your real product.
- Backgrounds and obvious fantasy. An abstract skyline behind a festive offer, or a cartoon mascot flying over the city. ASCI treats purely decorative and plainly unrealistic elements as label-free.
- Planning work. Content ideas, outlines, shot lists and first drafts of briefs.
The catch is accuracy. Google's guidance reminds publishers that generative models predict likely words rather than look up facts, so prices, opening hours, GST details and product claims need checking against your own records every time.
When does AI content need a label?
A label is needed when realistic synthetic content shapes a buying decision and leaving it unlabelled would mislead. ASCI, YouTube and Meta draw the line in similar places, though not identically.
| Use | Label needed? | Whose rule |
|---|---|---|
| AI-written caption, blog or ad copy | No | ASCI; YouTube lists scripts as production help |
| Colour fix or cleaned-up background on a real photo | No | ASCI, MeitY |
| AI model or brand ambassador who doesn't exist | Yes | ASCI |
| Realistic walk-through of a tower not yet built | Yes | ASCI's own example |
| Real person's face or voice recreated, with consent | Yes | ASCI |
| Realistic AI voiceover or video in a reel | Yes | Meta, YouTube |
How to label in practice
Use the platform's own setting first. Meta says it requires people to use its disclosure tool for organic posts with photorealistic video or realistic-sounding audio that was digitally created or altered, and may apply penalties if they don't. YouTube asks the same question at upload, under "AI use" in YouTube Studio. Under the IT Rules, large platforms may also ask you to declare "AI-generated: Yes/No" before posting and check your answer, so answer it truthfully.
Then add a plain caption line, because platform labels vary. ASCI suggests wording such as "Video created using AI" or "Video enhanced using AI", and allows other wording that tells customers accurately what was synthetic. "Background created with AI; the product is real" is clearer than a hashtag. Keep the label and metadata your AI tool adds, too: the IT Rules require AI tools they cover to embed them where technically feasible and not let users strip them.
Real estate needs extra care. A Thane developer's AI fly-through of an unbuilt project needs an AI label under ASCI, on top of the MahaRERA details every project ad already carries, which our page on real estate website design explains.
What should a business never do, even with a label?
ASCI is clear that a label doesn't rescue a misleading ad. These uses stay prohibited whatever disclosure you add:
- Fake endorsements. An AI "customer" praising your gym, or a real person shown recommending something they never used.
- Exaggerated results. ASCI's example is an AI before-and-after body for a weight-loss product. A Bandra skin clinic showing an AI-perfected "after" face falls in the same trap.
- Settings presented as real when they aren't. A Lonavala homestay advertising an AI infinity pool it doesn't have.
- Copyrighted work used without permission. Prompting for a famous illustrator's style or a film character to sell your product.
- Deepfakes and look-alikes. Any real person's face or voice without consent, celebrities included.
The last point carries legal weight beyond advertising codes. The Delhi High Court has granted a run of personality-rights orders, including for Karan Johar in September 2025 and R. Madhavan in December 2025, restraining people from using a star's name, face or voice through AI for commercial gain. The IT Rules also treat AI-made fake documents as unlawful, so an AI-generated certificate, award or press clipping on your website is a serious problem, not a design shortcut.
AI text on your website: what Google expects
Google's guidance on generative AI content doesn't ban AI writing. It says AI can help with research and with structuring original content, but that using it to generate many pages without adding value may break the spam policy on scaled content abuse. It also asks publishers to fact-check everything, including page titles, meta descriptions and image alt text, because those appear in search results.
In practice, one useful page written from your real experience beats twenty AI pages that repeat what is already online. Our guide to SEO for small businesses in India lists "dozens of AI-written blog posts" among the tactics to skip. If an SEO company is selling you an AI-content package, the questions in how to choose an SEO company in Mumbai apply.
Online stores have one extra rule. Google's page says Merchant Center requires AI-generated product images to carry IPTC "TrainedAlgorithmicMedia" metadata, and AI-written product titles and descriptions to be labelled as AI-generated in the feed.
Who owns AI content, and what about customer data?
The Copyright Act says that for a computer-generated work, the author is "the person who causes the work to be created". How that applies to modern AI output is still being argued in India and abroad. Read each tool's terms on commercial use, keep your prompts and source files, and for a logo or trade mark use original human-made design and ask a lawyer. Our guide to who owns your website covers getting copyright assigned in writing when an agency makes the work.
Customer data needs the same care. Under the Digital Personal Data Protection Act, your business stays responsible for personal data even when a vendor or tool processes it for you. Don't paste customer lists, phone numbers or order histories into a tool whose terms let it keep or train on what you type. Use business accounts with clear data settings, or remove names and numbers first.
A sign-off checklist before anything goes live
ASCI says responsibility stays with the advertiser, so use one routine for every AI-assisted post, ad or page:
- Check every fact, price, offer and claim against your own records.
- Ask whether a customer could mistake any image, voice or video for something real.
- If yes, and it affects what they buy, add the platform's AI label and a caption line.
- Confirm that every real person shown or heard has agreed in writing.
- Remove anything showing results, settings or features the product doesn't have.
- Make sure no artist's style, character or competitor's campaign was copied.
- Keep the prompts, files and approvals together in case anyone asks later.
FAQs
Do I have to say my captions or blog posts were written with AI?
No rule in India requires a label on AI-written text. The IT Rules cover realistic audio and visuals, and ASCI lists generating ad copy among uses that need no disclosure. You are still responsible for every claim in the text, so check facts before posting. Google suggests explaining how content was made where automation plays a large part.
Can we make an AI avatar of our founder for ads?
Yes, with the founder's written consent and their own reference photos or voice samples. ASCI requires a label when a real person's face or voice is replicated for personalised messages, even with consent. Keep the avatar saying only what the founder has approved, and never use it for reviews or results the business can't back up.
If an agency makes our AI ad, who is responsible?
You are, as the advertiser. ASCI's chief executive said when releasing the guidelines that responsibility remains with advertisers to keep the communication honest and compliant. Put the labelling and consent rules in the brief, ask the agency which parts were AI-generated, and approve the final version yourself rather than letting it be posted automatically.
Do these rules apply to a small business posting on Instagram?
Yes. ASCI handles complaints about ads in every medium, including websites and social posts, whatever the size of the business. Meta's and YouTube's disclosure rules apply to every account. A removed post or a restricted account costs far more than adding a label.
We make AI images, captions and brand films with a human approval step before anything is published, and our AI content service page explains how we handle likeness, copyright and labelling. If you are planning AI content and want a second opinion on what needs a label, send us a short note and we will tell you honestly whether AI suits the job.
The weekly email
One useful email a week
Our best reads on websites, search and growing your business online, every Monday. No spam, unsubscribe in one click.