AI shifts ad targeting from tracking clicks to inferring intent, budget, and personal circumstances from conversational data.
ChatGPT and Meta AI now feed chat interactions into ad systems, raising questions about consent, retention, and advertiser access.
India's DPDP Act sets consent rules for personal data, but its core obligations take effect only in 2027, after AI ads have already launched locally.
Every click has always told advertisers something. A search for shoes or a scroll past a travel site gave away interest, nothing more. AI changes what a single sentence can reveal. A chat about a tight budget, a health worry, or a new baby carries far more than any click ever did, and platforms are starting to put that data to work.
Most ad systems rest on one principle: the more a platform knows, the sharper its guess about what a person will buy. Cookies, small files left behind in a browser, work like sticky notes passed between sites. One page notes a visit to a running shoe store; another reads the note and shows a matching ad.
Browsers treat this differently. Safari and Firefox limit cross-site tracking by default. Chrome still allows third-party cookies, offers controls to manage them, and blocks them automatically in Incognito mode. Cookies, logged-in accounts, and device details all feed the same system. AI now adds a sharper layer on top.
AI spots patterns across small, unrelated actions. Someone searching for baby products and reading about infant sleep might get flagged as a new parent, even without saying so directly. The guess could be wrong, yet it still shapes what that person sees next.
Three steps matter here. Collection happens when a platform receives information: a search, a click, a chat. Inference happens when AI estimates interests or plans from that data. Activation happens when an ad system uses that estimate to choose what to show.
Not every inference reaches an advertiser; that depends on the platform's own policy. The real question is which guesses get stored and acted on, often without the person ever knowing one was made.
A search is a few words. A chat often lays out an entire situation: worries, plans, budgets, health questions.
A University of Washington study found that 82% of respondents rated chatbot conversations as sensitive or highly sensitive, ranking above email and social posts. That reflects perception rather than proven misuse, but it shows how much weight people place on these exchanges.
Meta has started drawing on this space directly. Since December 16, 2025, interactions with Meta AI have helped shape content and ads across Facebook and Instagram, rolled out in most regions aside from the UK, the EU, and South Korea.
Meta states that topics such as religion, health, politics, sexual orientation, and racial or ethnic origin stay out of ad targeting. Users cannot opt out of having their AI chats factored in, though ad personalization settings remain adjustable. That is a company statement about intended use, not a technical limit on what the system could infer.
ChatGPT ads went live for Indian users in late August 2026, shown to logged-in adults on Free and Go tiers as labeled sponsored placements beneath responses. OpenAI states that ads do not influence ChatGPT's answers, that advertisers never receive private conversations, and that ads will not appear near health, mental health, or political topics. The feature now runs across more than 40 countries.
Three distinct questions apply whenever a platform makes such claims. Does it use a conversation to select an ad? Does the advertiser receive that conversation? Does the platform retain it for other purposes? An assurance on one point says nothing about the other two, and policies vary widely between companies.
An ad for an affordable laptop, shown to someone actually shopping for one, helps. An ad that senses financial stress and nudges toward an unplanned purchase does something else entirely. The underlying tools can be identical; what shifts is how personal the signal is, how clearly it is disclosed, and whether a person can control it.
| Factor | Relevance-Based Ad | Influence-Based Ad |
|---|---|---|
| Signal used | Stated interest or search | Inferred emotional or financial state |
| Transparency | Usually visible to the user | Rarely disclosed |
| User control | Can be adjusted in settings | Often not controllable |
| Risk | Low to moderate | Higher, especially for vulnerable users |
Several methods reduce certain risks without removing them entirely.
| Method | How it works | Limit |
|---|---|---|
| On-device learning | Analysis happens on the device; less data leaves it. | Some personal data may still be transmitted. |
| Group-based ads | Advertisers see a group, not an individual. | Small groups can still point to one person. |
| Contextual ads | Matches the page's content, not browsing history | Offers no personalization |
| Data minimization | Collection limited to a defined purpose | Depends on how narrowly "purpose" is set |
India's Digital Personal Data Protection Act, 2023, governs how digital personal data gets processed, allowing it based on consent or specified legitimate uses, with consent tied to a stated purpose.
The DPDP Rules were notified in November 2025 on a phased schedule. The Data Protection Board came first, a consent manager framework follows in November 2026, and the main obligations and penalties apply from May 2027. ChatGPT ads reached India in August 2026, ahead of those core obligations.
The law does not automatically rule out inference-based advertising either; whether a specific practice complies depends on which provision applies and how the data moves through the system.
The practical test is whether people actually understand how their information gets used. Agreeing to use a chatbot is not the same as agreeing to every later use of what gets shared inside it.
Also Read: Google September Android Drop: New AI Features Added to Phones
A few habits limit exposure without eliminating it. Keeping full names, addresses, bank details, and health information out of chatbot conversations helps. Checking memory, chat deletion, and ad personalization settings separately, rather than assuming one covers the others, closes a common gap.
Limiting app permissions (a torch app rarely needs location data), using a tracker blocker, and reviewing the interest profile a platform has built all add friction against casual data collection.
AI will not end personalized advertising, nor should it. What it demands is a higher bar: if a platform can predict what someone needs, it should be able to explain what it predicted, on what basis, and how to say no. Prediction without that explanation is where trust starts to break down.
Also Read: Why Every App is Adding AI: How It Changes What You Do Online
The next phase of ad privacy will not be decided by better cookie controls. It will be decided by whether platforms treat an inference with the same weight as a fact and whether users ever get to see the guess being made about them before it shapes what they encounter next.
You May Also Like:
1. Does OpenAI use ChatGPT conversations to target ads in India?
OpenAI states it does not use chat content to influence responses and that advertisers do not receive private conversations, though ads do appear as labeled placements based on account-level signals for Free and Go tier users.
2. Can users opt out of Meta AI chats shaping their ads?
No. Users cannot opt out of having Meta AI interactions factored into ad and content personalisation, though they can still adjust how personalised those ads appear through settings.
3. Does deleting a chat also delete related ad data?
Not necessarily. Chat deletion, AI memory, and ad personalisation are typically separate settings, so removing a conversation does not guarantee any advertising record tied to it is also erased.
4. Is AI-inferred ad targeting legal under India's DPDP Act?
The Act requires consent tied to a stated purpose, but its core obligations and penalties only take effect from May 2027, after ChatGPT ads already launched in India in August 2026.
5. What is the real difference between relevant ads and invasive ones?
The tools are often identical; what changes is how personal the underlying signal is, whether it's disclosed to the user, and whether the person has any real ability to control or reject it.