Pennsylvania’s new advertising law makes synthetic persuasion more visible. The larger challenge is keeping promises about real products worth believing.

On October 22, Pennsylvania approved a law on AI-generated advertising. It requires disclosure for certain synthetic advertisements that create false impressions about goods or services. It has exceptions and takes effect after 60 days. This is a meaningful intervention, but it establishes neither that labels will prevent deception nor that a labelled advertisement is truthful. The task ahead is to make product promises verifiable, even as convincing appearances become easier to manufacture.

Imagine a small restaurant advertising a sumptuous meal assembled on a screen. “AI-generated” tells a customer something about the picture’s origin. It says little about whether the portion, ingredients or dish resemble what will arrive. A buyer needs help interpreting the promise, as well as the picture.

This distinction matters most when a disappointing purchase consumes money a household cannot comfortably waste. Consumer agency is the ability to choose on dependable information and obtain a remedy when a seller misleads. Giving people another warning to decode can leave that responsibility largely on their shoulders.

The case for cheap creative tools deserves a fair hearing. Advertising has always used illustration and artifice. A neighbourhood shop could gain access to attractive promotional material without a large production budget. Restrictions that treat every invented image as deception could protect established brands from smaller competitors.

Yet appearances can become cheap while checking them remains expensive. If a careful manufacturer and a careless one can present equally persuasive evidence of quality, customers may struggle to reward the better product. Sellers could then face stronger incentives to improve the advertisement than the thing advertised. That is a plausible competitive mechanism, not an observed result of this new law.

The gains would also be uneven. Software suppliers can charge for generating campaigns; platforms can earn from distributing them. Buyers and merchants may bear the cost of investigating discrepancies and handling complaints. A saving in content production becomes a social gain only to the extent that it survives these additional costs.

For creative workers, the likely pressure is on particular assignments: routine product images, variations and promotional layouts. Businesses may commission less of that work while needing more documentation of actual products and human checking. Neither mechanism establishes a net employment outcome. The practical question for employers is whether savings fund reliable communication or simply remove the people who notice misleading details.

A 2025 study in PNAS Nexus offers a useful caution. Across two preregistered experiments involving 7,579 Americans, the tested labels reduced belief in claims conveyed by misleading AI-generated images. Simple AI-origin labels, however, generally did little to participants’ stated willingness to engage with the posts. These were survey responses, not purchases, and cannot establish how Indian consumers would respond.

Pennsylvania’s legislation preserves other legal remedies. That matters: disclosure need not become permission to mislead. India’s Consumer Affairs department made a related principle explicit in its 2022 account of advertising guidelines: a disclaimer should not conceal material information or repair a misleading claim. The FTC’s 2024 rule addressing fake reviews, including invented AI reviewers, likewise concerns whether claimed experience exists.

For Indian markets, the design challenge extends beyond attaching English text to an image. Disclosures should work in the language and medium of the sale, including audio where needed. Small merchants should have inexpensive ways to show actual dimensions, contents and product photographs. Buyers should not need forensic expertise, a fast connection or repeated calls to obtain basic evidence.

Three futures follow from these choices. In a better one, affordable illustration sits alongside dependable product information, authentic reviews and accessible redress. Smaller firms compete through both reach and quality. In an uneven one, established brands buy reassurance while unfamiliar sellers struggle to earn trust. In a harmful one, familiar labels fade into the background and a contest in manufactured appearances displaces investment in what customers receive.

Which promises in your advertising can a customer independently check? Who owns the work of verifying them? Do lower creative costs still look attractive after complaints, returns and lost repeat purchases?

Platforms should test what buyers understand and make corrections easy. Merchants should keep evidence behind material claims and distinguish illustrations from representations of actual stock. Investors should ask whether cheaper persuasion improves lasting customer value. An AI label is useful when it begins that examination. The product must still withstand it.

Sources: Pennsylvania General Assembly, bill history and text; Wittenberg et al., PNAS Nexus; India Consumer Affairs, 2022 review; FTC, 2024 fake-review rule announcement.