When the audience disappeared

In the first week of May 2028, a large entertainment company announced its greatest audience ever. Its reported reach had passed two billion daily encounters. At the same time, its sales division told the board that human purchasing decisions attributable to those encounters had fallen for the sixth consecutive quarter. Nobody had falsified the figures. The impressions were real. The viewers were not. Most of them were personal agents gathering summaries and dismissing ninety-nine items out of every hundred before their owners had even woken up.

The change had not begun with a boycott. Two years earlier, the typical consumer still searched for a holiday, compared prices and opened the pages paid to appear at the top. By 2028 a household delegate was checking room inventory, refund conditions, weather probabilities and the reliability of local transport directly against a supplier's machine-readable contract. The advertisement never appeared on a screen. The consumer received three proposals, each already negotiated. The decisive economic encounter had taken place between software systems, far from the billboard and long before the person said yes.

The first companies to notice the reversal were not the search giants. They were small hotels, insurers, farms and freight agents. A merchant could increase advertising expenditure and attract no additional buyers because purchasing agents rated the underlying agreement, not the campaign. Brands with famous names and poor contract terms abruptly became less visible than obscure suppliers whose delivery records could be verified by machines. The long, costly work of creating public desire lost its monopoly on selling.

What replaced the advertisement was more troubling: the access schedule. Agents negotiated fees to examine offers. A manufacturer could buy priority in a thousand personal agents' decision queues, but the fee no longer guaranteed that a human would ever see the product. Platforms described the arrangement as efficiency. Merchants called it an invisible tax on participation in the economy. Within months the industry's most profitable product was not an image or a search keyword. It was the privilege of being considered.

Then came the great measurement fraud without a fraudster. Investment funds, broadcasters and governments had spent decades treating audience size as a proxy for influence. In June, the largest independent review discovered that two rival political broadcasts had each been 'seen' hundreds of millions of times by assistants assessing the content for their owners. Human exposure was a fraction of the published number. A minister could appear to command the attention of a nation while reaching hardly anyone beyond professional summary services. The political class discovered what media executives already knew: a million impressions were no longer a million minds.

The agents did not merely reject messages. They could reorder them. A household assistant might summarize a prime minister's speech before a family discussion, selecting the claims most relevant to mortgage payments and ignoring the claims about identity or national destiny. A different household received an opposite emphasis. The speech had been public; its effect was privately reconstructed. Governments began demanding a common record of what citizens had been shown. Civil-liberties organizations warned that a state-defined summary was another form of control. The old fight over censorship mutated into a fight over interpretation.

The winners in the new economy were those that owned reliable interfaces to the agents, not necessarily those that owned audiences. The losers included advertising exchanges whose bids were denominated in human attention, publishers financed by accidental clicks and creators whose influence lived inside a platform's numerical follower count. Money migrated toward verification, fulfillment, trust and the ability to guarantee a result. A new kind of publisher did not ask, 'How many people read this?' It asked, 'How many decisions did this change, and which agent authorized them?'

By October, the entire advertising industry was attempting to sell proof of human encounter. Some companies introduced certified live broadcasts where a person had to remain present to unlock a commercial message. Others invented experiences too valuable to summarize. Physical concerts, unrepeatable human performances and communal screenings suddenly commanded prices that downloadable entertainment could not. Paradoxically, the faster content became infinite, the more expensive one shared moment could become.

The great split of 2028 was therefore not between those who used SI and those who rejected it. It was between people whose agents could negotiate on equal terms and people represented by cheap, opaque assistants subsidized by advertisers. One class owned the right to filter the world. Another rented a filter and never knew whose interests it served. The treaty writers of 2031 would eventually describe this as the first border of the synthetic age. It was already in place, invisible, around every human decision.