BoSacks Speaks Out: When the Audience Is No Longer Human

Every once in a while, the old publishing dinosaur in me rears its head, looks around and says, “What the hell?”

That was my reaction to Shelley Seale’s INMA article, “BCN, Time share lessons in selling ads to machines.”

I am not a digital holdout. I became digitally integrated before many of my publishing peers. I embraced online distribution, email newsletters, websites and, more recently, artificial intelligence. I understand that publishing has always changed with the available technology.

But this latest development gives me pause.

Advertising to human beings was complicated enough. Now publishers are developing advertising specifically for machines. Not advertising created by machines, which is already happening, but advertising designed to be read, processed and acted upon by AI agents and bots.

According to the article, more than half of today’s web traffic is not human.

At Time, bot and agent traffic sometimes exceeds human traffic. In response, Time has divided its website into two versions: one for people and another, stripped-down, machine-readable version for AI systems. It has even begun serving sponsored content directly to those machines.

BCN in Germany is taking another route. Its GEO Brand Impact product attempts to improve the likelihood that a brand will be mentioned, cited or recommended in answers generated by artificial intelligence. GEO, or Generative Engine Optimization, is rapidly becoming the AI-era successor to SEO.

I understand the business logic. If people are increasingly asking ChatGPT, Google and other AI systems for answers instead of searching through lists of links, brands naturally want to influence those answers. Publishers, meanwhile, need to replace traffic and revenue being lost as searches incre

From a commercial perspective, it is ingenious. From a publishing perspective, it is also disturbing.

We may need a new definition of publishing.

For centuries, publishers created and distributed information for human beings. Even when search engines entered the equation, the goal was still to help people discover and visit the publisher’s work. Search engines were the road. The reader remained the destination.

AI threatens to reverse that relationship. The machine may now become the publisher’s immediate audience, while the human receives an answer assembled somewhere else. The reader may never see the original publication, visit its website or even recognize where the information came from.

That changes the equation.

It also raises a question that the publishing industry cannot afford to duck: Where does legitimate brand information end and manipulation of the answer begin?

BCN sees an opportunity to use the authority and trust of established media brands to help advertisers influence what large language models cite and recommend. Time is producing sponsored information specifically structured for machine consumption. Both approaches may be clearly labeled and carefully monitored, but the implications are larger than another clever advertising product.

A human reader can see an advertisement and understand that someone paid for it. An AI-generated answer may blend information from publishers, advertisers, forums, videos and countless other sources into a single confident response. Will the person receiving that answer know which facts were promoted, which were independently reported and which appeared because a brand paid to improve its visibility?

The machine may not have emotions, as the article notes, but it is hardly neutral. It reflects the information it receives, the sources it favors and the systems designed to rank those sources.

I am not arguing that publishers should ignore AI agents. That would be commercial suicide dressed up as principle. If machines are becoming gateways to information, publishers must understand how those gateways operate. They should make their journalism readable and properly attributable. They should demand compensation when their work creates value. They should also explore responsible new revenue streams.

But publishers must protect the distinction between editorial authority and paid influence. If trust is what makes a publication valuable to an AI system, selling that trust carelessly will eventually destroy the very asset being monetized.

Perhaps machines are a new audience. Perhaps they are better understood as powerful intermediaries between publishers and people. Either way, the ultimate audience is still human. Machines do not buy cars, vote, subscribe to magazines or care whether the information they receive is true. People do.

The publishing dinosaur in me is not objecting to change. He has survived quite a lot of it. He is simply asking whether we fully understand what we are building.

Advertising to machines may become a substantial new business. But before we celebrate it as publishing’s next great revenue stream, we had better establish clear standards for disclosure, attribution and the separation of advertising from independent information.

Otherwise, publishers may discover that they have taught the machines how to sell without teaching them whom to trust.