By Franklin Rios, CEO of Next Net
Optimization has a way of making bad ideas look sophisticated. Give a marketer enough dashboards and machine-learning muscle, and almost anything can acquire the sheen of inevitability.
But optimization has always had one inconvenient dependency: It needs to know what it is optimizing for.
That dependency is becoming a much bigger problem as brand discovery shifts from search results pages and publisher sites into AI-generated answers, summaries and recommendations. Marketers naturally want to “show up” in ChatGPT, Claude or Gemini. Showing up is a terrible strategy, however, if the brand does not understand where it is showing up, why it is being included, what is being said about it and whether the system has understood the business correctly in the first place. That may sound obvious, but it is not behaving that way in the market.
In traditional search, the optimization target was at least legible. A marketer could care about rankings, revenue or a host of other metrics. Those signals were imperfect and occasionally worshipped with an excessive level of devotion. But they were visible. A human searched, saw links, clicked or did not click, landed on a site, converted or did not convert.
AI-mediated discovery breaks that tidy little chain. The consumer may never scroll through a list of links. The answer engine may summarize the category, select a few sources, exclude others and satisfy the user before a brand’s analytics platform ever sees a session. Meanwhile, the brand’s website is no longer just a destination for human prospects. It is also raw material for crawlers, retrieval systems and agentic tools that may read, summarize and reuse information without behaving like anything a legacy marketing stack was built to recognize.
This creates a measurement problem with strategic consequences. If site traffic rises, is that more human demand, more bot activity or more machine retrieval? If traffic falls, is the brand losing relevance, or are users getting what they need from AI answers without clicking through? If a brand is cited frequently, is it being cited for the right reasons, in the right context, with the right associations? Or has it become the internet’s favorite example of something it would rather not be known for?
These are not academic questions. Automated systems optimize toward signals. If the signals are polluted, incomplete or misunderstood, automation will not pause for a discussion about semantic accuracy. It will simply scale the wrong response with impressive efficiency.
That is the peculiar danger of this moment. The machine does not need to be malicious. It just needs a bad brief. A brand might decide it needs more AI visibility and flood the web with content designed to attract citations, only to discover that the content reinforces generic category language instead of the specific buying contexts that matter. A B2B company might optimize for broad awareness when its real opportunity lies in surfacing for a narrow set of high-value personas. A retailer might chase answer-engine mentions while losing control of product, pricing or availability context. A financial services firm might produce technically accurate content that fails to establish trust with the actual decision-makers it needs to reach.
In each case, the problem is optimization before orientation. That is why the current brand and advertiser tech stack needs rethinking. Marketing technology needs to evolve because the unit of discovery is changing. Marketers now optimize how machines interpret the brand on behalf of people.
That requires a more disciplined measurement model than the one many marketing organizations currently use. Site analytics, search visibility, content performance and conversion data still matter, but they no longer describe the whole discovery environment. They capture the parts of the journey that remain visible to the brand. Increasingly, some of the most important influence happens before that visibility begins.
The practical challenge is to reconnect measurement with meaning. The marketer’s job is to understand which signals still represent human intent, which signals reflect machine activity and which signals reveal how the brand is being interpreted in the market.
That work is less glamorous than launching another automated optimization program, which is unfortunate, because “clean up your assumptions before scaling them” does not look great on a conference agenda. But it is where the leverage is. AI systems will keep changing how information is retrieved, summarized and presented. Marketers need to know what they want to be known for, which audiences matter most and where the current data stops being a reliable proxy for reality.
