AI Search Turns Prompts Into Customer Research, MarTech Says

By Elena Vasquez

Two users can now submit identical prompts to the same AI assistant and receive different answers, because the assistant holds context from earlier conversations — the core argument of an analysis published on MarTech titled "AI search is turning prompts into customer research."

The piece argues that AI search has not made consumer needs more complex. It has given people permission to express the complexity that was already there. And that shift turns every prompt into a stream of customer insight that brands can harvest.

What changed in how people search?

Before search engines, people with complicated questions asked friends, visited a library, or sat down with an expert such as a travel agent, the author writes. Search engines then trained users to compress those needs into keyword strings like "best family hotel Tenerife" — four words that omit most of the underlying brief.

The MarTech analysis contrasts that with a modern AI query:

• Party composition: two adults, two young children.
• Room requirement: separate sleeping area for the children within the same room.
• Priorities: good food, a car rental, swimming pools.
• Hard constraint: the kids' pool must be heated.

That level of detail previously went to a travel agent, who might still have struggled to answer every part of it, the author notes.

Why do bigger questions matter to marketers?

A keyword can indicate intent. A fuller conversation can expose the drivers and barriers behind that intent, according to the article. Circumstances, anxieties, motivations, previous experiences, priorities, and trade-offs can all enter the discovery process itself — intelligence that historically emerged only later, often in a conversation with a salesperson.

The content implication is direct. If people search with this specificity, brands need content with the depth and nuance to answer situational questions, not just broad category pages. "Traditional search became very good at understanding what we were looking for," the author writes. "AI search has the potential to understand what we're actually trying to achieve."

What does AI memory do to the prompt?

The analysis identifies a deeper shift: the prompt is no longer the complete expression of intent. As people settle on one or two AI tools for daily use, those tools build context from previous conversations — habits, preferences, past likes and dislikes.

The running shoe example illustrates the mechanism. One user has previously discussed a knee injury with their assistant; the other has not. The prompts may be word-for-word identical, yet the injury context can produce two different recommendations. "The knee injury doesn't have to appear in the prompt to be relevant to the answer," the author writes.

Memory and personalization capabilities vary between platforms and depend on individual settings and permissions, the article cautions, so this will not apply to every AI search experience. The direction, however, is clear. A user may eventually say: "You know that hotel we stayed at in Mexico that I loved? Find me the Tenerife version" — a query where the meaning lives entirely outside the prompt.

Are prompts the new keywords?

The article reports heavy industry interest in prompt tracking (monitoring what AI tools say about a brand in response to standard queries), driven by years of keyword-ranking habits that reached SEO reports, marketing reports and, in some businesses, board reporting.

Typical tracked questions include:

• What does ChatGPT say when somebody asks this question?
• Does our brand appear?
• Which competitors are mentioned?
• How frequently are we recommended?

The author does not dismiss the practice: "I'm not suggesting we shouldn't track these things. There's absolutely value in doing so." But if two people can ask the same question and legitimately receive different answers, a standardized prompt can only ever show one version of the possible experience. The article recommends supplementing prompt tracking with site data: which pages receive arrivals from AI platforms, and what increasingly specific content attracts those visits.

How does relevance become situational?

"Best running shoes" creates one kind of competition. But an AI that knows the user is training for a first marathon, has a knee injury, values comfort over speed, prefers a particular fit and will spend up to $150 effectively redefines "best" for that individual.

That dynamic rewards niche focus, the analysis argues: content built for specific needs and circumstances rather than the highest-volume topic, plus presence in niche influencer channels, specialist publications and communities. The aim, the author writes, is to give AI enough reason and evidence to understand when a brand might be the best choice for somebody.

Why is this a business strategy question?

The instinctive question — "How do we get AI to recommend our brand?" — is narrower than the opportunity, according to the article. The more valuable questions include:

• Who are we genuinely brilliant for?
• Under what circumstances are we the right choice?
• Why do people pick us that we might not even realize?
• What evidence exists to prove it?
• Are there unmet needs we can now see that could change the proposition itself?

The author's agency calls its approach behavior-led search, on the premise that the biggest opportunity comes from understanding the human on the other side of the technology. Some of the richest insight may not come from a search tool at all, the article notes, but from salespeople, customer service teams and frontline staff who already hear the oddly specific questions websites fail to answer.

"Those aren't SEO problems. They're business opportunities," the author writes. The opportunity for brands is not simply to appear for more prompts — it is to understand people deeply enough to be the right answer when it matters.