The narrative in recent AI news has moved from how powerful models are and how many tokens they burn to what agents can handle on our behalf. I’m more interested in something less flashy: when consumers hand “find someone for me” over to AI, a business’s marketing copy may suddenly become a specification document.
And the worst thing in a specification document is “anything goes.”
I lean toward a somewhat irritating view: To get recommended by AI, businesses may first need to make themselves easier to rule out. The more specific the need, the less useful “we offer comprehensive services” becomes for matching. It sounds all-powerful, but break it down and there isn’t a single condition you can verify. Engineers who see this kind of requirements document usually go make coffee first.
Customers Aren’t Asking for Your Keywords Anymore
Imagine a renter discovering that their air conditioner is dripping at night. Instead of searching for “recommended air conditioner repair in Taipei,” they ask an assistant:
Find me a technician who can come to Zhongshan District tomorrow evening and is willing to repair a split-system air conditioner that’s over ten years old. The rental can’t undergo major dismantling, and tell me the on-site diagnostic fee first.
This is a hypothetical scenario, but it lays out the difference clearly: the customer is handing over a bundle of constraints, not a string of search terms.
Business A has a beautiful website that says “professional team, dedicated service, guaranteed quality.” Business B has a bare-bones page, but it lists service areas, how to book evening appointments, the scope of inspections and repairs for older units, and diagnostic fees. It also explains which kinds of leaks may require dismantling and reinstallation.
Business B isn’t necessarily better at repairs. But when it comes to deciding who fits this particular need, the clues Business B has left are more usable. There’s no mysterious AI wizardry here. Business A has supplied advertising slogans; Business B has supplied material for making a judgment.
“Don’t Be Too Definite” Is Starting to Come at a Cost
Traditional marketing often worries that spelling out conditions in too much detail will drive customers away. Don’t post prices yet, don’t mention limitations yet—let people send a private message first. There are reasons for this approach: services require estimates, and they require communication.
But if consumers ask AI to put together a shortlist first, that private message may never happen. Missing information doesn’t necessarily mean flexibility; it may simply mean the assistant can’t confirm anything. Whether it rules a business out, asks follow-up questions, or guesses wildly depends on the tool and its data sources; “AI will understand” cannot be an operating strategy.
More troubling still, complete specifications don’t mean reliable capabilities. An ordinary business that explains itself clearly may be easier to identify than a highly skilled technician whose online information is stuck five years in the past. What AI recommendations amplify is sometimes the ability to organize information, not entirely service quality. That gap is where marketers actually have work to do.
Rather than churn out another twenty articles for a technician on “Five Tips for Air Conditioner Maintenance,” I’d rather organize the questions he answers on the phone every day: which models he has repaired, which areas he won’t travel to, what the quote includes, and exactly which fault each case resolved. Structured data can help machines parse information, but it is no guarantee of a recommendation; if the website, business information, and actual conditions for accepting jobs don’t line up, even the prettiest markup merely arranges the contradictions neatly.
Visibility still matters, and reviews, word of mouth, and source credibility haven’t left the stage either. It’s just that there is still a stretch of evidence between “lots of people have seen me” and “there’s a reason to choose me for this particular need.”
That air conditioner technician may still dislike writing articles. He picks up the phone and opens with: “I can take a look at units over ten years old, but send me a photo of the drain pipe first. The only slot left tomorrow evening is 8:30. Let me tell you the diagnostic fee up front.”
He says these things every day. Yet his website’s homepage still displays four big words: Service Comes Above All.