Try something for me. Open ChatGPT, or Perplexity, or Google’s AI Mode, whichever machine your phone has been quietly steering you toward, and ask it to recommend the best companies in your industry, in your region. Phrase it the way a real buyer would. “Who should I hire to…” or “What’s the best…”
Go ahead. I’ll wait.
If you just watched an AI confidently recommend three of your competitors, a company two states away, and a business that closed in 2023, congratulations. You’ve discovered the most important marketing problem of 2026. Your buyers discovered it about eighteen months ago. They just didn’t send a memo.
/ 01The quiet part, out loud
While everyone was arguing about whether AI would take our jobs, it took our search bar instead.
Gartner projected that traditional search volume would fall 25% by 2026 as buyers shift to AI assistants. Pew Research measured what happens when Google shows an AI summary above the results: click-through rates drop from 15% to 8%, which means nearly half the clicks are gone. Adobe’s analytics team tracked AI referral traffic to retail sites growing more than 700% in a single year.
Put those numbers side by side and the story is plain. Your customers did not stop researching. They stopped clicking. Somewhere between their question and your website, a machine now reads the entire internet, synthesizes an answer, and hands it over in a paragraph. If you’re in the paragraph, you’re in the deal. If you’re not, you don’t exist, and no amount of ad spend fixes nonexistence.
This is a structural change in how demand finds supply, and it arrived at precisely the moment most mid-market companies finished optimizing for the old system. Nobody’s fault. Just expensive.
/ 02Why the machine skips you
Your website was built to persuade humans. AI doesn’t read persuasion. It reads structure, specifics, and evidence, and it treats most marketing copy the way you treat a cold email that opens with “I hope this finds you well.”
Consider the sentence at the top of thousands of mid-market websites: “World-class solutions for discerning clients.” A human reads that and shrugs. A language model reads it and finds nothing to hold. No industry, no geography, no numbers, no claim it could verify. That sentence could describe a law firm or a hot tub dealer, and the machine, unable to tell the difference, cites neither.
The research on answer-engine behavior is remarkably consistent about what does get cited. Content that asks the buyer’s actual question and answers it directly. Specific numbers with named sources. Company facts that match everywhere they appear, from your site to the directories to the reviews. And corroboration from third parties, because a claim only you make is a claim the machine discounts.
The machines have opinions about trust. They formed them without asking you.
What you can actually do this quarter
None of this requires a moonshot. It requires somebody paying attention, which in my experience is the scarcer resource.
Interrogate the machines monthly
Ask ChatGPT, Perplexity, and Google’s AI the ten questions your buyers actually ask. Write down what comes back. That transcript is your new brand audit, and it’s more honest than your last one. The machine has no reason to spare your feelings.
Publish answers, not ads
Take your buyers’ real questions and make them headlines. Answer in the first sentence. Support with numbers and named sources. Be the answer, not the ad, because the ad is increasingly a thing the machine summarizes past.
Fix your facts everywhere
Same name, same address, same claims, same numbers, on your site, your LinkedIn, the directories, the reviews. AI cross-references. Inconsistency reads as unreliability, and unreliable doesn’t get cited.
Earn mentions you don’t control
Trade press, industry associations, podcasts, local business journals. The machine weighs what others say about you more heavily than what you say about yourself. So does everyone, honestly. The machine just does it at scale.
Assign an owner
Not a vendor, and not “the agency handles digital.” A named person accountable for one question: what do the machines say about us, and is it getting better? Reviewed monthly, like cash flow. What gets measured gets cited.
The part nobody wants to hear
Most companies will read a list like that, nod, and do none of it. Not out of laziness. Every item on the list sits between departments. The website belongs to marketing, the directories belong to nobody, the press belongs to a PR firm you stopped calling, and the strategy belongs to whoever has ten free hours, which is no one.
That gap between knowing and doing is where this whole shift will be won or lost. MIT’s researchers found that 95% of enterprise AI pilots deliver zero measurable ROI, and the failures rarely trace back to bad tools. They trace back to tools with no operating system around them. No owner, no cadence, no connection to revenue.
So the real takeaway is smaller than a strategy and bigger than a tip. Somebody in your company should own being the answer. If a name came instantly to mind, this piece cost you five minutes and confirmed you’re ahead of most. If you’re still thinking, that silence is worth sitting with.
The companies that win the next decade will be the ones the machines learned to trust first. Those opinions are forming right now, this quarter, with or without your participation.
I’d participate.
If the question is narrower, whether an FAQ page still earns a place in those answers, do FAQ pages still help us show up in AI answers? takes it on.
This post is part of AI search visibility: every question, in the order it comes up, a reading guide to the whole series.