In March 2026 I opened a discovery call expecting to run our usual introduction, and the founder on the other end stopped me inside the first minute. He said he had already decided to work with us and only wanted to confirm a start date. When I asked what had convinced him, he did not mention our sales deck, a referral, or a case study. He said he had asked an AI assistant which firms handled his exact problem, read what it returned, checked us against the two other names it listed, and made his choice before we ever spoke. That call reset how I think about strategic marketing for a small business, because the real decision had happened in a room we were never invited into.
We tracked that thread across 9 clients and over 6 weeks, and the pattern held more often than I was comfortable with. Buyers were arriving at the first conversation already sorted into yes or no. The companies winning were not the ones running the loudest campaigns. They were the ones an AI assistant could describe accurately when a buyer asked a plain question about their category. What we got wrong, for years, was treating the website and the ad account as the storefront. The storefront had quietly moved to a place we did not control and were barely measuring.
The mechanics were less mysterious than they sounded. When someone asks an assistant to recommend a vendor, the model assembles an answer from what it can find and trust across the open web. If your company answers the real questions in your category in clear, sourced, first-person language, you show up in that answer. If your public material is vague brochure copy, you are invisible at the exact moment a buyer is deciding. One of our clients had a polished site that said almost nothing specific. A smaller competitor with a plain blog that answered ten concrete questions was the one the assistant named. That was the whole game, and we had been optimizing the wrong surface.
Fixing it was unglamorous operations work, not a rebrand. We started by writing down the fifteen questions our clients’ buyers actually ask before they buy, the ones that come up on every sales call. We pulled those straight from HubSpot call notes and the recurring threads in our shared Slack channels, because the questions were already there, just never turned into public answers. Then we had each founder answer those questions in their own voice, with a real number or a real example attached, and published them under a named author rather than an anonymous brand account. We were not writing marketing copy. We were writing the reference material a machine would read on the buyer’s behalf.
The part I underestimated was how much the author mattered. Early on we published a few of these answers under a generic company byline, and they did almost nothing. The same content, republished under the founder’s name with a title and a face, started getting picked up. In retrospect this is obvious. A buyer trusts a person more than a logo, and it turns out the systems that summarize the web lean the same way, because a named human staking a claim reads as more accountable than a faceless page. The lesson for any small business is that authority is a person, not a brand, and pretending otherwise leaves the strongest signal you have on the table.
The measurement took some invention, because the old analytics were built for clicks, not for answers. We could not see inside a ChatGPT session or a Google AI overview, so we did the low-tech thing and just asked. Every new lead across those accounts got one added line on the intake form asking how they first heard about the company, and we sorted the free-text answers by hand, a little over 200 tags in all. It was tedious, but it was the only honest read available, and it beat guessing. What the tags showed was that referrals and AI answers had started to blur. People would ask an assistant, get a name, then ask a friend to confirm it, and report the friend as the source even though the machine had built the shortlist. The old attribution model had no column for that step, so it had been crediting the wrong one for months.
There was a leadership cost to this that I did not expect, and it is the part most operators skip. Being the answer means the founder has to spend real time being quotable, and that time has to come from somewhere. For two of our clients the honest move was to cut a channel that felt productive but was not, a weekly webinar in one case and a large paid-social budget in the other, and move that time and money into first-party answers. Those calls were uncomfortable, because the cut channels had defenders inside the company. What made them possible was the tagging. Once the count of deals that started with an AI answer sat on a dashboard next to the spend, the reallocation stopped being an argument about taste and became an argument about evidence.
None of this is a reason to abandon the fundamentals. The clients who benefited still had a working product, a clear offer, and a real sales process. What changed was the order of operations. Instead of spending first on interruption and hoping the website closed the gap, they spent first on being genuinely findable and correct in the place buyers now start, and let the paid channels do the narrower job of reaching people who had not begun looking yet. The businesses that treated AI visibility as a side experiment kept paying to reach buyers who had already, quietly, ruled them out.
If I had to compress the whole 6 weeks into one instruction for another operator, it would be this. Go find the last ten deals you lost and ask, honestly, where the buyer was when they decided. More and more, the answer is not your funnel. It is a conversation with a machine that summarized your category for them. The strategic marketing job now is to make sure that summary is accurate, specific, and attached to a real name, because that summary is doing the selling whether you help write it or not. We at FORKOFF spent years getting good at the old storefront. The useful, uncomfortable lesson from that March call was that the storefront had moved, and the only real question left was whether we would be the ones to furnish it.
Be the first to comment