The Student Who Became the Storefront
Paul had run an estate planning law firm in Portland for seventeen years. He was not a marketer by nature but he had learned over time that publishing useful content was how people found him online. He wrote articles about probate timelines, about the difference between a will and a trust, about what happens when someone dies without an estate plan in Oregon. He wrote them himself, at his desk, after his last appointment on Tuesday evenings, because he believed that people deserved plain-language answers to questions that frightened them.
Those articles brought him clients for a decade. A person would search for an answer to a question about their aging parent's estate, find Paul's article, read it, trust him from the writing alone, and call. The content was the introduction. The introduction became the relationship. The relationship became the work. Paul wrote more articles. The system rewarded him with more introductions.
In 2023, something changed in the introductions. They kept coming, but fewer of them. The articles still ranked. The traffic numbers still showed visits. But the calls were different. The people who called had already read everything. They had already decided what they wanted. They arrived without the preliminary questions that used to be the opening of a relationship. They arrived as if they had already had a long conversation with someone who knew Paul's subject matter thoroughly.
They had. They just had not had it with Paul.
What the Training Actually Built
Search engines have indexed publicly available content since the early days of the web. That was always understood. What was not fully understood, because it was not fully disclosed, was the secondary use of that indexed content as training data for large language models.
The articles Paul wrote on Tuesday evenings, the questions he answered, the plain-language explanations he crafted for people who were frightened about their parents' futures, went into a body of publicly available text that AI systems used to learn how to answer questions about estate planning. The systems learned from practitioners like Paul across every field. They learned from dentists explaining procedures, from HVAC technicians explaining maintenance, from financial advisors explaining retirement accounts, from chiropractors explaining treatment protocols.
The AI that now sits at the top of search results and answers these questions in AI Overviews learned to answer them, in large part, from the people whose clients it now intercepts. The student became competent. The competent student became the first answer a searching person receives. The first answer is where the relationship used to start. Now the relationship starts with the AI and sometimes ends there too.
Paul's articles did not disappear. His rankings did not collapse. The traffic did not go to zero. What changed was the shape of the funnel. The top of the funnel, where curiosity became engagement and engagement became trust, was partially absorbed by a system that learned its subject matter from people like Paul and now uses that knowledge to answer Paul's potential clients before they need to find Paul.
Why This Is Different From Every Other Algorithm Change
Every few years, a major algorithm change reshuffles the organic search landscape. Sites gain or lose traffic. Strategies that worked stop working. New approaches emerge. Practitioners adapt. This is the normal rhythm of building a digital presence in a world where the platforms change the rules.
The AI shift is different in kind, not just in degree, and the difference is in the direction of the value flow.
Previous algorithm changes redistributed traffic. A slow site lost ground to a fast site. A thin-content page lost ground to a deep one. The traffic moved between competitors. Another business, often another small business in the same space, received what the losing site no longer got. The market restructured but the market's participants were still the same participants.
The AI shift does not redistribute traffic between competitors. It absorbs traffic into the platform. The person who would have visited Paul's article, then called Paul, then become Paul's client, now receives the answer from the AI Overview and in a meaningful percentage of cases considers the question answered. The traffic does not go to a competing law firm. It does not go anywhere. It stays on the search results page, held by a system that learned the answer from people who built their businesses on the expectation that providing good answers would bring them clients.
The implicit contract of the past twenty years was: publish your expertise openly and the platform will send you the people who need it. That contract was never written down. It was structural. The platform needed content to be useful. Practitioners needed clients. The arrangement worked for both. The AI layer restructured the arrangement without renegotiation. The practitioner still publishes. The platform now serves the answer itself. The practitioner's role changed from source to training data, and the transition happened gradually enough that most practitioners did not notice until the call volume had already declined.
The Three Things the AI Cannot Do
The AI that answered Paul's potential client's questions about probate timelines cannot represent them in probate court. It cannot review the specific language of their parent's existing will and identify the clause that will cause a dispute. It cannot sit across a table from a grieving family and explain what happens next in plain language that accounts for the specific family dynamics in the room. It can answer the general question. It cannot do the specific work.
This is where the businesses that are navigating the AI shift well have found their footing. The informational content that the AI absorbed was always the entry point, not the value. The value was always the specific, personal, expert application of knowledge to a situation that an AI cannot fully understand from a search query. The AI made the entry point less exclusive. It did not make the value less real.
What changed is that the entry point no longer automatically leads to the practitioner. The client now arrives having done more research, having fewer elementary questions, and in some cases having formed a preliminary judgment from the AI interaction before the practitioner says a word. The practitioners who are doing well are the ones who understand this arrival condition and have adjusted their digital presence to meet clients at a later point in their decision rather than at the beginning of their research.
The second thing the AI cannot do is convert a visitor once they arrive on a website. The AI Overview answers the question in the search results. It does not book the appointment. It does not place the call. It does not fill the contact form. When a person decides they need a practitioner rather than an answer, they still go to a website and that website still determines whether they call or leave. The AI restructured the discovery phase. It did not touch the conversion phase. The conversion phase belongs entirely to the business whose website the visitor lands on.
The third thing the AI cannot do is replace the trust that comes from a specific practitioner's specific perspective over time. Paul's Tuesday evening articles did something that a generalized AI answer cannot do: they sounded like Paul. They had his particular way of explaining a complicated thing in plain terms, his specific patience with the subject matter, his voice. That voice built relationships with people who had never met him. The AI produces accurate summaries. Accurate summaries do not build the same kind of trust that a distinctive voice does, and a distinctive voice in a specific niche, published consistently, is still one of the most durable competitive assets a practitioner can build.
What the Practitioners Who Are Doing Well Actually Changed
They did not stop publishing. They changed what they published and why.
The content that the AI absorbed was primarily informational: what is, how does, what happens when. The AI can answer those questions now. The content that the AI cannot absorb effectively is experiential: what I have seen, what this specific situation looked like from inside, what the general principle looks like when it meets a real and complicated case. First-person case-based content is still the practitioner's exclusive territory because the AI was not in the room when it happened.
They also changed where they put their energy within their digital infrastructure. The practitioners who are converting well from the visitors who do arrive have fast websites. They have measurement systems that tell them what is actually happening when a visitor arrives, not a GA4 approximation filtered through ad blockers and iOS privacy restrictions. They know their real conversion rate rather than their reported one. They have a contact path that works on a phone in three seconds rather than a form that loads slowly and times out on a mid-range Android.
The AI shift reduced the volume of visitors arriving at the top of the funnel. The practitioners who responded by improving their infrastructure at the bottom of the funnel are the ones whose revenue did not drop proportionally with the traffic. A practitioner who converts 6 percent of arriving visitors rather than 3 percent needs half the traffic to produce the same number of clients. The AI took some of the traffic. Better infrastructure took back the revenue.
The Auditor's Take
Paul did not need to fight the AI. He needed to stop operating on the assumption that the top of the funnel would always work the way it worked in 2015. It does not. The AI is real, it is here, and the content that fed it is not coming back.
What Paul still has is the thing the AI cannot take: the work itself, the voice, the specific expertise applied to specific situations, and the infrastructure that converts the visitors who do arrive. The AI made the discovery phase less reliable as a traffic source. It did not make the work less valuable or the website less important as a conversion tool.
The businesses that are struggling are the ones still trying to win at the top of the funnel with the same informational content strategy that the AI absorbed. The businesses that are growing are the ones that understood what changed, adjusted their content toward experience rather than information, and built infrastructure that converts the visitors they still receive at a rate that compensates for the ones the AI intercepts.
The training happened. The student is working. The question now is what you build with the infrastructure you still own.
Based on patterns observed across multiple audits. All identifying details are illustrative. The diagnosis is always free.
Ready to Fix This on Your Site?
Run your site through PageSpeed Insights and send me the score. Free diagnosis. No pitch. Just the numbers.