A buyer asks a question your website was built to answer and gets a complete response without clicking anything. No results page to scan, no link to follow. The answer arrives assembled, and the vendors named inside it receive the only visibility that exchange produced.
That interaction is becoming common, and most of the marketing industry is still selling the old model at full price. I have watched this pattern in software: a distribution channel everyone depends on changes its physics while vendors keep selling maps of the old terrain.
Here is what the evidence shows, what I expect, and where I would move budget.
What the data shows
Three findings, each with its limits stated.
- Clicks fall when an AI summary appears. Pew Research tracked the browsing of 900 US adults in March 2025. When Google showed an AI summary, users clicked a traditional result in 8% of visits, against 15% without one, and clicked a link inside the summary in 1% of visits. This is consumer behavior on Google, and B2B buyers were not separated out.
- B2B referral traffic from AI is still small. Octane11, analyzing 400 million B2B website sessions from September 2025 to June 2026, found AI search under 2% of B2B search referrals, although about 9% of closed-won deals had an AI search touch. The data is vendor-reported. It also shows why the effect hides: much of the influence never arrives as a click.
- Buyers use AI, then check it. Forrester's 2026 buying research says generative AI search is now a starting point for vendor research, and that buyers validate it with peers and trials because the answers are often incomplete or unreliable.
What the evidence does not show is the pace. Nobody can tell you the share of your search value that disappears by 2027. The direction is measured. The speed is a forecast.
What loses value
Classic SEO had one job: win the click. Rank on the results page, earn the visit, convert on your site. The tactical stack grew around that loop: keyword-targeted service pages, high-volume blogging calendars, word-count arms races, content written for crawlers first.
As answers replace clicks, the tactics that only worked inside the click lose value: the thin service page, the 500-word post built around a keyword, the publishing calendar measured in volume. Generic content was already weak. Now that AI can produce unlimited competent prose, it cannot differentiate anyone.
My expectation is that a growing share of a typical B2B search budget will return less each year. I cannot put an honest number on it, and I would be suspicious of anyone who does.
What compounds: being the source instead of the destination
The question survives. Buyers will keep asking which vendor handles their problem and how to evaluate options. What changes is who assembles the answer, and the machine builds it from sources it trusts. Your job becomes being the source. Four assets matter, and they are the ones practitioners focus on. Controlled studies that isolate each one do not exist yet.
Entities over keywords. The models reason about who you are: a specific company with specific offerings and people, described consistently across your site, your profiles, your structured data and the rest of the web. A company that reads as one coherent entity is easier to name.
Answers over articles. Content structured as direct, complete, quotable answers to real buyer questions is easier to retrieve and cite. Two hundred precise words beat two thousand that circle the point.
Depth over volume. One connected body of work on a specialty, with case detail and the judgment only practitioners produce, gives an engine more to cite than fifty generic posts.
Instruments over prose. Calculators, assessments and diagnostics produce the answer instead of describing it. A competitor's AI cannot generate yours.
| Loses value as clicks fall | Compounds as answers rise |
|---|---|
| Thin service and location pages | One coherent entity across site, profiles and structured data |
| 500-word posts built around a keyword | Direct, quotable answers to real buyer questions |
| Word-count arms races | Demonstrated depth in a defined specialty |
| Publishing calendars measured in volume | A connected body of work with a point of view |
| Content written for crawlers | Content written to be quoted by a machine |
| Pages that describe an answer | Instruments that produce the answer |
Figure 1: Where to move the displaced share of the budget. The left column is what many agencies still sell.
Why start before the curve is obvious
Assets on the right take sustained work to build, and the payoff arrives gradually. A company that starts while the shift is still partial has them in place when it matters. A company that waits for its traffic to visibly collapse starts from zero, while competitors that are already described clearly and consistently absorb the attention it can no longer see leaving.
The cost of being early is small, because these assets also help today. The cost of being late is invisibility that is slow and expensive to reverse. That asymmetry is the case for acting, and it holds even if the shift arrives slower than predicted.
Run the ten-minute audit yourself
Ask ChatGPT, Gemini and Perplexity the questions your buyers ask. Which vendors lead the category? How should a company like mine evaluate options? Ask each question several times, because the same prompt rarely returns the same list. SparkToro's testing of nearly 3,000 prompts found identical brand lists in fewer than 1 in 100 repeat runs, while the strongest brands kept recurring. So track how often you appear, not where you rank.
Then ask whether the description of you is accurate. That transcript is your real visibility report, and it is usually humbling. Reallocate from the content treadmill toward entity work, answer-shaped content and one real instrument.
Buyers are still searching, and many are doing it with AI before they click anything. The question is whether the machine assembling their answer knows who you are.