Consumer use of AI tools may have stopped growing, even as professionals keep adopting more of them. That is the central finding in new desktop-usage research from Datos and SparkToro, analyzed by SparkToro cofounder Rand Fishkin and reported by Search Engine Land on Thursday. The gap between the two groups is wide enough to explain why AI feels universal inside marketing circles while behaving like a niche tool in the broader population.
The clearest signal is ChatGPT’s trajectory on desktop. Datos measured ChatGPT reaching roughly 37 percent of United States desktop users in September 2025, then slipping to about 34 percent by March 2026. A six-month decline is not a collapse, but it is the opposite of the steep adoption curve the product showed through 2024 and early 2025. The plateau holds across the EU and UK as well, where desktop usage runs about ten percentage points higher than the United States but follows the same flattening shape.
The professional picture looks different. Claude, the assistant from Anthropic, posted four straight months of growth from December through March and showed a 373 percent lift among business-to-business professionals compared with the average United States population. ChatGPT, by contrast, was 15 percent less likely to appear among retail-shopping consumers than among the average American. Claude did not rank in the top four AI tools for that consumer retail audience at all.
That divergence is the part SEO teams should sit with. Search and content professionals spend their working hours inside AI tools and inside LinkedIn feeds where AI is the dominant subject. Both surfaces overrepresent professional behavior. A practitioner extrapolating from personal experience will badly overestimate how many of their actual customers are running product research through a chatbot rather than through a search box.
The strategic error this enables is a premature reallocation of budget. A team that believes consumer AI adoption is still climbing fast will move spend toward generative engine optimization, the practice of optimizing for LLM-driven answers, and away from classic ranking work. If real consumer demand is flat near a third of desktop users, that reallocation overshoots. The traditional search box is still where most non-professional buyers start, and it has not lost the share that AI-first narratives often imply.
The opposite mistake is just as available. For products that sell to professionals, especially business-to-business software and services, the data argues the other way. An audience showing a 373 percent lift in Claude usage is plausibly running vendor comparisons and shortlists through assistants right now. Visibility inside those answers is a present concern for B2B sellers, not a future one, and underinvesting there cedes ground while the behavior is still forming.
A measurement caution applies to the research itself. Datos tracks desktop activity, and a meaningful share of consumer AI use happens inside mobile apps that desktop panels do not capture. The plateau is real for the surface measured. It is not proof that total consumer AI use has stopped growing across every device. Fishkin’s analysis is a corrective to the inflated professional read, not a declaration that consumer adoption is finished.
The actionable split is clean. Audit who actually buys from you before setting the AI-search budget. If the buyer is a consumer shopper, keep classic search fundamentals funded and treat generative-answer optimization as a measured experiment. If the buyer is a professional, move faster on answer-engine visibility, because that audience is already there.
ATTRIBUTION: Reported by Search Engine Land on 2026-05-21.
