/ai-search840 words

AI Overviews on Commercial Queries Grew 71%, Semrush Data Shows

Commercial keywords triggering AI Overviews grew 71% in five months, Semrush found, as zero-click searches rose 34.5% in a CMU experiment and brand mentions diverged from citations.

The future of SEO: Where search is heading in 2027
The future of SEO: Where search is heading in 2027AS_Photography / pixabay
  • Commercial-intent keywords triggering AI Overviews grew 71% on average from November 2025 to April 2026 (Semrush, 600,000+ U.S. keywords).
  • A CMU/Indian School of Business 2026 experiment found AI Overviews raised zero-click searches 34.5% and cut organic clicks 39.8%.
  • In a 3,981-domain Semrush study, only 13.2% of AI appearances paired a citation with a brand mention.
  • ChatGPT has 1.2 billion weekly users; Google AI Mode has passed 1 billion monthly users.
  • Google AI Mode queries average 7.22 words versus 4.0 for regular searches.

By Elena Vasquez

The share of commercial-intent keywords triggering Google AI Overviews grew by an average of 71% between November 2025 and April 2026, according to a Semrush study of more than 600,000 U.S. keywords. That single figure anchors Semrush's new forecast on where search — and SEO — is heading in 2027.

The trend is confirmed at the feature level. Google has extended AI Overviews from informational queries to commercial ones, and now auto-expands some answers without a "Show more" click, with follow-up fields that carry users into AI Mode. A page can rank near the top of organic results yet never appear in the AI Overview's cited sources. The two channels are separate. Semrush says marketers must track both.

What does AI answering do to clicks?

A 2026 field experiment by Carnegie Mellon University and the Indian School of Business measured it directly. Displaying AI Overviews increased zero-click searches by 34.5% and cut outbound organic clicks by 39.8% compared with hiding them.

Those numbers deserve scrutiny. The study measures behavior when the feature is toggled on and off, not long-run business outcomes. Semrush itself concedes the gap: a buyer may see a brand in results, never click, and later search for it directly or shortlist it. Traffic alone no longer captures SEO's effect. Impressions, branded search volume and conversions become the working proxies. Each proxy is indirect. None proves causation.

Where do buyers actually research now?

A Semrush survey of 1,030 U.S. consumers who had tried AI tools found 77% use AI alongside search engines. Only 4% relied mostly on AI. Some 33% started with search, 26% started with AI, and 18% switched between the two throughout their research.

Semrush brands the response "search everywhere optimization": build visibility on the platforms that matter, not on every platform. The phrasing is vendor framing from a company selling visibility tools. The underlying multi-channel behavior — Google, YouTube, Reddit, ChatGPT, then vendor sites — is what the survey data actually supports.

How visible is your brand inside AI answers?

The scale is real. ChatGPT reports 1.2 billion weekly users; Google's AI Mode has passed 1 billion monthly users. Semrush studied 3,981 domain appearances across ChatGPT, Gemini, AI Overviews and AI Mode. The breakdown: 61.7% were citations without a brand mention, 25.1% were mentions without citations, and only 13.2% included both.

Read that carefully. An AI platform can cite your page while recommending a competitor's product. Citations, mentions and recommendations do not travel together. This is the strongest argument in the piece for generative engine optimization (GEO) and answer engine optimization (AEO) — and the easiest metric for vendors to oversell.

Are keywords still the unit of demand?

Maybe not for long. In a Semrush analysis of ChatGPT prompts, 65%–85% of prompts in most months matched no keyword in its database of more than 27 billion keywords. The study included non-search interactions, a caveat Semrush itself flags. Query fan-out compounds the shift: AI systems decompose one question into many subtopic searches, rewarding topical coverage over isolated keyword targets.

What does 2027 look like?

Three developments frame the forecast. First, more interactive search: Google AI Mode queries average 7.22 words versus 4.0 words for regular searches, and ChatGPT prompts per session rose 50% over the four months ending February 2026. Multimodal input — Google Search Live, Copilot Vision, image and voice queries — extends visibility opportunities beyond the first response.

Second, agents. ChatGPT's shopping research feature and Google's Universal Commerce Protocol (UCP) already let agents compare products, check inventory and fill carts. Adoption lags the plumbing: a 2026 Forrester survey found 75% of consumers uncomfortable with autonomous agent purchases, though Semrush's survey found 22% had already bought inside an AI tool.

Third, skills. Semrush analyzed 3,900 U.S. SEO job listings in November 2025: 31% of senior roles mentioned AI versus 22.3% of other roles, and 6.3% of senior listings named AEO or SGE explicitly. Project management (31%), communication (26.5%) and A/B testing (23.9%) led senior skill requirements.

What stays fixed?

Semrush lists three constants: understand the human need, inform the machine, influence the customer journey. Google has stated that existing SEO best practices apply to AI Overviews and AI Mode without special optimizations — a confirmed engine statement, unlike the tool-based visibility studies above. Indexed pages remain a prerequisite for AI Overview links. Inaccessible pages limit what ChatGPT, Claude and Perplexity can retrieve.

The report closes with five priorities: fundamentals, topic research, original content, brand authority, and measurement beyond rankings. Much of the measurement advice routes through Semrush's own Position Tracking, Site Audit and AI Visibility Toolkit. The recommendations are conventional; the packaging is commercial.

What to monitor next

Watch the citation-versus-mention gap, the share of purchases completed inside AI tools, and whether AI Mode query length keeps climbing. Those three numbers will tell you whether 2027 arrives on schedule.

via papers.ssrn.com (Original)

More from Elena Vasquez

Elena Vasquez

Show full bio

Correspondent covering people, launches and day-to-day practice in Search rankings and algorithm updates, generative engine optimization at SERP Journal.

221 articles