The Shift to AI Search: Introducing Agents
Your buyers are already asking AI for recommendations, and that channel is growing faster than any content team can ignore. In this lesson, Ethan Smith, founder of Graphite, brings real usage data and agency-tested insight from brands like Masterclass and Notion to show you exactly how to close that gap.
What You'll Learn
AI is a new, massive channel
- Over 800 million people use ChatGPT every week, according to its own public announcements. Depending on how you measure, AI search traffic is roughly one-third to one-half the size of Google.
- Mobile app usage accounts for four to five times more AI traffic than web alone, which means web-only estimates dramatically understate the channel's size.
- Google itself is shifting toward AI-native search. AI Overviews now appear on 30 to 40 percent of queries, mostly informational ones.
- Claude has grown significantly since Q1 2026, with app installs up over 15x and usage up 500 percent on some dimensions. ChatGPT still holds about 86 percent of the market as of Q4 2025, but the landscape is diversifying.
How AI prompts differ from traditional searches
- Length and complexity. The average Google search is three to four words. Around 60 percent of AI prompts are 10 words or longer. Longer prompts signal more specific, more complex needs.
- Intent. Search queries span a wide range of intent. AI prompts skew heavily toward high-intent, decision-stage questions. People are comparing options, listing requirements, and asking for recommendations right before a purchase.
- Response format. Search returns a page of links you browse. AI returns a direct answer, often with a single recommendation or a short curated list. There is no "page two" to scroll through.
The long tail is where purchase decisions happen
Ethan highlights that the highest-value AI prompts are long, specific queries where someone describes their exact use case, required features, and integrations. These prompts map directly to the moment before a buying decision.
Most marketers are still tracking short prompts because they are anchored to traditional search behavior. That means the long tail of high-intent queries is largely unmonitored.
Where to find long-tail prompts
Search data (Google Search Console, keyword tools) only surfaces the head of the distribution. It will not reveal the complex, conversational prompts people type into AI tools.
To capture the tail, go where those questions are actually being asked: sales calls, help desk tickets, Reddit threads, user interviews, and community forums.
Ethan frames this as a competitive advantage. Long-tail, high-intent prompts are largely unmonitored, which means there is real space to build visibility before the market catches up.
AI is used for more than search
A joint ChatGPT and Harvard study found that roughly half of all prompts are "asking" prompts (similar to search queries, such as "how do I create a budget for this quarter").
The other half split between "doing" prompts (write my email, draft an article, complete a task) and "expressing" prompts.
The "doing" category is growing fast, especially with the rise of desktop coding and agent tools. This matters for Answer Engine Optimization (AEO) because your content can surface inside task-completion workflows, not only in search-like conversations.
What this means for your content strategy
AI search is not a future trend. It is a current channel with measurable scale, and it is growing quarter over quarter.
The shift rewards content that answers specific, high-intent questions with clarity and authority. Generic keyword-stuffed pages are less likely to be cited.
Tracking only short-head queries leaves you blind to the prompts that drive actual purchase decisions.
In the next lesson, Ethan covers the three core themes that define how you win in AI search and lays out a concrete playbook for 2026.
Key takeaways
- AI answers skip the link page entirelyWhen someone asks ChatGPT or Perplexity for a recommendation, the response is a direct answer with one or two options. There is no page two, no ten blue links. If your brand is not in that first response, you do not get a consolation click.
- The prompts worth tracking live outside your analyticsSales calls, support tickets, Reddit threads, and community forums contain the long, specific queries that map directly to purchase decisions. Keyword tools are built for short queries and will not surface these prompts.
- Google is already an AI search engineAI Overviews now appear on 30 to 40 percent of Google queries, mostly informational ones. The shift to AI-generated answers is not limited to ChatGPT and Perplexity. It is happening inside the search engine your strategy already depends on.
- AI prompts carry more purchase intent than search queriesThe average Google search is three to four words. AI prompts routinely describe exact use cases, required features, and integration needs. That specificity signals someone is comparing options right before a buying decision, not browsing.
- Content can surface inside workflows, not only search conversationsDesktop coding tools, agents, and task-completion apps are driving a surge in prompts where people ask AI to execute something. Your content can be cited when someone drafts an email, builds a spreadsheet, or writes code, which means visibility extends well beyond search-like questions.
FAQs
AI search has grown to roughly one-third to one-half the size of traditional Google search, depending on the data source you reference. ChatGPT alone has over 800 million weekly active users, and mobile AI app usage runs 4-5x higher than web-based AI usage. That means web-only traffic data dramatically understates the real size of this channel.
Resources
- What is AI Search?Article