Having been involved in the digital space since the mid-1990s, I have seen a lot of “next big things” come and go. I watched the web go from hand-indexed lists of sites to search engines, from directories to Google, from desktop to mobile. But I have never seen anything as disruptive—and as full of opportunity—as AI search.
Think back to the plumber example from the beginning of this book. For the last two decades, we have all been trained to do the same thing: you see a leak in your basement, you pull out your phone, and you search Google for a plumber. Everyone from a ten-year-old to an eighty-year-old does it the exact same way. That behavior was so universal that it became the foundation of this entire book.
For the first time since the start of the internet, that behavior is changing. ChatGPT and AI search have already changed how we find information, discover new things, build our ideas, and yes—research products and businesses. Instead of typing a few keywords and scanning a page of blue links, more and more consumers are simply asking a question and getting an answer. Sometimes that answer comes from an AI assistant like ChatGPT or Gemini. Sometimes it comes from Google itself, in the form of an AI Overview sitting on top of the results your dealership used to compete for.
Now, let me be clear about something upfront, because I don’t want you to walk away from this chapter with the wrong idea: traditional search is still very important, and it is still growing. Google remains, by far, the front door of the internet—as I write this, Google still drives roughly 88 percent of all search-driven visits to websites. Everything you learned in the SEO and PPC parts of this book still matters. But AI search is changing our behaviors quickly, and the dealerships that understand it early will have an outsized advantage—just like the dealerships that took SEO seriously in 2010 or paid search seriously in 2005.
I have also seen this transformation from the inside. At SmartSites, AI—specifically Claude with MCP connections to our data—has given us a genuine breakthrough over the last year. With access to thousands of GA4, Search Console, Google Ads, Microsoft Advertising, and Meta accounts, we can now analyze data in real time at a scale that simply wasn’t possible before: finding SEO content gaps, identifying which content actually generates sales (not just traffic), discovering the ad copy and campaign structures that truly perform, and building automations that make our team faster and smarter. We are more efficient than we have ever been, and we are building things we never imagined possible. The important lesson—and a theme you’ll see throughout this chapter—is that the best results don’t come from AI alone. They come from experienced people working strategically, using AI to access and analyze all the data at their disposal. AI makes great marketers more powerful; it doesn’t replace them.
So what exactly is AI search, how does it work, and what should your dealership be doing about it? That’s what this chapter covers. And just like we did with SEO and PPC, let’s start with a little history.
The History of AI Search
While AI has been part of search for a long time (Google has used machine learning in its ranking systems since at least 2015), the AI search era truly began in November 2022, when OpenAI released ChatGPT. It became the fastest-growing consumer application in history, reaching one hundred million users in about two months. For the first time, ordinary people—not just engineers—were having conversations with an AI and getting useful answers.
Microsoft moved first among the search engines. In February 2023, they launched Bing Chat (later renamed Copilot), making Bing the first major search engine with AI built directly into the experience. As we discussed in the Beyond Google chapter, Microsoft has a long history of finding ways to compete with Google in a game of inches—and this time, they genuinely beat Google to the punch.
Google, whose entire business is built on search, was not going to sit still. In May 2023, they previewed the Search Generative Experience (SGE) in their Labs program. A year later, in May 2024, SGE graduated and rolled out to everyone as AI Overviews—the AI-written answer box that now appears at the top of roughly half of all Google searches. In 2025, Google went a step further and launched AI Mode, a fully conversational search experience that lives in its own tab right inside Google, letting users ask follow-up questions the way they would with ChatGPT.
Then the advertising followed the eyeballs, the way it always does. Through 2025 and into 2026, Google began showing ads inside AI Overviews and AI Mode. In February 2026, Google launched shopping ads inside AI Mode, and OpenAI began piloting ads inside ChatGPT. (We’ll cover the advertising side later in this chapter and in the Beyond Google chapter.)
If that timeline feels fast, that’s because it was. Google spent twenty-five years perfecting the ten-blue-links experience. AI search rewrote it in about three. It is the biggest change to search since Google started selling ads on search results—and just like then, most businesses are reacting too slowly.
AI Overviews, AI Mode, and AI Assistants
”AI search” gets thrown around loosely these days, so let’s get precise. There are really three different things people mean when they say it, and your dealership needs to understand all three:
AI Overviews. This is the AI-written summary that appears at the top of a regular Google search, above the traditional results. It synthesizes an answer from multiple sources and—this is the important part—cites those sources with links. AI Overviews now appear on roughly 48 percent of Google searches in the US. Your customer is still “Googling,” but the first thing they see is an answer, not a list of links.
AI Mode. Google’s fully conversational search tab. Instead of one search, the user has a conversation: “What’s the best three-row SUV under $50,000?” followed by “Which of those holds its value best?” followed by “Who has one near me?” AI Mode is almost entirely zero-click—studies put it above 90 percent—meaning the user typically gets everything they need without ever visiting a website. And the walls between these experiences are coming down: in mid-2026, Google began letting users hand off from an AI Overview straight into an AI Mode conversation, and started adding AI agents that can track ongoing tasks and send notifications. Expect the line between “a Google search” and “a conversation with Google” to keep blurring.
AI assistants. ChatGPT, Gemini, Microsoft Copilot, Perplexity, Claude. These are destinations in their own right—apps and websites that consumers open instead of Google. Research shows that around one in four new-car buyers now uses an AI tool at some point during their vehicle research, and among those AI users, ChatGPT dominates with roughly a two-thirds share.
AI Search and Traditional Search Play Different Positions
Here’s a nuance that most of the panicked industry coverage misses, and understanding it will save you from both overreacting and underreacting: consumers don’t use AI search the same way they use traditional search. They use them at different stages of the journey, for different jobs.
Watch how differently people talk to each one. The average Google search is still three to four words—“toyota dealer near me”—because twenty years of Google trained us all to think in keywords. The average ChatGPT prompt runs around 23 words, often much longer, because it’s a conversation: “I have two kids and a long commute, what’s the most reliable three-row SUV under $40,000?” Long, personal, exploratory questions are research questions. Short, decisive queries are action queries. That difference tells you exactly where each tool lives in the funnel.
The data backs this up from every direction. Semrush’s ongoing AI Overviews research found that in early 2025, over 91 percent of queries triggering an AI Overview were informational; commercial research queries have been growing fast since (the share of commercial searches showing an AI Overview grew 71 percent over six months), while transactional searches showing AI Overviews actually declined. In automotive specifically, Ekho’s 2026 vehicle research study found 30 percent of buyers used generative AI to research their vehicle—more than twice as many as used online marketplaces—but they still completed the journey on dealer and retailer websites. And remember EMARKETER’s finding from earlier: 95 percent of AI platform-driven purchases still finish on a brand or retailer site. Our own dealer client data tells the same story from the click side: informational click-through collapsed while branded, local, and shopping click-through rose.
Put simply: AI search is where your customer forms the shortlist. Traditional search is where they act on it. The research phase—“what should I buy?”—is moving into conversations with AI. The action phase—“where do I get it?”—still leans heavily on classic search, maps, and your website.
That’s why your strategy needs both halves of this book. If you’re invisible in AI answers, you’re not in the consideration set when the shortlist gets formed—you’ve lost before the customer ever types your brand into Google. But if you win the AI mention and neglect your traditional search presence, your website, or your VLAs, you lose the handoff—the customer AI sent your way gets intercepted by the competitor who still owns the bottom of the funnel. AI search sets the table; traditional search serves the meal.
Here’s the conceptual shift that makes all of this matter, and I want you to remember this sentence even if you forget everything else in this chapter: traditional search ranks pages; AI search synthesizes answers and cites sources. For twenty years, the game was to rank #1. The new game is to be retrieved, trusted, and cited—to be inside the answer, not below it.
GEO, AEO, and the New Vocabulary
Whenever an industry shifts, a new alphabet soup follows. Let me demystify the two acronyms you’ll hear most, because you will absolutely hear them from agencies pitching your dealership:
GEO (Generative Engine Optimization). The practice of optimizing your website and your broader web presence so that generative AI engines—AI Overviews, AI Mode, ChatGPT, Gemini, Copilot, Perplexity—include and cite your business in their answers. If SEO is about ranking in the list, GEO is about being part of the answer.
AEO (Answer Engine Optimization). An older cousin of GEO that predates the ChatGPT era—it originally referred to optimizing for featured snippets and voice assistants (“Alexa, who’s the best Honda dealer near me?”). Today the terms are used almost interchangeably, and honestly, the distinction matters less than the work itself. Some practitioners also say “AI visibility” or “LLM optimization.” It’s all describing the same goal.
The good news—and it’s a big one—is that GEO is not a replacement for SEO. It’s a layer on top of it. The AI engines learn about your dealership from the same web that Google crawls: your website, your Google Business Profile, your reviews, the articles and lists that mention you. Almost everything you invested in through Part 1 of this book—unique content, authoritative links, technical health, a well-maintained GBP, a strong review profile—is exactly what the AI engines feed on. Dealerships that did SEO right for years have a massive head start in AI search. Dealerships that cut corners are about to have that exposed in a whole new set of places.
How the AI Engines Decide What to Say
Just like we pulled back the curtain on the Google algorithm in Part 1, let’s demystify how these AI tools actually produce an answer—because once you understand it, everything I recommend in this chapter will feel obvious rather than mysterious.
An AI assistant like ChatGPT or Gemini is a large language model (LLM)—a system trained on a massive snapshot of the internet. Ask it something timeless (“what’s the difference between a CVT and an automatic?”) and it answers from that training. But ask about anything current or local—inventory, prices, “dealers near me”—and the model does something different: it runs searches behind the scenes, reads the results, and builds its answer from what it finds, citing its sources. The industry calls this retrieval-augmented generation. You don’t need to remember the term; you need to remember what it means: for the questions that matter to your dealership, the AI is reading the live web—your website, your Google Business Profile, your reviews, articles about you—and paraphrasing what it finds.
The industry has a name for that behind-the-scenes step: grounding—connecting the model to live web data and trusted sources so its answers are accurate and current—and can say where a fact came from. Microsoft has leaned into this harder than anyone: Bing’s search index serves as the factual foundation that grounds Copilot and many other AI experiences that license Microsoft’s data. Remember that, because it tells you where your leverage is. You can’t optimize the model—but you can absolutely optimize the sources it grounds against: your website, your business profiles, your reviews. (It’s also one more reason not to ignore Bing: be well represented in Bing’s index and you are well represented in every AI experience it grounds. More on that in the measurement section and in Useful Tools.)
That leads to three practical truths:
First, the AI’s picture of your dealership is a mirror of your web presence. If your hours are inconsistent across the web, the AI may state them wrong. If your reviews mention pushy sales tactics, the AI may repeat that to a shopper who never reads a single review. If you have no dedicated page describing your commercial truck department, the AI has almost nothing to retrieve when someone asks who services commercial fleets nearby.
Second, AI systems make mistakes—confidently. The industry calls these hallucinations: the AI states something false as fact, like a dealership being permanently closed or carrying a brand it doesn’t. You can’t call OpenAI to complain. What you can do is control the sources the AI reads—which is what the GEO playbook later in this chapter is for.
Third, you can’t keyword-match a conversation. In the SEO part of this book, we talked about targeting keywords—and for classic search, that still works. But SparkToro’s research found that when 142 different people were asked to write a prompt with the same shopping intent, barely any two prompts looked alike. People get creative, specific, and personal with AI (“my wife hates road noise and we have a golden retriever—what SUV should we look at?”). Nobody can build a keyword list for that. What the AI does is extract the underlying intent and retrieve content that covers it. The practical shift: stop writing pages for exact phrases, and start covering topics and situations—the questions, use cases, and comparisons a real buyer brings—thoroughly enough that whatever words they use, your content is the best match.
The Machines Are Signing Their Work: AI Watermarking
One more how-it-works development belongs here, because it quietly raises the stakes of the content conversation we started back in Part 1. In the summer of 2026, the major AI companies began watermarking what their models write. The European Union’s transparency code for AI-generated content took effect in July 2026 with roughly 190 companies signed on; Google had already been embedding a statistical watermark (called SynthID) in Gemini’s text since 2024, and Anthropic announced in August 2026 that Claude’s output would carry a similar one, with the rest of the industry expected to follow.
A text watermark has nothing to do with a visible label. When a model writes, there are thousands of small moments where several words would fit equally well; a watermarked model makes those choices in a subtle statistical pattern that a detection tool can later recognize. You can’t see it, a reader can’t feel it, and light editing won’t fully remove it—though a genuine rewrite in your own words will.
What does this mean for your dealership’s marketing? Be skeptical of anyone who claims to know for certain. As I write this, no search engine has said it demotes content simply because a machine wrote it, and Google’s official position remains that it rewards helpful content however it’s produced. But follow the incentives. The AI engines are in the business of serving original, trustworthy information, and a page of unedited AI output adds nothing an AI engine doesn’t already have. Google’s spam systems already target mass-produced, low-value content at scale. Watermarking hands every platform a cheap, reliable way to spot exactly that.
So the practical advice hasn’t changed since Part 1—the stakes have simply gone up. Use AI in your content process the way we use it at SmartSites: to research, to analyze, to draft. Then make sure what you publish contains what no model can generate on its own—your real prices, your real inventory, your customers’ actual words—finished and stood behind by a person who knows your store. Pages built that way have nothing to fear from detection, because they aren’t pretending to be something they’re not.