# From the U.S. Auto Retail AI Tracker: The North Star Metric for AI Marketing—IPS

-01- U.S. Car Buyers Are Quietly Changing the Way They Buy Cars

Cox Automotive released its latest *AI in Auto Retail Tracker* at the end of July 2026 (Q2 2026 data, covering 1,502 consumers intending to buy a vehicle within the next 12 months and 483 dealers). The one line in this report that should make every marketer pause is:

**63% of purchase intenders plan to use AI to complete their next vehicle purchase.**

This is not an edge behavior of “young people trying something new”; it is a structural shift in the mainstream consumer journey. What is even more interesting is *how* users are using AI—they are not using AI to bypass dealers, but to move toward dealers more effectively:

• 26% of AI users use it to generate “a list of questions to ask the salesperson once in the dealership”;
• 24% say AI makes them “better prepared” when they arrive at the store;
• Only 17% list “reducing interaction with salespeople” as a benefit—the lowest-scoring option of all.

This means that the customer walking into the showroom tomorrow has already done their homework. AI has not eliminated human touchpoints; it has reshaped the “cognitive pre-stage” before those touchpoints. Consumers sort out their needs, compare models, and filter configurations in a conversational interface, then walk into the physical world with a “mental checklist.”

This is precisely the judgment that industry research has repeatedly emphasized: what generative AI changes is not a single-point tool, but the very entry point of the consumer decision journey.

As Professors Tan Beiping and Jin Liyin point out in their book *Generation: AI Productivity Reshapes the New Marketing Paradigm*—when “recommendation” happens in a chat box, the brand’s main battlefield shifts from the search engine results page to the answers of large models.

-02- A Huge Gap on the Industry Side: 82% Use AI, 71% Still Don’t Understand AI Search

In sharp contrast to the rapid migration on the consumer side is the sluggishness on the supply side.

The same report shows that AI adoption among U.S. dealers has reached 82%, with automating routine tasks, customer follow-up coordination, and content generation as the three mainstream use cases. It looks like “everyone is using AI.” But look at another set of numbers:

• Only 29% of dealers have begun adjusting for “AI-driven search/recommendation”;
• 32% admit they “know they should act but haven’t started”;
• 26% say they “aren’t clear what this means”;
• 23% feel “this isn’t on my to-do list.”

In other words, 71% of dealers have not yet taken the first step in response to “consumers using AI to choose cars.”

And anxiety is spreading: 52% of dealers feel they are behind their peers in AI adoption (up from 47% in Q1). The report’s wording is precise—“not because they are actually behind, but because the goalposts keep moving and the noise keeps getting louder.”

Even more worth pondering is the “scissors gap” between expectations and reality:

• Driving sales/revenue growth: expected 69%, actually experienced 22%;
• Improving employee efficiency: expected 54%, actually experienced 26%;
• Improving profit: expected 51%, actually experienced 9%.

ROI and cost concerns (45%) and accuracy concerns (45%) tie as the biggest obstacles. The industry’s questioning has matured from “Should we use AI?” to—“Please prove it works for the business.”

-03- The Marketing Science Perspective: The Essence of All AI Anxiety Is a “Measurement Gap”

The most easily underestimated set of data in the report is about “measurement”:

• 41% of dealers cannot say clearly whether AI has met expectations (up from 37% in Q1—it is getting worse);
• One-third of AI users either do not measure AI’s effects at all or cannot explain how they measure them.

This is the core diagnosis from a marketing science perspective: most of today’s anxiety, doubt, and wavering around AI is not essentially a technology problem, but a measurement problem.

The same data contains the best counterevidence. The report singles out “data-driven decision-making” dealers for comparison:

• Using metrics to evaluate AI effectiveness: 71% of data-driven dealers vs. 46% of others;
• Believing AI is critical to long-term success: 79% vs. 56%;
• Actually achieving revenue growth from AI: 34% vs. 18%.

The gap is nearly double. This is not a coincidence, but another validation of the most basic law of marketing science: what gets measured gets managed; what gets managed can be turned into growth. Data-driven players who “measure first, optimize second” are not only more confident in their understanding, but also actually capture more revenue in results.

For auto brands, this law has a brand-new projection in the AI era: when consumer decisions move upstream into the chat windows of large models, what brands need to measure is no longer just “how people see you,” but also “how AI recommends you.”

-04- IPS: The North Star Metric for AI Marketing

This is precisely the original purpose of the IPSmoto project (www.ipsmoto.com).

If the first principle of marketing science is “everything begins with measurement,” then in the AI era, the first measurement question to answer is: when consumers ask AI “Which car should I buy?”, is your brand recommended? And how is it recommended? We call the quantified answer to this question IPS (Intelligent Promotion Score).

IPSmoto’s monitoring system has several key design elements:

• Three IPS indicators: IPS (Intelligent Promotion Score), VS (Visibility Score), and RS (Ranking Score), characterizing the three levels of a brand in AI answers—“being mentioned, being seen, and being ranked near the top”;
• Nine core automotive attributes: cabin, intelligent driving, safety, range, space, performance, and other dimensions, breaking down the “reason structure” behind AI recommendations;
• 16 global markets (China, the U.S., Japan, Germany, the U.K., France, etc.) × 11 mainstream large models (GPT, Claude, Gemini, DeepSeek, Tongyi, Doubao, Kimi, Wenxin, Perplexity, etc.) with 24/7 continuous monitoring;
• 500+ brands and 7,000+ models, with high-frequency prompt testing that simulates real car-buying intent across age groups such as young adults, families, business professionals, and older buyers;
• AI Citation Index: tracking the authoritative media sources that major models deeply trust—because AI’s “trust” is essentially a citation weighting of the content ecosystem.

IPSmoto advocates a “dual-metric system of IPS + NPS”: NPS measures human users’ willingness to recommend, the North Star of brand health for the past two decades; IPS measures AI’s willingness to recommend, the new North Star for brand growth in the next decade. Human feedback and AI feedback together form a complete closed loop for brand equity measurement.

Back to the U.S. market data: 63% of car buyers will use AI to choose cars, and AI recommendations are by no means neutral—they can be measured, attributed, and optimized. Whoever establishes a baseline IPS measurement first will be the first to obtain the “control group” for marketing science experiments in the AI era.

-05- Three Action Recommendations for Brand Owners

Combining the insights from the Cox report with IPSmoto’s monitoring practice, we recommend that Chinese auto brands (especially those going global) immediately do three things:

1. Put “AI recommendation performance” into the regular dashboard. Just as you monitor public sentiment and search indices, regularly measure the IPS of your brand and models across markets and models. Without a baseline, all optimization is shooting in the dark.
2. Manage the “source assets” of AI trust. Cox data shows that 68% of dealers believe “tools with built-in AI are more valuable,” and 63% trust embedded AI more—trust is transmitted in layers. For brands, the authoritative media revealed by the AI Citation Index are the “opinion leaders” of the AI era. Whether your content appears in sources deeply trusted by AI matters more than how many impressions you buy in determining how AI evaluates you.
3. Respond to “Is AI worth it?” with data. Do not become one of the 41% of practitioners who cannot explain AI’s effects. Set measurable goals for every AI marketing action—whether it is a change in IPS, growth in AI citations, or the share of dealership leads from customers who used AI to assist their decisions. The tradition of marketing science has not changed: hypothesis, measurement, attribution, iteration—only the testing ground has moved from shelves and search pages into the chat windows of large models.

Conclusion

One line in the Cox report is worth remembering: “The dealers getting real value from AI are not treating AI as a separate strategic department, but as an iteration of their existing work.”

The same is true of marketing science. AI has not created a new essence of marketing; it has simply made “being recommended” a scalable object of measurement for the first time. 63% of U.S. car buyers have already stepped into AI’s chat window—is your brand in AI’s answer?

This is the question IPS is designed to answer.

Data source: Cox Automotive, *AI in Auto Retail Tracker*, Q1 2026 (n=504 dealers) & Q2 2026 (n=483 dealers; n=1,502 vehicle intenders), July 2026.