Prompt-stopping

The Lead

AI answers are picking the big guys. Let’s talk through what smaller teams can do about it.

We read two studies that came out this week, both have a lot they can lend to automotive.

The first, from Vaer AI on behalf of Lightspeed Commerce, ran 20,000 shopping prompts through ChatGPT, Google AI Mode, and AI Overviews. With web search off, national chains took 63 to 70 percent of recommendations. Small and independent stores showed up in about 10 percent, even though they were roughly 38 percent of the stores the models mentioned. Turning search on helped, but chains still took about half of the top picks.

The second, from NP Digital, looked at 9 million AI answers across nine platforms and more than 400 brands. On bottom-of-funnel commercial prompts, the "who should I buy from" questions, 82 percent of citations went to third-party sites. (from my own anecdotal research you’d get Carvana or Carmax). Brands' own pages got 3 percent.

Neither study looked at car dealers. But the pattern maps onto our business, and Denise Galiatsatos called it on our podcast, which we’ll release in the coming weeks, before either one was published. She said the old goal was a thumb-stopping ad, and the new one is being the answer when someone asks ChatGPT or Gemini. Carson named it on the spot: prompt-stopping.

Here's what the data says about getting there.

1. "Local" doesn't save you

   The most useful finding in the Lightspeed study: adding "local" or "near me" to a prompt barely changed results, because the models counted a nearby branch of a national chain as local. Adding "independent" more than doubled small-store picks, from about a third to nearly four-fifths.

   Translate that to auto. A shopper asking for a Honda dealer near them can easily get sent to the nearest big-group store, or a national used-car retailer with a lot in town. What moves the result is how clearly you describe yourself.

   If you're a single-point or family-owned store, say so in plain text where models read: your About page, Google Business Profile, VDP footer, and third-party listings. Include the owner's name, years in the community, and the town. Put it in words, not in a logo or banner image.

2. Specific questions favor the big guys

   The same study found that specific product requests favored large retailers even more than broad ones. That describes every serious car shopper. Nobody asks for "an SUV." They ask for a 2023 RAV4 XLE Premium under 30,000 miles.

   Big groups and national retailers win those prompts because their inventory data is structured and everywhere. Craig Smith from Visor.vin told us on the show about King Ranch F-150s listed as XLTs. If your VDP does that, no model will match you to the shopper looking for a King Ranch. Your spec data is now your ad copy.

3. Other people are writing your AI answer

   82 percent third-party means the answer about your store is mostly built from what others say about you: review sites, YouTube, Reddit, local press, and marketplace listings. In NP Digital's data, YouTube gave the biggest visibility lift of any third-party source at 2.8x, ahead of LinkedIn and Reddit.

   That's good news for dealers. Walkarounds and buying-process explainers are the easiest content you can make. Post them to YouTube with full transcripts, and say the store name out loud.

4. Your own pages pay off most when they land

   When a brand's own page did get cited, its visibility rate jumped 5x, and owned citations stuck around longer than third-party ones. The catch: up to 75 percent of answers that cited a brand's page never said the brand's name.

   That's the VDP problem in one number. If your page answers the question but your name isn't in the part that answers it, the model uses your work and credits nobody. Put the store name in the sentences that answer real questions: what's included, what the warranty covers, and how the out-the-door process works.

What to do this month

• Write your store identity in plain text on the places models read most.
• Pick 10 prompts a shopper in your market would actually ask. Run them in ChatGPT, Gemini, and Google AI Mode, log who shows up, and repeat monthly.
• Build a GA4 segment for AI referrals. Gemini just started adding UTM tags to its links, so that traffic is finally measurable.
• Audit 20 VDPs for trim and option accuracy.
• Shoot three YouTube walkarounds with transcripts.

Your website is still the home game. AI answers are an away game on someone else's field, and right now the big groups have the advantage there too.

Sources

• ChatGPT and Google AI favor large retailers in shopping answers
• What 9 million AI answers reveal about search visibility

Borrowed: Cakes Body built a media company instead of buying more ads

Cakes Body is a beauty brand founded in 2022 by twin sisters Taylor Capuano and Casey Sarai. It hit $100 million in sales in three years, mostly on founder-led TikTok rather than paid ads.

Now that paid social keeps getting more expensive, they've launched Cakes Media, an in-house content arm run by a former Red Bull content executive and funded out of the brand marketing budget. The first release was a 57-minute YouTube documentary that has passed 371,000 views. Every episode gets clipped into short-form for TikTok and Reels. Capuano says the goal is to "operationalize virality," meaning a repeatable content system instead of a lucky hit.

Here's what a dealer can take from it. One long piece of content, filmed once, can feed a month of short posts. Most stores do the reverse: they scramble for a new clip every week and run out of ideas by week three.

The dealer version: follow one car from trade-in to delivery.

Pick one interesting trade and film it at each step:

• the appraisal and the number offered
• what the tech found in inspection
• what recon cost and why
• the photo shoot
• the pricing decision
• the hand-off to the next owner

Edit it into a 20 to 30 minute YouTube video, then cut 10 to 12 clips from it. That's three posts a week for a month. You need a phone, a used car manager who's willing to talk straight about numbers, and an afternoon of editing.

This lines up with what Denise Galiatsatos said on our show. The dealership videos that work feature the store's own people, not hired influencers. Shoppers rarely see how a car gets made ready, and that look behind the curtain is what builds trust.

It also connects to this week's Lead. YouTube was the top third-party citation source in AI answers. A full-length video with a transcript and your store name in it gets you on social and in AI answers at the same time.

What to measure

• views per clip
• visits to the featured car's VDP
• how many shoppers mention the video in the CRM source field

Source: Beauty brand Cakes wants to turn its TikTok virality into an owned media strategy (Digiday)

Signals

An Alaska independent dealer agreed to a $500,000 settlement for leaving its $499 doc fee out of its advertised online prices. More than 5,000 buyers going back to 2018 get $100 each. Check your VDPs, listing feeds, and inventory ads this week, before a state attorney general checks them for you… Auto Remarketing

Gemini has started adding UTM tags to its outbound links. Google's John Mueller confirmed he's seeing it too, but Google hasn't documented it. Gemini app traffic has mostly landed in GA4 as "direct" until now. Build an AI-referral segment in GA4 this week so you have a baseline before it changes again. Search Engine Journal

In Stephens' Q3 dealer sentiment survey, F&I (+34% net) and parts and service (+22% net) were the biggest upside surprises. Traffic (-38% net) and new-vehicle gross (-29% net) were the biggest misses. 46% of dealers called September their worst recent month for new gross. If the back end is carrying the store, Q4 budget belongs in service retention and equity mining, not just conquest. CDG News