Car wash with 3.3 stars just won the AI answer.
When Annie Jackson, Director of Revenue and Growth Operations at GatherUp, asked Google for a touchless car wash the size of an SUV in Norfolk, VA, Google returned the business with 3.3 stars and answered approval and inquiry hours above the star rating itself. The question rating is higher than average.
That example builds on a session Jackson presented with Jason Wertham, Vice President of Security Performance Review at GatherUp: AI tools compile their own description of each location from reviews, listings, and social web chatter, and repeat that description to customers who have never accessed your website.
Jackson and Wertham went through four quick case studies that they use to reveal what ChatGPT, Google AI Overviews, and Ask Maps are saying about a multi-location product today, then building, managing, and securing solutions that change the answer.
The recap below summarizes the findings. Information, handout, and live research handout in full session. Watch it on demand.
How Are Customers Using AI to Find Local Businesses Now?
They ask a full question and accept a summary answer. In GatherUp consumer data collected in the fall of 2025, 55% of consumers spoke with Google or Bing AI abbreviations, 48% asked ChatGPT about a local business, and 31% asked multiple times.
Jackson’s car wash question shows what that looks like in practice. Instead of “a car wash near me,” he says it was “an untouched car wash for my SUV in Norfolk, VA.” Google withdrew from its 300 million locations and 500 million review participants to restore a single business, with the length of approval and 24/7 hours answered on line.
“Google answered my questions, but this business actually comes up as a 3.3 star,” Jackson said. “Given the context of my question above the star rating.”
Wertham added that these tools now look at who you are and when you ask. “The time of day when you do this query on Google Maps can affect which businesses are returned in those results,” he said. An LLM that knows you own an SUV or a large dog uses that context for all future local queries, whether you mention it again or not.
Session testing starts from this behavioral change: four prompts, from a brand name question to a site-by-site survey, that show you the answer your customers are already seeing. See the four research commands.
Does Google Updates Feed AI Answers?
Not the listing itself: Google, Yelp, and other major directories block LLM crawlers from reading review content on business profiles. Updates still support local level and list conversion; they only put AI answers in places where LLMs can crawl.
“The major directory service providers, Google, Yelp, and others, do not allow LLM tools like ChatGPT and Claude to scrape or crawl review data from business listings,” Wertham said. “You’ll notice that they don’t cite certain reviews from those platforms.”
Similar updates can be cleared when you republish them. Post them on social media channels or embed them in review widgets on your site and, in Wertham’s words, “they are now fair game for LLM tools to be pulled.”
That determines which questions you can overcome. When a customer asks for “popular” or “most reviewed” businesses, LLM searches for the review text it can access. Reviewing the contents of the directory does not contribute anything to that answer.
“If you’re relying on review platforms to do that for you, it’s not going to be enough,” Wertham said.
Action item: Republish your updates when AI can’t crawl them. The session covers which widget and social placement make review content readable, including how to conduct business feedback and review.
Wertham also breaks down first-party review capture, survey responses that don’t reach Google, and why one 11,000 customers matter. That part lives in the recording that is most needed.
Does Your Star Rating Still Matter for Search AI?
It is important under review and speed. No AI response to session audit examples yielded an average star rating; all excerpts from review content.
Consumer data points in the same direction: 45% of users prioritize recent reviews over star ratings, 60% trust detailed written reviews over rating-only reviews, and 70% prefer a review request within 72 business hours.
Wertham noted that consumers often override Google’s default “most relevant” review type and switch to “most recent,” because the latest review predicts the experience they’ll get. The maximum built in age reviews is less weighted than the current, stable stream.
“I’d rather go to a business with 1,000 reviews and a 3.9 or 4.2 review than 30 and a 5.0,” Wertham said.
The session organizes the response into three work streams, creating, managing, and protecting: creating a consistent list and review volume, managing responses and monitoring within a 72-hour response window, and protecting the rating you received against policy-violating updates and burying tactics such as review updates. Go through the structure, manage, protect the release.
Why AI Gives a Different Answer for Your Business Every Time?
Because LLM answers behave like a gambling machine, a single question is never a reliable read. Jackson cited a SparkToro study where different people asked LLMs the same question across devices and accounts, and the results didn’t come back in the same order.
“Asking an AI a question is like a slot machine,” Jackson said. “It will return the same data, but each time it will look different.”
Rank is the wrong metric for AI visibility. The total number of citations, the range of sources that provide feedback, predicts whether your product appears at all. Your brand can miss one device’s response entirely and lead the next.
Google published its guide to preparing for productive AI and updated it this month. Wertham flagged one change with teeth: the AI slop penalty. Google now finds low-value AI-generated content and, in his words, “punishes businesses” for it. Regular AI blog posts and glorified FAQ scratching posts now call you instead of being ignored.
Action item: Run your research commands in incognito or temporary chat mode so that the saved context stops shaping your results, and reuse them on a schedule. The session shows a monthly resume method for measuring whether your visibility work is delivering feedback.
Q&A: The Most Helpful Questions from the Webinar
Q: What is the fastest thing I can do this week to change what AI says about my company?
“Talk about your listings. Make sure your listings are accurate and consistent, no matter what platforms you’re on. Then make sure you’re promoting your reviews to third-party directories where you find them. Post them on your social media, post them on a section of your website.” Jason Wertham, 49:23 in a much needed record.
“Make sure you have the basics down. Get the basics down, make sure they’re set, and then you can move on to the complexities.” Annie Jackson, 50:30 on the much needed recording. Jackson illustrated this point with a local restaurant whose Facebook page listed the owner’s cell phone number; he didn’t know why the calls kept coming.
Q: How long before content changes are reflected in AI responses?
Annie replied: small facts go fast, standing is slow. Store hours and phone numbers update quickly, but “what’s known about it will take a while,” usually two weeks to a month with a longer tail than that. You identified your own website as the fastest: new offers must appear on your channels first, because reviews will not announce you. Full answer at 54:19 on the much needed recording.
Q: My weak spot has old bad reviews that keep popping up. Should I wait until they are old?
Jason answered: age fades review relevance naturally, but keyword-heavy reviews and updates from Local Guides rank longer, and emoji reactions keep the review from slipping even without high volume. Updates that violate policy are always controversial at any age; his security update team removes more than ten years of updates every time. Reliable adjustments are capacity and speed, as the latter surpasses content in long-term relationships. Full answer at 47:33 in the much needed recording.
Q: How should franchisors handle this if each franchisee controls their own profile?
Jason answered: Franchise reputation management has not been successful in a consistent space, because each franchisee has ownership of their inventory while the brand takes AI feedback. Establish best practices, provide white-label tools or partners to use, and put the playbook in their hands. He also suggested conducting audits on behalf of franchise owners and coaching them on the consequences, as a wrong AI response from one location costs the entire brand. Full answer at 55:46 in the much needed recording.
Watch the Full Webinar
The most sought-after session consists of four emergency audit tips and a downloadable manual, comprehensive design, management, release prevention, monthly assessment method, security review, policy breach review, and a gift for Jackson’s AI narrative audit. Watch the full webinar if you want.
