When should a local business use AI in its marketing?

The short answer? Use it backstage. Avoid it onstage.

Anything a customer never sees is fair game: the first draft you rewrite, the inbox you're behind on, the pile of reviews you’ve been meaning to read. Anything a customer does see, especially images, is where AI starts costing you money. The question was never how much AI you use. It’s whether anyone can tell.

That distinction got more expensive this year, and there are numbers behind it.

The owners aren’t wrong about the budget

If you’ve spent any time in the local Facebook groups lately, you’ve watched a version of this argument. Someone posts a flyer with a suspiciously glossy plate of food. Someone else says it’s AI. The owner replies that they’re running on nothing and can’t pay a designer for a Saturday event.

That answer is honest. It’s also completely ordinary. U.S. Bank surveyed 1,000 small business owners this spring and found 75% of them using generative AI in the business. The U.S. Chamber of Commerce has tracked the climb: 23% in 2023, 40% in 2024, 58% in 2025. Marketing is the most common use, not a side use. In the U.S. Bank data, 56% use it for marketing and sales and 51% for content creation. Business.com’s 2026 survey put the average time recovered at 5.6 hours a week, rising to 7.2 for managers.

So the person defending themselves in the comments is describing the median American small business. They aren’t cutting a corner nobody else is cutting.

The customers aren’t wrong either

Here’s what doesn’t come up in those threads.

Fractl surveyed 1,008 U.S. consumers in the second quarter of 2026 and asked whether heavy AI use would change how much they trusted a brand they already liked. Forty percent said their trust would go down. The same question a year earlier got 20%. It doubled in twelve months.

The penalty isn’t spread evenly. Fifty-four percent of Gen Z said their trust would drop, against 33% of Gen X and 32% of boomers. Women were more likely to penalize it than men, 44% to 34%.

A separate study points the same direction. Klaviyo surveyed 8,000 consumers across eight countries in December 2025 and found that when AI is visible in a brand’s marketing, only 7% say it makes them trust that brand more. Thirty-two percent say it makes them trust the brand less. The rest are neutral, which is its own answer: nobody is being won over by it.

Nobody in that argument is being unreasonable. Both sides are describing something true.

The line is who sees it

Sort every AI use into one of two piles.

Backstage. The customer never sees the output, only the result of you having more time. This is where AI is close to free money:

• First drafts of captions, emails, and event descriptions that you then tweak or rewrite in your own words

• Reading through a year of reviews and telling you what people keep mentioning

• Replying to routine emails so you’re not doing it at eleven at night

• Working out what to say before you say it

• Naming, brainstorming, and every version of “help me think about this”

Onstage. The customer looks directly at the thing AI made. This is where the 32% lives:

• Generated images of your food, your storefront, your staff, or your products

• Written reviews or testimonials of any kind

• Photographs of a thing that does not exist

One rule covers most of it: never generate an image of something a customer could walk in and compare against reality. A rendered burger you don’t actually serve is the fastest way to turn a curious person into a suspicious one, and suspicion doesn’t stay contained to the flyer. It spreads to the menu, the reviews, and the prices.

What to do instead, on no budget

The flyer problem is real. The solution is less expensive than either side of the argument assumes.

Your phone camera beats any generated image, and it costs nothing. Shoot the actual plate, near a window, sometime between ten and two, against the plainest background you own. Take thirty frames and keep one. That photograph is worth more than a perfect render precisely because it’s a little imperfect. It proves the thing exists.

Then put your own photo into a free Canva template rather than a generated background. Pick one typeface and two colors and use them on everything for a year. Consistency is most of what people read as “designed.” Most local businesses lose more to using six fonts across six flyers than they’d ever lose to a modest template.

If you want real design help, the region has more of it than people realize. Virginia Highlands Community College in Abingdon, ETSU in Johnson City, and Emory & Henry in Emory turn out students who need portfolio work and charge accordingly. A standing arrangement with one student for a season costs a fraction of an agency and often produces something better than a one-off freelance job, because they learn your business.

And use AI for the words. Write the caption with it, rewrite it in your own voice, shoot the picture yourself. That hybrid is nearly undetectable, and it’s honest.

Does any of this affect whether AI recommends you?

Indirectly, and not the way people assume.

Assistants like ChatGPT and Perplexity don’t inspect your flyer and dock you for it. What they read is your Google Business Profile, your reviews, your website, and what other people have published about you. But a business that customers have quietly stopped trusting writes fewer new reviews into existence, gets mentioned less, and gives those systems less to work with. The trust problem shows up later, in the record the assistants are reading.

The businesses being recommended by name in this region are the ones with a current profile, real reviews, and a site that plainly says what they do and where they do it. None of that requires a design budget.

Common questions

Will people really be able to tell?

With images, more often than businesses expect, and the tell is usually a small thing: hands, text on a sign, a reflection that doesn't work, food that looks too symmetrical. With writing, people are less able to identify it specifically and more likely to just find it boring, which costs you the same thing.

What about an AI-generated logo?

Different situation, and less risky than a generated photo, because a logo isn’t claiming to document reality. The problem is ownership, and it isn’t a gray area anymore. A purely AI-generated image with no human author can’t be copyrighted in the United States. The D.C. Circuit ruled that the Copyright Act requires human authorship, and the Supreme Court declined to hear the appeal in March 2026. You’d likely still build trademark rights through using the mark in business, and trademark is the protection that matters most for a logo. But you’d be giving away the copyright layer for nothing. For a mark you intend to keep for years, paying a person once is worth it. We’re not lawyers, and if the mark is central to your business, ask one.

We already used AI images. Did we damage something permanently?

No. Replace them with real photographs as you go and don’t make an announcement about it. Nobody is keeping a file on you. The businesses that get hurt are the ones that keep doing it after customers start commenting.

Curious what ChatGPT, Google’s AI, and Perplexity currently say about your business? We’ll check and send you what we find, free: AI Visibility Check

Sources

All figures below were checked against primary sources on July 25, 2026.

- U.S. Bank, Small Business Perspective Survey (4th annual, released June 18, 2026). 1,000 U.S. small business owners, fielded Feb 27 – Mar 17, 2026, margin of error ±3.1%. 75% using generative AI; marketing and sales 56%; content creation 51%. [usbank.com/SBOsurvey.html](https://www.usbank.com/SBOsurvey.html)

- U.S. Chamber of Commerce / C_TEC + Teneo Research, Empowering Small Business: The Impact of Technology on U.S. Small Business (Aug 2025). Generative AI use: 23% (2023), 40% (2024), 58% (2025). [uschamber.com](https://www.uschamber.com/technology/empowering-small-business-the-impact-of-technology-on-u-s-small-business)

- Business.com, 2026 Small Business AI Outlook. 5.6 hours saved per week on average; 7.2 for managers. [business.com](https://www.business.com/articles/ai-usage-smb-workplace-study/)

- Fractl, Q2 2026 survey of 1,008 U.S. consumers and 150 marketers. Trust decrease from heavy brand AI use: 20% (2025) → 40% (2026); Gen Z 54%, Gen X 33%, boomers 32%; women 44% / men 34%. [frac.tl/ai-statistics](https://www.frac.tl/ai-statistics/)

- Klaviyo with Datalily, 2026 AI Consumer Trends (March 2026). 8,000 consumers across eight countries, fielded December 2025. Visible AI in marketing: 7% trust more, 32% trust less, remainder neutral. [klaviyo.com](https://www.klaviyo.com/solutions/ai/consumer-trust-in-ai)

- Thaler v. Perlmutter, D.C. Cir. (Mar. 2025); cert. denied, U.S. Supreme Court, March 2, 2026. Human authorship required for U.S. copyright.