What changes for a small business?

The cost of producing something that looks finished has dropped sharply. A page of copy, a logo concept or a working layout can now appear in minutes. That changes who can start, and it changes what a professional is being paid to do.

Drafts get cheaper

A first draft is no longer the expensive part. A business owner can get an outline, ten headline options or a rough FAQ before lunch. Use that. A rough draft helps you notice what you actually want to say, and it makes conversations with a designer or writer more specific.

What remains expensive is deciding what is true, what fits the business, and what to leave out. Those are judgment calls, and a tool can't be held to them.

Sameness rises

When everyone asks similar tools similar questions, they get similar answers. You have probably started to notice the pattern: the same soft promises, the same three-part headings, the same glossy stock-style images. The result isn't wrong. It just could belong to any business in any city.

A quick test: cover your logo and read your homepage. If a competitor could publish it unchanged, the page isn't doing enough work.

Original knowledge matters more

The content that holds its value is the content a model doesn't have: the question customers ask you every week, the mistake you've watched people make, what your work costs and why. A caterer knows how much notice a lunch order needs. A salon knows how long a color appointment takes. Write those things down, because that is where the difference between you and a generic answer lives.

Search answers change how people find you

People now get summarized answers inside search, sometimes before they ever see a list of links. Google's AI Overviews and AI Mode show links to help people explore a topic, and AI Mode may show a wider set of them. A person might meet your business as a cited link in an answer rather than a blue link they chose from a list.

For site owners, Google's guidance is less dramatic than the headlines. It says there are no extra requirements or special optimizations for these features, no new files, and no special schema. A page has to be indexed and eligible for a snippet. Clicks and impressions from these features are counted in the Web search type of the Search Console Performance report, not in a separate one.

So the sensible plan is the plain one: keep important content in text, link your pages to each other, keep structured data consistent with what the page says, and keep your Business Profile current. We cover how we approach this on our AI search optimization page.

Chat assistants move onto websites

A trained assistant can now answer routine questions on your own site at any hour. We offer AI chatbot training as an add-on: it is trained on your business knowledge, FAQs and policies, tuned to your tone, and can hand leads to your email or CRM.

Before adding one, ask who is responsible when it gets something wrong. A chatbot on your site speaks for your business. Give it a narrow job, feed it answers you have approved, and make it easy to reach a person.

What doesn't change?

People still buy from businesses they trust, and trust still comes from things that are hard to fake. AI made polish cheap, so polish alone now tells a visitor very little.

  • Trust. Named clients, real addresses, live work and awards with a source link still do what they always did.
  • Verification. Models predict likely words. They don't look facts up in the way people assume, so they can state something false with total confidence. Your hours, prices, credentials and legal claims need a human check every time.
  • Real photos. A photo of your actual team, room or work is evidence. An AI image is an illustration, and it can set the wrong expectation for someone who then walks through your door. The Audiology Center of Burbank site handles this well: its footer says which photographs show the real team and office and which imagery is illustrative.
  • Accountability. When a page gives bad advice or a chatbot makes a promise, a customer isn't going to blame the software. A named person at your business has to own what gets published.

What have we built with AI ourselves?

We have built two AI products of our own, and both raise questions a good demo doesn't answer. Neither one is client work, and we have no results to quote from them.

Ask Burbank is our own AI-powered guide to Burbank, covering local businesses, events and city services, with an AI concierge named Robbie. It's currently in private preview, so we make no claims about how it performs with the public. Building a guide to a real city puts the verification question front and center, because local details change and a confident wrong answer sends someone to the wrong place.

Hacobian Financials is an in-house household-finance app with an AI advisor. The advisor is a chat that sends the household's questions to Anthropic's Claude API. That single sentence raises the question every business should ask of any AI feature: what data leaves your system, where does it go, and have the people involved agreed to it? The app is private, which is also why we don't publish a screenshot or a live link.

If you're considering something similar, our web app development page explains how we scope custom tools, including AI-enabled ones.

What does Google say about AI-generated content?

Google treats AI-assisted content under the same rules as everything else, and it stresses review. Its guidance on generative AI content says these tools can be useful for research and for structuring original content, and it warns that producing many pages without adding value may violate its spam policy on scaled content abuse.

Three more points from that page are worth acting on:

  • Review everything. It calls for fact-checking AI output before publishing, and it extends that review to titles, meta descriptions, structured data and image alt text.
  • Tell readers how it was made. It suggests explaining how automatically generated content was created, because that gives readers useful context.
  • Stores have extra rules. For ecommerce, AI-generated images need IPTC metadata marking them as AI-made, and AI-generated product titles and descriptions must be labeled under Merchant Center policies.

A simple AI policy for a small business

You can cover most of the risk with five rules on one page. Write them down, even if your whole team is you.

  1. A person checks every fact. Prices, hours, dates, names, statistics and anything medical, legal or financial get verified against a real source before publishing.
  2. No invented proof. Never let a tool write testimonials, client results or statistics. If it offers a number, find where the number came from or delete it.
  3. Label what readers would want to know. Mark AI-generated images and illustrations as such, and say so when a page is substantially automated.
  4. Add what only you know. Before anything goes live, add a real example, a real photo or an opinion you would defend.
  5. Protect private data. Don't paste customer details into a tool until you've read how it handles them.

Written by Romik Hacobian, founder of Media City Design, a Burbank web design studio since 2021.

Sources