AI Search for Micro Businesses: The Smallest Setup That Gets You Cited by Answer Engines

Most AI-search advice is written for companies with an engineering team. You have you, maybe a partner, and a website you barely have time to update. Good news: the highest-value AI-search work for a micro business needs zero infrastructure. The semantic search and embeddings machinery the big guides obsess over is real, but for you it is the last 20 percent. The first 80 percent is being legible to the engines, and that is a content job, not a database job.

Here is the honest version. An answer engine like ChatGPT or Perplexity answers a customer's question by reading pages, breaking them into passages, and pulling the passage that best answers the question. If your page answers the question clearly in one place, you get pulled and cited. If your answer is buried in a paragraph that also talks about three other things, you get skipped. You do not need to run any of the retrieval machinery yourself to win here. You need to write like the machine reads.

#What answer engines actually read

They read structure. A question with a direct answer under it. A heading that matches how people ask. A short, self-contained block that makes sense even when it is lifted out of the page and shown on its own. That last part matters most: the engine displays your answer without the rest of your page around it, so the answer has to stand alone.

• Put the answer first. Lead with the direct answer, then explain. Do not make the reader (or the engine) dig for it.
• One question per block. A heading that is a real question, followed by a two-to-four-sentence answer, is the single most retrievable shape on the web right now.
• Say the thing plainly. "We charge 150 dollars for a standard drain cleaning in the Dallas area" gets cited. "Our transparent, value-driven pricing reflects our commitment to quality" gets skipped. Specifics get pulled. Adjectives get ignored.
• Add the facts an engine needs to trust you: your service area, your hours, your prices, your process. Concrete facts are what a retrieval system ranks highest, because they answer real questions.

#Schema markup: the one technical thing worth doing

There is exactly one piece of code-adjacent work that pays for itself at your size: structured data. It is a small block of JSON that tells engines what your business is, where you operate, and what questions you answer. FAQ schema and LocalBusiness schema are the two that matter for a micro business. Your web person can add them in an hour, or your site builder may have a toggle for them. We cover the three types that actually help in the schema markup post.

Pair the FAQ schema with a real FAQ section written in the plain, one-question-per-block shape above. The schema tells the engine "here are my answers," and the content gives it answers worth citing. That combination is most of the game for a business your size.

#When a micro business actually needs semantic search

Rarely, and that is fine. Semantic search, the kind Supabase documents with embeddings and vector columns, earns its place when you have enough content that visitors cannot find what they need by browsing. Think fifty or more pages, a real catalog, or a document library. Below that, a plain search box and good page structure beat a vector database you have to maintain.

Signs you have crossed the line: people email you questions your site already answers, but they could not find the answer. Your site search returns nothing useful. You have a body of how-to content, past projects, or products that is too big to browse. If none of that is true, skip this section with a clear conscience and put the time into the content work above.

#The cheapest real setup, if you do cross the line

If you genuinely need search over a lot of content, here is the smallest possible version. Supabase has a free tier. pgvector is built in. An open-source embedding model runs at no per-use cost. You can stand up meaningful semantic search for zero dollars a month until you have real traffic. The full mechanics are in the internal-search post, but the shape is small.

-- One table. Your content plus its embedding.
create extension if not exists vector;

create table pages (
  id bigint generated always as identity primary key,
  title text,
  url text,
  content text,
  embedding vector(384)
);

-- Find the closest matches to a question.
create index on pages using hnsw (embedding vector_cosine_ops);

That is the entire storage layer. You embed each page once with a free model (Supabase documents this exact quickstart in generating text embeddings), and you query by nearest match. No separate search service, no monthly search bill, no infrastructure to babysit. If you outgrow the free tier, the Pro tier is 25 dollars a month and covers far more than a micro business will use.

#The five questions your customers actually type

You do not have to guess what an answer engine will ask your content. Your customers already tell you, in their emails, their calls, and the search box. For a local service business, the questions are almost always the same five shapes, and each one should have a plain, self-contained answer somewhere on your site.

1. "Do you do [specific thing] in [specific place]?" Service plus area. Answer it with a sentence that names both: "Yes, we repair tankless water heaters across the north Dallas suburbs."
2. "How much does [thing] cost?" A number or a range beats "it depends" every time. "Most standard installs run 1,200 to 1,800 dollars" gets cited. Silence on price gets skipped.
3. "How fast can you come out?" Availability. "We usually schedule within two business days, and we keep same-day slots for emergencies."
4. "Are you licensed and insured?" Trust. State it plainly with the fact, not a vibe. Engines and customers both look for it.
5. "What makes you different from the other guys?" The one honest sentence you would say out loud, not a slogan. Specific beats grand.

Answer those five in plain language, each in its own block with a heading that matches the question, and you have covered the majority of what an answer engine would ever pull from a business your size. This is not a technical project. It is an afternoon of writing the truth clearly.

#Why getting cited matters more than getting ranked for you

Old-world SEO was about ranking on a page of ten blue links. The new reality is that a customer asks an engine a question and gets one answer, sometimes with two or three sources named. If you are one of those named sources, you win the customer with zero ad spend and no ranking battle. If you are not, you are invisible no matter what position you would have held on the old results page.

For a micro business, this is good news. You cannot outspend a national chain on ads or outrank them on domain authority. But you can out-answer them on the specific, local, honest questions that a chain's generic content does not touch. "Emergency drain cleaning in [your town], typical cost, same-day availability" is a question a huge competitor answers vaguely and you can answer exactly. The engine prefers the exact answer. That is your edge, and it costs nothing but clear writing. The wider shift is covered in AI Overviews and zero-click search.

One more thing in your favor: answer engines reward being specific and current, and a micro business can update a price or an availability note the same day. A national competitor's content goes through committees. Yours goes through you. Use that speed.

#How to tell if AI engines already mention you

Before you change anything, find out where you stand. You do not need a tool or a subscription. You need ten minutes and the engines themselves.

1. Open ChatGPT, Perplexity, and Google, and ask each the five questions above as a customer would, with your town in the query. See whether you appear, whether a competitor appears, or whether the answer is generic.
2. Search your own business name and note what the engines say about you. If the summary is wrong or thin, that is a content gap you can fix directly on your site.
3. Ask "who offers [your service] in [your town]" and see who gets named. If it is only directories and not businesses, there is an opening for a business that answers the question directly on its own pages.

Write down what you find. That is your baseline. Do the content work below, wait a few weeks for the engines to recrawl, and run the same ten-minute check again. The change in what the engines say about you is the only measurement that matters at your size, and it is free.

#The one-page rewrite, before and after

Here is what the change looks like in practice. Before: a services page opens with "With years of experience and a commitment to excellence, our family-owned team is dedicated to providing top-quality solutions tailored to your needs." An engine reads that and finds nothing to cite. There is no fact in it. No service, no place, no price, no answer to any question a customer would type.

After: the same page opens with "We install and repair water heaters, softeners, and whole-home filtration across [your town] and the surrounding suburbs. Standard water heater installs run 1,200 to 1,800 dollars. We schedule within two business days and keep same-day emergency slots. Licensed, bonded, and insured in [your state]." Every sentence is a fact an engine can lift and a customer can act on. The philosophy can still live below, as supporting context. But the retrievable answer comes first. That single reordering, repeated across your key pages, is most of what AI search asks of a business your size.

#How long until an engine notices

Set your expectations on the timeline, because the wait trips people up. After you publish the clearer content, engines have to recrawl your site before the change shows up in their answers. That is usually days to a few weeks, not minutes. Google's crawl is the fastest signal; the AI engines follow on their own cadence. Do not judge the work the same afternoon you ship it.

Speed it up with the basics. Make sure your sitemap is current so crawlers find the changed pages, and if you use Google Search Console, request indexing on the pages you rewrote. Then run the ten-minute check from above every couple of weeks. The change you are looking for is not a ranking number. It is the engines starting to say something more accurate and more specific about your business, in answers customers actually see. For the crawl-and-index mechanics, Search Console from zero to actionable is the fastest way in.

#Your weekend checklist

1. Rewrite your top five pages so the answer comes first and each section stands on its own. Start with the pages that describe what you do, where, and for how much.
2. Add a real FAQ section built from the actual questions customers ask you. One question per block, plain answers, real numbers.
3. Add FAQ and LocalBusiness schema markup (or turn it on if your site builder supports it).
4. Make sure your service area, hours, and contact details appear as plain text on the page, not only inside an image or a map widget.
5. Only if you have 50-plus pages and a findability problem: stand up the free-tier Supabase table above and wire a search box to it. Otherwise, stop here.

That is a real weekend of work, and it moves the needle more than most of what agencies will try to sell a business your size. The mistake micro businesses make with AI search is reaching for the complex tools first. The order is backwards. Get legible, get structured, get cited. Add the vector database only when your own content has outgrown a browse-and-find site.

If you want the content work done for you, or a second opinion on whether you have actually crossed the line into needing search, book a consultation. And if your business is bigger than you let on, the versions for small and growing teams and for agencies pick up where this one stops. The broader zero-click reality is in AI Overviews and zero-click search.