If you run a small or medium-sized firm with no in-house IT, buyers reach you in a different way than they used to. Instead of scanning ten blue links, they ask ChatGPT or Google AI Overviews things like "find a reliable accountant near Hamburg that works with tradespeople" or "who can build a booking system for a small rental firm?"
A good Google ranking still helps, but it's no longer the whole game. You also need to be readable and recommendable by AI assistants. That combination is SEO and GEO: classic search optimization plus generative engine optimization, sometimes called AEO (answer engine optimization). This guide explains how to show up in ChatGPT search, how AI search picks businesses to mention, and what to change on a typical small business website.
Why a Google ranking alone is no longer enough
When someone asks an assistant for a recommendation, it reads a handful of sources and writes one short answer naming two or three firms. Plenty of people stop at that answer and never scroll further. If your business isn't among the sources the assistant trusts and understands, you miss out even if your old ranking was decent.
This hits smaller firms hardest. A plumber in Dortmund or a dental practice in Leipzig rarely has a content team. The site may be five pages, last touched two years ago. That used to hold its own against bigger rivals well enough. For AI search it's usually too thin: there is nothing clear for the model to quote.
AI tools favour pages that state plainly who you serve, what you do, where you work and how to start, so a clearer site helps more than a bigger one, and clarity costs less than volume.
How ChatGPT search and AI Overviews choose sources
Think of an AI assistant as a busy receptionist. It won't recommend a supplier it cannot describe in a sentence or two. It looks for direct answers, consistent facts, and signs the information is current and written by a real business.
What the model actually reads
An assistant does not "know" your firm. It retrieves pages, snippets, listings, and sometimes machine-readable files, then summarizes them. It pays most attention to short factual statements about services and service area, headings that match the question asked, question-and-answer blocks it can quote, and consistent contact details across your site and listings.
A paragraph about "passion for excellence" gives it nothing. A sentence like "We build booking and invoicing web apps for rental firms in Germany with 10 to 80 vehicles" gives it everything.
What makes it trust one firm over another
Nobody outside the search companies knows the exact formula, and anyone promising a guaranteed spot in AI answers is overselling. In practice, assistants prefer sources that are specific, consistent, and easy to verify.
Keep facts identical everywhere. If your homepage claims one service area, your contact page another, and your directory listing a third, the assistant has to guess, and may well recommend a competitor whose pages agree with each other. A few concrete lines about quotes, lead times, or handover also show you are a real operation. Plain, checkable facts win.
The practical setup: SEO and GEO for a small business site
Most of the work is tidying what you already have so Google and AI tools can read it. Below is the setup we recommend for a typical site of five to twenty pages. It is usually days to a few weeks of tidying, not a rebuild.
1. Give every service one clear page
One page per service beats one long "services" page. A builder might split extensions, loft conversions, and repairs. Each page should answer five questions near the top: who is this for, what problem does it solve, what is included and excluded, where is it offered, and what is the next step?
Write headings as customer questions. "What does a replacement for our Excel tracker include?" beats "Our approach". End with one call to action, such as a quote form or booking link, and link up to the matching main service page.
2. Add FAQ blocks that answer buying questions
FAQs are a cheap way to give an assistant a clear answer to quote. FAQ markup no longer earns rich results for most businesses, but it still helps machines read your answers. Write four to six per service page in the words buyers use: how long it takes, what you need from them, how ownership of the result and data is handled, what leaving looks like, and roughly what it costs.
Answer each in two to four sentences and be honest about limits. A short "good fit if / not a fit if" note saves everyone time and reads as trustworthy. Keep the same format across the site and mention your town or trade once, naturally.
3. Make the site readable for AI crawlers (llms.txt, schema, sitemap)
This is the part owners hand to whoever looks after the website. If you do not have anyone yet, our SEO, AEO and GEO work covers exactly these items. The technical work is small and removes guesswork for crawlers: one stable URL per page with an XML sitemap submitted to Search Console, basic structured data for the business and FAQs, an up-to-date robots.txt, and an llms.txt file, a proposed, unofficial plain-text index of your key pages that some AI tools may read.
Think of llms.txt as a table of contents for models. It does not guarantee inclusion, and whether AI tools use it is still open, but it lists your preferred pages in one place. This rarely needs a rebuild.
Which AI crawlers to allow in robots.txt
Your robots.txt decides which crawlers may fetch your pages. Common AI-related user agents include OAI-SearchBot (ChatGPT search), GPTBot (OpenAI model training), PerplexityBot, Google-Extended (Gemini, not AI Overviews) and CCBot (Common Crawl). To be cited in ChatGPT search, allow the search crawler; blocking training crawlers is a separate choice. Vendor policies change often, so check the current crawler documentation before you publish, and ask whoever maintains the site to review the file with you.
Our own setup as an example
We run SolutionPlus the way we recommend to clients. SolutionPlus UG (haftungsbeschränkt) in Berlin is the contract partner, project management works in German and English, and our engineers in Pakistan do the build. The SEO, AEO and GEO work we offer clients is described on our SEO, AEO and GEO page.
Our site follows the pattern above: one page per service framed around a buyer problem, short FAQ blocks in a consistent format, stable URLs with sitemap and machine-readable files including llms.txt, and one primary call to action per page. Clear pages first, clever tooling only where it earns its keep.
Checklist: is your site ready for AI search?
Start here:
- Each main service has its own page with audience, scope, area, and next step
- Headings written as customer questions, not internal labels
- Four to six short FAQs per service page with direct answers
- Name, phone, email, and area match across site and listings
- Quote or contact path works on a phone in under a minute
- Sitemap submitted and robots.txt checked
- Basic structured data for business and FAQs
- llms.txt listing your best pages
- One primary call to action per page
- Content reviewed regularly so facts stay current
If you tick fewer than half, you'd be better off starting with structure and FAQs, as they carry most of the weight. The cost depends on page count, how much rewriting is needed and the state of the site, so we give a written estimate after a quick look at it (Get my estimate gives a first idea). For full rebuild pricing, see our website development cost guide.
Who this setup is not for
This setup is not for every site. If you sell mainly through marketplaces, if your pages change daily by feed, or if you need guaranteed placement in ChatGPT answers, spend the budget elsewhere first. It fits best when buyers choose a supplier by reading a few clear pages, which is how most local services and small B2B firms win work.
How to measure visibility in AI answers
You cannot track AI answers in Search Console alone. Run a small monthly prompt panel instead: ask the same five to ten buying questions in ChatGPT search and Google AI Overviews and note when your firm is named or cited. In analytics, watch referral traffic from chatgpt.com and perplexity.ai alongside brand mentions, and treat all three as directional signals rather than exact ranks.
What to do next
Pick your most profitable service, rewrite that one page using the structure above, add its FAQ, and check it on a phone. Then repeat for the next service.
If you would rather hand it over, our SEO, AEO and GEO work applies the same setup we use ourselves. Send my request with your site address and your main service. Reply within one business day.
Frequently asked questions
What is GEO, and how is it different from SEO?
SEO helps pages rank in classic results. GEO, sometimes called AEO (answer engine optimization), helps AI assistants understand, trust and quote your pages when they write an answer. Clear structure, FAQs and consistent facts help both.
Do I need llms.txt on a small business website?
It may help, but whether AI tools read it is still open. The file is a proposed, unofficial plain-text index of your key pages. Add it after your service pages and FAQs are clear, not before.
How long before changes show up in AI answers?
There is no fixed timetable, and no honest supplier can promise placement in ChatGPT answers. Tools recrawl at their own pace. Clear, consistent pages tend to be picked up faster.
Can I do this without an in-house IT team?
Yes. Most of the work is wording and structure an owner can approve, and the technical items are short tasks for whoever maintains the site. If you hire help, ask for contract, IP and code ownership terms in writing before work starts.
Can you guarantee that we will appear in ChatGPT answers?
No, and neither can anyone else. Nobody outside the search companies knows the exact selection formula, so anyone promising a guaranteed spot is overselling. What you can do is make your pages specific, consistent and easy to verify, which improves your chances.




