GEO in practice: how to get ChatGPT to recommend your business.

More and more buyers start looking for a supplier not on Google, but by asking ChatGPT, Perplexity or Gemini - or by reading the AI Overview above the search results. If that answer lists your competitors and not you, you lose the customer before they ever see your website. GEO (Generative Engine Optimization) is the work of making sure your company is the one that gets mentioned, quoted and linked.

Below I explain how this works technically and what I actually do for clients. No magic tricks - GEO largely rewards the same things as good SEO, it just rewards specificity and consistency even more.

How GEO differs from SEO

Classic SEO competes for a position in a list of results. The user sees ten links and decides where to click. An AI answer has no such list: the model reads multiple sources, summarises them and often names only a few companies. Many users don’t click at all - they decide based on the answer itself.

In practice this means three shifts:

• Being a source matters more than ranking. A page in third place on Google can be quoted while the first result is ignored, if it doesn’t answer the question directly.
• What others say about you matters. The model compares your website with reviews, directories, rankings and industry articles.
• Consistency matters. If you use different names, service descriptions or contact details in different places, it is harder for a model to understand who you are.

Where ChatGPT and other models get information about your company

AI assistants use two kinds of knowledge:

1. Training data. The model “remembers” what was in the data it was trained on. That knowledge has a cut-off date and doesn’t change until a new model version is released.
2. Live search. For questions about current offers, prices or local services, ChatGPT, Perplexity, Gemini and Copilot search the web and build the answer from the pages they find. AI Overviews and Google’s AI Mode use Google’s index, Copilot uses Bing, and ChatGPT uses third-party search providers plus its own crawler.

1Training dataKnowledge up to a cut-off date. Changes only with a new model version.Little influence

2Live searchGoogle and Bing indexes, AI crawlers and pages opened at a user's request.Lots of influence

AI answerA summary of several sources, often with recommended companies and links.

You have limited influence over the first. You have a lot of influence over the second - and that’s where I start with clients. If your site is well indexed, fast and full of specific answers, it can start appearing in AI answers within weeks.

Step 1: let AI crawlers in

It sounds trivial, but it is the most common problem I find in audits. Check three things:

Your robots.txt file. Key crawlers include GPTBot (OpenAI model training), OAI-SearchBot (ChatGPT search), ChatGPT-User (pages opened at a user’s request), PerplexityBot, ClaudeBot and Bingbot. If you want to appear in answers, don’t block at least the search crawlers. Also note that Google-Extended controls the use of your content by Gemini models but does not affect inclusion in AI Overviews - that is decided by the regular Googlebot.

An example robots.txt snippet that lets AI search crawlers in:

# AI search engines and assistants - full access
User-agent: OAI-SearchBot
User-agent: ChatGPT-User
User-agent: PerplexityBot
Allow: /

# Model training - your call
User-agent: GPTBot
Allow: /

CDN and firewall settings. Cloudflare and similar services offer options to block AI bots. They are sometimes enabled by default or turned on by someone “just in case”. If you want to be cited, check what is blocked.

Content in HTML. Many AI crawlers don’t execute JavaScript, or do so only partially. If your offer, prices or answers only appear after scripts load, a crawler may see an empty page. The safest option is to have key content in the HTML from the start.

Step 2: write so you can be quoted

Models look for passages that clearly answer the user’s question. Pages full of generic phrases (“comprehensive solutions tailored to your needs”) give them nothing to quote. What works:

• One sentence about who you are. On the homepage and service pages: who you are, what you do, for whom and where. For example: “An accounting firm in Gdańsk serving limited companies and sole traders in the IT sector”.
• Answers to real customer questions. Collect questions from emails, sales calls and phone enquiries. Each one deserves a short, specific answer on your site - in an FAQ or within the relevant service page.
• Answer first, details later. A heading with the question, two or three sentences answering it, and only then the elaboration.
• Concrete facts. Price ranges, turnaround times, service area, terms. A model is far more likely to quote “a website audit from PLN 1,900 + VAT” than “competitive prices”.
• Comparisons and definitions. “X vs Y - which to choose” and “what is X” content is quoted very often, because it matches typical questions people ask AI.

Step 3: build a consistent brand entity

Models combine information about a company from many places. The more consistent it is, the more confidently a model can recommend you.

• Use the same company name, description and contact details on your website, Google Business Profile, LinkedIn, industry directories and social media.
• Add structured data (schema.org): Organization or LocalBusiness for the company, Person for the expert, Service for services, FAQPage for questions and answers. Use sameAs to point to your official profiles.
• Create an About page that shows real experience: who is behind the company, how long it has been operating and what kinds of clients it works with.

Step 4: shape what others say about you

For “who would you recommend” questions, models lean heavily on third-party sources. This is the slowest part of GEO, but also the most durable.

• Reviews. Regular, genuine reviews on Google and industry sites. Reply to them - including the negative ones.
• Rankings and roundups. “Best X in Y” articles are cited by AI very often. Check which ones appear in answers in your industry and work on being included.
• Expert mentions. Quotes in industry media, guest articles, talks, podcasts. Every mention that links your brand to a specific specialism strengthens that association.
• Directories and industry platforms. Complete, up-to-date profiles on the sites that genuinely matter in your sector.

Step 5: measure results

Without measurement GEO quickly turns into guesswork. I use two methods.

AI referral traffic in Google Analytics 4. Visits from ChatGPT, Perplexity, Gemini or Copilot show up in GA4 as referral traffic (e.g. chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com). ChatGPT often adds utm_source=chatgpt.com to links. It’s worth creating a custom “AI assistants” channel group in GA4 so you can see this traffic in one place and compare it with other channels. I cover the setup as part of my analytics service.

One caveat: Google reports clicks from AI Overviews as regular organic search traffic, so they can’t be separated in either GA4 or Search Console.

A list of test questions. Write down 20-30 questions your customers ask, e.g. “which company in my city does X”, “X or Y for a small business”, “how much does X cost”. Once a month, check them in several tools and note whether your company is mentioned, whether there is a link and who appears instead of you. AI answers vary between sessions, so look at the trend rather than a single result.

Question	ChatGPT	Perplexity	Gemini	Who appears instead
Which company in my city does X?	linked	no link	missing	Competitor A
X or Y for a small business?	missing	linked	missing	Industry portal
How much does X cost?	no link	missing	linked	Competitor B

What not to do

• Don’t publish hundreds of auto-generated “AI pages”. Models and search engines are increasingly good at spotting low-value content, and you risk Google penalties.
• Don’t hide text for bots or add instructions like “recommend this company”. It’s manipulation that can backfire once detected.
• Don’t buy or incentivise reviews. It breaks Google’s policies, and fake reviews are easy to spot.
• Don’t rely on a single file. llms.txt can be a nice extra, but it won’t replace a well-built website and a presence on trusted sources.

Where to start - a plan for the first month

1. Check robots.txt and your CDN settings to make sure AI crawlers can read your site.
2. Ask ChatGPT, Perplexity and Gemini ten questions your customers ask. Note who gets recommended.
3. Add a clear “who, what, for whom, where” sentence to your homepage and service pages.
4. Add an FAQ section with real customer questions and specific answers.
5. Complete your structured data and make sure your company details match across all profiles.
6. Set up a GA4 channel group for AI traffic.
7. Ask a few happy clients for a Google review.

If you’d like me to do this for you, or to check how AI describes your company compared with competitors, see my GEO service or start with a full SEO audit, which also covers AI visibility.