How to get your business cited by ChatGPT, Gemini and Perplexity

How AI assistants choose what to cite

When someone asks ChatGPT, Gemini or Perplexity a question that needs current or specific information, the assistant usually runs a web search behind the scenes, reads a selection of results and writes an answer with links to the pages it used. Those links are the citations. Questions about local services, prices, suppliers and recommendations nearly always trigger this kind of search.

Two things decide whether you appear. First, your page must be in the search index the assistant uses and must rank well enough to be among the results it reads. Second, your page must contain a clear, trustworthy passage that answers the question, so the assistant can use it. A good page that is hard to find, or an easy-to-find page with a vague answer, will rarely be cited.

There is also a second, slower route. Large language models learn from vast amounts of text during training, and businesses that are mentioned often and consistently across the web become part of what the model “knows”. That shapes which names it suggests even before it searches. This broader work is what we call GEO.

Where each assistant gets its sources

The details change often and the companies do not publish everything, but the broad picture is reasonably clear:

The practical lesson is that two indexes, Google and Bing, sit behind a large part of AI search. Many businesses in Denmark have never opened Bing Webmaster Tools, yet Bing visibility affects Copilot and, at least in part, ChatGPT. Registering the site there and submitting a sitemap takes a few minutes.

Step one: let the right crawlers in

Check your robots.txt file and any firewall or bot protection on your hosting. Some security plugins and content delivery networks block unfamiliar bots by default, which can quietly keep AI search crawlers out.

It helps to understand the difference between crawlers for search and crawlers for training:

• GPTBot collects content that may be used to train OpenAI’s models. Blocking it does not, according to OpenAI, stop your pages appearing in ChatGPT search results.
• OAI-SearchBot is the one that makes your pages eligible for ChatGPT search answers. Blocking it can keep you out of them.
• Google-Extended is a control token, not a separate crawler. It lets you opt out of some uses of your content in Gemini; Google says it does not affect inclusion in Google Search.
• PerplexityBot indexes pages for Perplexity’s answers.

Whether to allow training is a business decision. Whether to allow the search crawlers is simpler: if you want to be cited, let them in.

Step two: publish pages worth quoting

Assistants cite passages that answer a question cleanly. The pages that get quoted most often tend to share a few features:

1. One clear topic per page. A page about wedding photography in Copenhagen should not also try to cover corporate headshots and product shoots.
2. A direct opening. The first two or three sentences say what the service is, who it is for, where it is offered and, if you can, what it typically costs or how long it takes.
3. Questions as headings. Use the wording buyers type or say, then answer underneath in a self-contained way.
4. Concrete facts. Areas served, languages, lead times, what is included and what is not. Assistants prefer specific statements to marketing language.
5. Visible authorship and contact details. Who runs the business, where it is registered, how to reach it.
6. Fresh information. Out-of-date prices or services are a reason for an assistant to choose another source.

Structured data (Organization, LocalBusiness, Service, FAQPage) supports this by stating facts in a machine-readable form. Our citation-ready pages service builds pages to this pattern.

Common mistakes that keep businesses out of answers

A handful of problems come up again and again when we look at why a capable business is never mentioned:

• Services hidden in PDFs or images. Price lists and menus uploaded as images or scanned documents are hard for crawlers to read. Put the key facts in normal page text.
• Everything on one page. A single “services” page listing twenty offers in one line each gives an assistant nothing substantial to quote about any of them.
• Content that only loads with JavaScript. Some AI crawlers do not run scripts as fully as Googlebot, so text that appears only after scripts run may be invisible to them.
• Chasing new file formats before the basics. Proposals such as an llms.txt file get attention, but support among the major assistants is limited and uncertain. Clear, indexable pages matter far more.

Fixing these is rarely glamorous, but it often does more for citations than any new content.

Step three: be the same business everywhere

An assistant is more confident recommending a business when the facts about it agree across many sources. If your website says one address, your Google Business Profile another and an old directory listing a third, you look less reliable than a competitor whose details match everywhere.

For a Danish business, the places to check include:

• Your CVR registration and the details shown on virk.dk.
• Google Business Profile and Bing Places.
• Business directories such as Krak and Proff.
• Review platforms such as Trustpilot, and any industry-specific review sites.
• LinkedIn, Facebook and Instagram profiles.
• Membership listings in trade associations and chambers of commerce.

Use the same business name, address, phone number, description of services and website link in each place. Describe what you do in similar words, so the pattern is easy to recognise.

Step four: earn mentions on other sites

Your own website can say anything. Independent sources carry more weight, both for search rankings and for the assistants that read them. When a journalist, a partner, a trade body or a respected blogger writes about your business, it confirms that you exist and are relevant.

Useful, realistic sources of mentions include local and trade media, guest articles for industry publications, partner and supplier pages, event and conference listings, podcasts, and “best of” or comparison articles written by others. Comparison and list articles appear especially often in assistant answers to recommendation questions, so it is worth knowing which ones exist for your field and whether you are included.

Avoid paid link schemes and fake reviews. They breach Google’s guidelines, Danish marketing law restricts fake reviews, and they can damage the trust you are trying to build.

Step five: check and track your citations

There is no complete dashboard for AI citations yet, so measurement is partly manual:

• Keep a list of 15 to 30 questions your buyers ask, in Danish and English, and ask them in each assistant every month. Note whether you are mentioned, whether you are linked, and which competitors appear.
• In your analytics, look for referral traffic from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. ChatGPT often adds a utm_source=chatgpt.com tag to links it shows, which makes those visits easier to spot.
• Ask new enquiries how they found you, and include AI assistants as an option.

Answers vary between users, sessions and days, so judge trends over months rather than reacting to a single response. If you want a structured starting point, our search team runs these checks as part of AEO and GEO work.