In short
- AI answers rely on several sources. If you only appear on your own website, you get named less often.
- For most businesses the key sources are the Google Business Profile, business directories, review sites, chambers and associations, and trade and local media.
- A Wikipedia article is within reach for only a few businesses and must not be written by the company itself. Wikidata is often the more realistic route to an unambiguous entry.
- Paid mentions without disclosure and fake reviews are banned in the EU.
Why mentions count for AI
When someone asks ChatGPT, Perplexity or Gemini for a provider, the system searches the web for suitable pages and summarises them. What counts is not only what a business writes about itself, but whether other sources confirm it: is the firm listed by its trade body, does it have reviews, has the local paper reported on it, does a trade portal mention it?
Research supports this. In a study on Generative Engine Optimization by researchers from Princeton University and other institutions, presented at KDD 2024, targeted methods increased the visibility of content in AI answers by up to 40 per cent, with differences by subject area. For businesses this means: content and source coverage can be influenced, but not with a single trick.
Which sources AI systems draw on
| Source | Examples | Why it counts |
|---|---|---|
| Profiles on large platforms | Google Business Profile, Bing Places, Apple Business Connect | Basic data such as address, opening hours, category and reviews |
| Business directories | Chamber and association directories, business listings, trade portals | Confirm that the business exists and what it offers |
| Review sites | Google reviews, industry-specific portals | Third-party experiences, often quoted directly in answers |
| Media | Local paper, trade journal, podcasts with transcript | Independent reports with high credibility |
| Knowledge bases | Wikipedia, Wikidata | Unambiguous link between name, location, industry and website |
| Forums and communities | Reddit, specialist forums, Q&A sites | Real opinions, often quoted for comparison questions |
The quickest way to see which sources really count in your industry is to ask several AI systems typical customer questions and note the linked sources. Often they are surprisingly few pages that keep coming up.
Directories: quality over quantity
It used to be: as many listings as possible. Today what counts is whether the listings are correct and appear in trustworthy directories. Here is how to proceed:
- Take stock: search for your company name, old names, old addresses and phone numbers. Note every listing.
- Align the data: name, address, phone, website, opening hours and services must be identical everywhere. Contradictions make AI systems cautious.
- Essential sources first: your chamber, trade body or professional association, the Google Business Profile, Bing Places and Apple Business Connect.
- Choose industry portals deliberately: better five portals that really are cited in your industry than fifty listings in empty directories.
- Delete duplicate and outdated listings: after a move or a name change, old data often stays online for years.
Wikipedia and Wikidata
Wikipedia is an important source for AI systems, but not a marketing channel. German-language Wikipedia only includes companies that meet certain notability criteria, for example at least 1,000 full-time employees, more than 100 million euros in annual revenue, a listing on a regulated market or a documented dominant or pioneering position. English-language Wikipedia requires significant coverage in independent, reliable sources. Most small and medium-sized businesses do not meet this, and that is no flaw.
Anyone connected to the company should not write or edit the article themselves, but propose changes on the talk page and disclose the conflict of interest. Paid contributions must be disclosed under the Wikimedia Foundation’s Terms of Use. Articles paid for in secret are regularly deleted and damage your reputation.
Wikidata is the structured knowledge base behind Wikipedia. An entry is possible if the company is a clearly identifiable entity that can be described using serious, publicly available sources, such as the companies register and reports. A clean entry with name, location, year founded, industry and official website helps machines identify your business unambiguously. Here too: factual, sourced, no advertising language.
Media and expert articles
Independent reports count especially for people and machines. Realistic routes for small and medium-sized businesses:
- Local media: anniversaries, apprenticeships, projects in the region, donations and work with local clubs are genuine occasions.
- Trade media and associations: expert articles, case reports, talks at industry events. Many associations publish contributions from members.
- Your own data: a small analysis from your daily work, for example on delivery times or price trends in your industry, is often quoted if it is clearly explained.
- Podcasts and interviews: ideally with a transcript, so the content is also available as text.
What you should avoid
- Paid articles without disclosure: editorial content that was paid for without this being recognisable counts as misleading advertising under the EU Unfair Commercial Practices Directive.
- Fake or bought reviews: they are explicitly banned in the EU, as is claiming that reviews come from real customers without taking reasonable steps to check.
- Link farms and mass directories: hundreds of listings in directories nobody reads bring nothing and can do harm.
- Self-written Wikipedia articles: they usually get noticed and are frequently deleted or tagged with warnings.
Sources
- Aggarwal et al.: GEO: Generative Engine Optimization, KDD 2024
- Wikipedia (German): Relevanzkriterien
- Wikipedia: Conflict of interest
- Wikidata: Notability
- Directive 2005/29/EC on unfair commercial practices, Annex I, as amended by Directive (EU) 2019/2161: EUR-Lex
- Google: Maps user-generated content policy, prohibited and restricted content