Visible in AI Answers: What Changes for Company Websites

What changes when customers ask an AI instead of a search engine?

A buyer is looking for a supplier of sheet metal parts in small batches. In the past he would have used a search engine, seen ten results and opened three websites. Today he asks an AI assistant and gets an answer with four names, short descriptions and perhaps two links. If your business is not among the four, this buyer has never seen your website.

Two overlapping developments are at work. Kenny and Pogrebna (2026) describe them in Harvard Business Review like this: language models such as ChatGPT or Copilot are replacing the search engine for some questions, and sources are cited only selectively or not at all. At the same time, search engines themselves place AI summaries above the results list, which draw visitors away from websites. Both reduce effort for the searcher and increase it for businesses.

For owners, the sober question is not whether this is happening but what it means for their own sales. The article SEO for small and midsize businesses has already described the basic rule: what a search engine understands well, an assistant understands well too. So the question is: what does a small business need to do beyond good content for search engines in order to be named in AI answers? To answer it, we first look at who gets named in AI answers, then introduce three stages of visibility, and go through your own website, evidence on other sites, and measurement. At the end comes the question of how urgent the topic is for a small business.

In an AI answer there is no second page. Either you are named, or you do not exist for that question.

Who gets named in AI answers, and who does not?

Specialists call the effort to be visible in such answers Generative Engine Optimization, or GEO. The term goes back to a paper by Aggarwal and colleagues (2024). They built a large collection of search queries from various fields and tested how changes to web content affect its visibility in AI answers. Certain changes increased visibility by up to 40 percent, but the effect varied strongly between fields.

How unevenly visibility is distributed is shown by an analysis by Kumar (2026). It examined more than 100,000 answers from AI search engines about more than 100 brands between March and May 2026. It is a preprint that has not yet gone through peer review, and the data come from a provider of visibility measurement. The orders of magnitude are revealing nonetheless.

Two further findings matter for small businesses. When the systems cited sources, about 78 percent were company websites. Your own website therefore remains the most important source. And the most frequently cited format was best-of lists and comparison overviews, at around 21 percent of all citations. If you are missing from such lists, you are often missing from the answer too. Finally, it is striking how unstable the framing was: whether a brand was portrayed positively or negatively changed about 6.7 times more often than whether it was mentioned at all.

Found, understood, recommended: three stages of visibility

For an assistant to name your business, it must be able to do three things. I arrange them in three stages: found, understood, recommended. Each stage builds on the previous one. You can influence the first two entirely yourself, the third only partly.

1. Stage 1: Found

   Your website’s content is reachable and readable for search engines and AI systems. No important content only in images or PDF files, no accidental blocks, fast loading.

2. Stage 2: Understood

   The systems can clearly determine who you are, what you offer, for whom and where. Identical details on the website, business profile and directories, structured data, clear pages per service.

3. Stage 3: Recommended

   There is proof a system can draw on for a recommendation: concrete answers with numbers, reviews, mentions on other sites, entries in comparisons and trade media.

In a conceptual paper combining literature and industry data, Crăciun, Traistaru and Crăciun (2026) arrive at a similar picture. As signals that decide on citations in AI answers they name, among others, verifiable content, broad mention in third-party media and consistent information about a company across all sources. This differs from classic search engine optimization above all in one respect: what others write about you carries more weight.

What does this mean for your own website?

Your own website does not lose importance; it gains an additional role. It remains the place where prospects inquire. And according to Kumar’s (2026) figures, it is the most frequent source AI answers rely on. A website that answers questions clearly serves both purposes.

One technical point deserves a separate note. Some websites use their control file for search engines, the robots.txt, to block the programs AI providers use to collect content. That may be intentional, for copyright reasons for example. But it can also be an unnoticed default. Have it checked what your setting is, and decide deliberately.

What else a website needs so that visitors inquire is in the checklist A website that generates inquiries. If you want to hand off the implementation of structure, technology and reporting, you can find support, for example, at quik-services, a provider of websites, SEO and lead generation and a business partner of Owners Council.

Proof beyond your own website: the third stage

The third stage is the hardest because it is not entirely in your hands. An assistant making a recommendation relies on more than your self-presentation. For small businesses, four sources are realistically within reach.

• The business profile. Complete details, services, photos and regular reviews. How to do that is described in the article Google Business Profile.
• Reviews and testimonials on third-party platforms. Industry portals, review platforms, partner pages. How to collect them is covered in the article Using testimonials and references.
• Mentions in trade media and associations. An article in a trade journal, a talk at an association event, an entry in the member directory.
• Comparisons and overviews. Best-of lists and provider overviews in your industry. Check whether you appear there and whether the details are correct.

There is a danger here that owners should know about. Anyone who depends on recommendations from platforms whose rules they do not set makes themselves dependent. That applies to AI assistants just as much as to marketplaces and networks. How to recognize and limit such dependencies is discussed in the article Dependence on platforms. The best safeguard remains your own contact with customers: a customer list, a newsletter, person-to-person referrals.

How do you measure visibility in AI answers?

Honestly: poorly, so far. Pettiette, Conde and Fischbach (2026) interviewed 16 marketing practitioners from various industries. Uncertainty about measurement was among the main challenges they named. The respondents saw the new discipline as a complement to search engine optimization, not a replacement. A literature review by Popescu (2026) reaches the same conclusion: it is an additional logic of visibility whose evaluation is not yet standardized.

For a small business, three simple measurements are enough, and none of them requires special tools.

The third measurement is the most valuable because it reaches all the way to the conversation. Add “AI assistant” as a separate source in your inquiry list, just like search engine, referral or trade show. How this list is structured and how to calculate cost per order by channel from it is described in the article Measuring marketing success.

How urgent is this for a small business?

Less urgent than some providers claim, and more important than many owners believe. For businesses whose customers come mainly through referrals and personal contacts, little changes in the short term. For businesses that have so far received many inquiries via knowledge questions in search, something can shift noticeably.

For everyone: what you do for visibility in AI answers does not harm classic search, quite the opposite. Clear answers, unambiguous details and genuine proof are good work regardless of who reads them. Expert articles from your own experience, as described in the article B2B content marketing, are the best raw material for this.

First steps for this week

Start with an inventory. It takes an hour and shows you which of the three stages you are on.

• Ask two or three AI assistants ten typical customer questions and note whether and how your business is named.
• Compare name, address, phone number and services across website, business profile and the most important directories.
• Rewrite the first paragraph of your most important service page so that it names service, audience and region in one sentence.
• Have it checked whether the website blocks AI programs, and decide deliberately.
• Find two comparison overviews or directories in your industry and check your entry.
• Add “AI assistant” as a source in the inquiry list.

In a quarter, repeat the sample check. Then you will see whether anything has moved and can decide whether more effort is worthwhile.

We started by asking what a small business needs to do beyond good content for search engines in order to be named in AI answers. The first two stages, being found and being understood, are craft and largely overlap with good search engine work, supplemented by unambiguous information about the company. The third stage depends on evidence and mentions outside your own website, and that is where the real hurdle lies for small brands. How quickly the effort pays off cannot yet be said with certainty, given how young the research is.