← All articles

GEO for B2B companies: how to get found in ChatGPT and Gemini

September 11, 2026 · by Wildbos

B2B companies show up in answers from ChatGPT, Gemini, Claude and Perplexity by getting three things right at the same time. Your site has to be readable for AI crawlers, you need content that answers your buyer's question directly, and your name has to appear outside your own domain. One of the three is rarely enough.

We ran our own site past four assistants in July, on the questions we want to be found for. Wildbos did not appear in a single answer.

The site was technically sound. None of the directories and platforms we checked mentioned us. Technique alone does not get you into the answer.

What does the work consist of?

Generative engine optimization is about which sources an assistant picks when it composes an answer. There are only a handful. There is no second page where you can still show up.

The work breaks down into four parts.

  • Technical readability. AI crawlers often do not execute JavaScript, so content that only loads client-side is invisible to them.
  • Citable content. Pages that ask the question the way your buyer asks it, and give the answer in the first paragraph.
  • Visibility outside your own domain. Review platforms, industry directories and comparison articles about your category.
  • Measurement. Asking the same questions every month and noting who gets mentioned, with the source URLs attached.

Those four hang together. A perfectly readable site without mentions elsewhere produces what our own measurement produced in July.

How does an AI assistant pick its sources?

An assistant searches, reads a limited number of pages and summarises them. It picks sources that answer the question literally and that it can read without a browser.

In practice that means three things for your site.

Let the crawlers in through your robots.txt. GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended and CCBot are the names that matter.

Put the answer at the top of the page, in text that sits in the HTML. An assistant rarely reads past the first screens.

Mark up what you are with schema.org. Organization, Service, FAQPage, BlogPosting, BreadcrumbList and Person do the work, plus Review and AggregateRating where you have real reviews.

What do you do in the first three months?

  1. Measure your starting point. Ten to fifteen questions your buyer actually asks, across all four assistants, and note who does get mentioned.
  2. Fix readability. Allow the crawlers, server-side rendering on your core pages, schema on your organisation and service pages.
  3. Write one page per buying question that answers that question in the first paragraph.
  4. Build visibility on the platforms that fit your type of company.
  5. Repeat the measurement every month and compare on citations per engine.

Expect three to six months before the effect is visible and stable. Models also change on their own, so an improvement can come from your work or from a model update.

Which platforms count per type of IT company?

Type of companyWhere mentions count
B2B software and SaaSG2, Capterra, Google Business Profile
IT services and MSPClutch, Sortlist, Google Business Profile
Hardware and technical componentsGoogle Business Profile plus industry-specific directories
Staffing and recruitmentGoogle Business Profile as the anchor

Trustpilot does not work as an anchor for B2B, because consumer reviews carry different weight. Paying for reviews is never the route. Two to four reviews a month from satisfied clients is a pace you can sustain for years.

What goes wrong at B2B companies

Three mistakes come up most often.

The first is treating technique as the finish line. The dev team delivers a clean score and then it stops. Our own measurement in July shows where that ends.

The second is content about the company instead of about the question. A page with your platform as the heading rarely beats a page that has the buyer's question as the heading.

The third is measuring the wrong number. A single answer in ChatGPT says little, because the same model answers differently tomorrow. Citations per engine across a full month do say something.

How do you know it is working?

Guaranteed positions in AI answers do not exist. A spot in an AI answer is not for sale.

What you can do is ask the same questions every month and put the outcome next to last month's. Citations per engine and share of voice, with the source URLs attached, are the only numbers you can track honestly. If someone promises you a position, you know enough.

Frequently asked questions

What is the difference between SEO and GEO?

SEO targets a position in a list of blue links. GEO targets appearing in the composed answer of an assistant, which fits only a handful of sources. The technical base overlaps, the measurement does not. With GEO you count citations per engine and share of voice.

How long before GEO produces results?

Expect three to six months before the effect is visible and stable. The readability of your site can be fixed within weeks. Visibility outside your own domain, and the trust that grows from it, takes longer.

Can a B2B company do GEO in-house?

It can, provided you have dev capacity for the technical side and someone who measures every month. The bottleneck is usually execution, not knowledge. Most internal teams get as far as an audit and then stall on publication and follow-up.

Further reading

Getting found in ChatGPT and Gemini starts with knowing where you stand today. Anyone who wants to see which answers do or do not mention their company can run the free AI visibility scan at wildbos.com/geo-scanner.