Skip to content
trending

Guides

GEO SEO: How to Check Whether ChatGPT Recommends Your Business

Learn how to measure whether ChatGPT, Gemini, Perplexity and Google AI Overviews mention your business and cite your website.

By 6 min read

👁 113 reads

GEO (generative engine optimization) is about helping ChatGPT, Gemini, Perplexity and Google AI Overviews recommend your business and cite your website. With traditional SEO, you compete for a place among ten links. With GEO, you want to be one of the few names a model mentions in a paragraph. In September 2026, I asked ChatGPT and Perplexity questions about 20 Polish businesses across different industries. ChatGPT did not cite the websites of 17 of them, even though all had working websites. Here’s how I measure visibility and what you can change.

At a glance

  • Measurement: Ask a consistent set of 30-100 customer questions several times in ChatGPT, Gemini, Perplexity and AI Overviews. One screenshot is not a measurement.
  • Two metrics: A mention (the business name appears in the answer) and a citation (the business’s domain appears among the sources). I count them separately.
  • What you can change: Bot access, website content and your business’s presence on sites that models cite. You cannot change knowledge already stored in a model.
  • Status: My September 2026 measurement and GEO course.

With a search engine, customers compare offers themselves: they open two or three tabs and read. In an AI answer, the model does the comparison before the customer sees anything. They get a paragraph with a few names and a ready-made recommendation, without a second page of results.

In my measurement, 60 ChatGPT answers with search enabled had an average of 3.3 sources, ranging from zero to eight. Perplexity included 12 sources in each of 60 answers. Pew Research Center measured the effect in Google: when an AI summary appears for a query, 8 percent of users click a regular result, compared with 15 percent when there is no summary.

AI models draw on three sources of information, and you can influence only two

  • Training data. What the model read about a business up to the end of its training period. You cannot update this information; you can only check its effects by asking the model with search turned off.
  • Live search. For questions about prices, businesses in a city and current events, the model searches the web. It breaks a question into several supporting queries, retrieves pages and selects excerpts. Bot access, website content and your presence on other sites matter here.
  • Cited sources. Some of the retrieved pages appear as links. This is the only layer you can see from the outside, so it is the one I measure.

Supporting queries explain why a good Google ranking is not enough. A page ranking first for “heat pump price” may have nothing to say in response to a follow-up question about service costs after five years. A business further down the results can win by publishing a table of service costs.

A mention is not the same as a citation

For each answer, I record two things separately: whether the business name appears in the text and whether its domain appears among the sources. Confusing the two leads to misleading reports.

Here’s an example from my measurement: for five medical and related businesses, I collected 30 answers. ZnanyLekarz appeared as a source in 18 of them. The model recommended a specific doctor but linked to the doctor’s directory profile, not the clinic’s website. For the business owner, that is still a recommendation, but one made through a third party. For me, it shows where to focus the work.

It can also work the other way around. A business’s domain may appear eleventh among Perplexity’s 12 sources, while the answer does not mention the business at all. Technically, that is a citation; in practice, it is invisibility.

How to measure GEO step by step

  1. Build a question set. Use 30-100 questions customers ask just before choosing a business: recommendations (“Who would you recommend for…”), comparisons, problems and prices. I draw them from the client’s sales conversations, Google Search Console queries and industry forums.
  2. Check four platforms. Use ChatGPT, Gemini, Perplexity and AI Overviews. Each searches and cites sources differently. With the same questions, Perplexity cited this site in 60-70 percent of checks, while ChatGPT and Gemini did so in zero percent.
  3. Repeat the queries. Ask each question several times and save the answer and its sources. The names in an answer can change each time you ask.
  4. Track the metrics. Measure the business’s share of answers, the position of its mention and the answer’s tone, compared with 3-5 competitors.
  5. Review competitors’ sources. List the sites models use to recommend competitors. This gives you a ready-made plan for work beyond your own website.

Make your website easy for AI models to extract information from

First, I check robots.txt and server logs. Bots relevant to GEO include GPTBot and OAI-SearchBot (OpenAI), PerplexityBot and Google-Extended. If you block a search bot, the model cannot retrieve your page when someone asks a question.

Next, I review the content. Models cite excerpts, so put the answer in the first sentence of a section, consider using questions as headings, and present prices and specifications in tables. Watch for elements that may disappear during content extraction: text in side panels, tabs loaded with scripts and collapsed accordions.

Finally, check that your business information is consistent across the web: your “About” page, pricing, services and contact details. Some assistants use Bing’s index, so I check Bing Webmaster Tools and submit changes through IndexNow, a step many businesses overlook.

Build visibility beyond your website with directories, comparisons and reviews

Models use directories, rankings and reviews because these make it easier to compare several businesses at once. A business with a good website but no presence in these places can lose to an average business that appears everywhere. That is why GEO work is at least half off-site: your Google Business Profile, Bing Places, industry profiles and inclusion in roundups. Don’t buy fake reviews.

GEO as a service: audits and ongoing work

A GEO audit can include a report and an action plan. A smaller initial scan can test 10 questions and show a business owner which companies models recommend in their area. Ongoing work can include monthly measurement. These are service options, not a promise of earnings. I do not promise a place in an AI answer; I offer measurement and implementation.

I explain the full process in my course, GEO: The Business ChatGPT Recommends: 29 lessons, 8 hours and 54 minutes, two complete audits (a service business and an online store), and a report template for company leadership. Lesson 1, on the difference between a search engine and an AI answer, is free. For a guide to traditional SEO for large websites, see AI courses.

If you’re interested in your website’s traditional Google performance rather than AI visibility, my step-by-step process, from initial questions to a report for company leadership, is in my article about an SEO audit.

Frequently asked questions

How is GEO different from SEO?

SEO focuses on ranking in search results; GEO focuses on getting a model to mention your business in an answer and cite your website. Some of the work overlaps, but GEO is measured differently: by your share of answers to a consistent set of questions, rather than rankings and clicks.

How can I check whether ChatGPT recommends my business?

Write down 30 questions customers ask before choosing a business. Ask each one several times in ChatGPT with search enabled, and record whether your business name appears and whether your website is among the sources. Repeat the process in Gemini and Perplexity, because the results can differ significantly.

Should I block GPTBot in robots.txt?

If you want your business to be visible in ChatGPT, do not block OpenAI’s search bots. Make a separate decision about bots that collect training data and those that retrieve pages when someone asks a question: OpenAI and Perplexity describe these as different bots in their documentation.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources

Kinetyra

// read next

More on this from QUANTSLANT

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

read nextAnthropic Files for IPO as AI Listing Could Set a RecordnewsAnthropic Files for IPO as AI Listing Could Set a RecordAnthropic has filed for an IPO on Nasdaq under the ticker ANT. Reports suggest the offering could raise as much as $100 billion.

Most read

  1. 1Best AI Video Generator Apps for Your Phone (2026)
  2. 2Claude Opus 5.5 Motion Graphics: 60s Showreel from One Prompt
  3. 3Wan 2.2 on RTX 4090: timings, VRAM and four test clips
// site map
Quantslant in numbers 1,423posts in archive22news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›