Getting found in ChatGPT and AI Overviews: what actually matters for an online store (and how to measure it)

There are dozens of Dutch articles by now on "getting found in ChatGPT". Most of them say the same thing: write clear content, use structured data, be an authority. All true. All also exactly what SEO has been saying for twenty years.

This article is about what specifically changes for an online store, and, more importantly, how you measure whether it works. Because an AI visibility percentage means nothing if you don't know which answers your competitor shows up in and you don't.

First, the sober basics: AI search engines search through Google and Bing

ChatGPT, Gemini and Perplexity don't make their answers up out of thin air. For current questions ("best running shoes under 150 euros") they search live, through Google or Bing, and summarize the results. That's called grounding.

The consequence: if you don't rank well in Google, you rarely show up in ChatGPT. Research by SurferSEO showed that roughly half of the sources in Google's AI Overviews also rank in the top 10 of the regular results. So GEO (generative engine optimization) is not a separate discipline. It's SEO, plus a few things on top.

That's good news for online stores that have their SEO in order, and bad news for anyone hoping GEO is a quick shortcut.

The four things that make the difference for an online store

1. Your product feed is now a search source too

For shopping questions, ChatGPT shows products with a photo, price and summarized reviews. Those come from product feeds and structured product data, not from your blog copy. What to check:

• Product schema on every product page: name, price, availability, ratings, GTIN.
• Merchant Center feed complete and up to date. The same feed you use for Shopping.
• Consistent product names across feed, page and schema. Discrepancies mean AI systems don't recognize your product as the same one.

2. Comparison and buying-guide pages get cited more often than product pages

AI answers to "which X should I choose" cite pages that explain the choice. An online store with only product pages and categories has nothing to offer there. A page like "Which [product group] fits [situation]", with factual criteria, does.

This is exactly where the Search Console analysis from our SEO article lands: the clusters with question intent are the pages you need for AI answers.

3. Facts have to stand on their own, paragraph by paragraph

AI systems don't pull in your whole page, they pull passages. A paragraph that opens with "as described above" is useless. A paragraph that opens with "A [product] made of [material] lasts an average of [x] years under [use]" is citable.

In practice: every section opens with the answer, then the explanation. Headings as questions. Tables for comparisons.

4. Mentions outside your own site carry a lot of weight

AI systems weigh how often and where a brand gets mentioned. Reviews on external platforms, mentions in comparison articles, a Wikipedia-worthy description of what you are. This is the part you don't solve with content on your own site, and the part we keep doing human work for.

How to measure whether it works

This is what's missing from most articles. Three layers, from simple to serious.

Layer	What you do	What you learn
Manual	Ask ChatGPT, Gemini and Perplexity the same twenty fixed questions every month. Note whether you get mentioned, and who does.	A first picture, free, but subjective
Search Console	Google doesn't report AI Overviews impressions separately, but a rising position with falling CTR is the signal that you're in the answer without getting the click.	Where you're "seen but not clicked"
Automated	An agent asks the same question set every week, logs the answers and the sources mentioned, and reports the shifts.	Which answers your competitor shows up in and you don't: the list you can work on

That last one is what we run for clients, for the same reason as our SEO reporting: not "how visible am I" but "where am I invisible, and for which question".

What to do this month

1. Check your product schema and feed for completeness. This is the quickest to fix and has the biggest effect on shopping questions.
2. Pick five buying questions customers ask before they purchase, and write factual answer pages for them.
3. Ask your twenty fixed questions and record the baseline. Without a baseline you won't know in three months whether anything worked.
4. Fix your SEO foundation before you start on GEO. AI systems search through Google; whoever sits on page three there doesn't get cited.

Frequently asked questions

Is GEO different from SEO? No. It's SEO plus: citable passages, complete product data and external mentions. Anyone who says you need a separate "GEO agency" is selling you the same thing twice.

Can I pay to be in ChatGPT? Not for regular answers. For shopping results, product feeds play a role, comparable to Shopping.

Will I lose traffic to AI Overviews? Probably, on informational questions. Dutch publishers and online stores report double-digit percentage drops in Google traffic (Emerce, May 2026). The traffic that remains is closer to the purchase.

Does an llms.txt file help? There's no evidence that the major AI systems use it. It can't hurt, but don't put time into it before your product data and content are in order.

---

Want to know which AI answers your online store currently does and doesn't show up in? We'll run a free baseline on twenty questions from your category.