E-commerce Businesses as Seen Through The Eyes of AI

E-commerce websites are quite a unique breed. They are structured a bit differently from other business websites and contain features not typically found on those, such as product pages. Online stores are also likely to be registered on platforms other than what service-based companies typically use. Because of these differences, it could be safe to assume that when a regular user asks AI about a store with functioning e-commerce, it will change its strategy on data gathering.

But does AI actually recognize these differences? Does it really have different priorities for e-commerce, or does it follow similar patterns to those it uses on service businesses?

We decided to explore where AI looks for information about online stores and determine what contributes to their successful LLM optimization. After all, we already noticed with service businesses that certain subpages and external websites consistently carried more weight than others. What we're curious to find out is whether e-commerce makes AI end up relying on the same familar sources, or if perhaps the not so familiar ones will be more valued.

First, let’s talk about intent

There is one key difference between service businesses and e-commerce. When people search for a service business, they usually want to know about the company itself – what it does, where it’s located, or how to get in touch. In contrast, when people search for an online store, they may be more interested in the sold products than in the company behind them. That’s why we tested two types of prompts in our research:

1. What do you know about the [business name] store from [city, state]?

2. What does the [business name] store from [city, state] sell?

The results of those tests, gathered using the GPT-5 model, showed a pattern:

AI tends to pick different sources depending on the user intent – it prioritizes external sites for general business queries, and business's own website for product queries.

Internal subpages and their varied roles

The e-commerce fundamentals

Every e-commerce website needs at least two essential elements – home page and product pages, which include singular pages for each item, and product lists based on filters.

AI found these two page categories to be the most useful overall. The home page was used as a source for 88% of the tested e-commerce businesses. It appeared in responses to both types of prompts for 69% of businesses, while being exclusive to business related prompts in only 3% of cases, and to product related ones in 16%.

At least one product page was used as a source for 67% of businesses. Unlike home pages, product subpages weren't selected as often for responses to both prompts – only 22% of the time. As it could be expected, those were found to be relevant for the second prompt only – in 42% instances a products or collections page was picked for the source list in response to just that prompt.

Pages designed for company details

Then we have all the pages meant to provide background information. We differentiated three more common types with slightly varying purposes:

1. About / Team – Found on 84% of reviewed websites and usually includes descriptions of the owner, the team, company’s history, or its goals. This was used as a source for 54% of businesses that had it on their website, which is a noticeable decrease compared to the 74% usage rate for service business websites.

2. Contact – Present on 83% of researched e-commerce websites. It usually provided a mix of phone numbers, e-mail addresses, physical locations, or a contact form. The contact pages were not a particularly popular pick for a source – it was used for 31% of businesses, which is again a decline compared to the47% usage rate for service businesses.

3. Location / Visit Us – These were considerably more rare, appearing on 28% of all websites, most often for businesses operating in multiple places. Location pages usually displayed a map and the company's address, with opening hours often mentioned as well. Despite their lower presence, they proved quite valuable for AI: it used them for 64% of e-commerce businesses that had them on their site. Location often substituted for a contact page when that only included either a form, or less of the same information.

Percentage of subpage types getting used as a source across two tested prompts

Noteworthy but less common pages

Other page types were not common enough for us to collect data for. However, we noticed two that could prove their use to AI.

• Services – not that many store businesses offered any services. If they did, they were usually mentioned in one page per service. AI could take note of those pages on both tested prompts.
• Events – we saw these mostly on hobby & entertainment store sites. These pages list any upcoming happenings that the business organizes or participates in. The chances for Events page to get picked as a source are nearly equal for both prompts as well.

External sources and industry-specific patterns

Hobby & entertainment – Niche directories can outshine

In this space (and in most of them as we’ll soon come to realize) MapQuest appears the most frequently. It served as a source for 72% of e-commerce businesses in hobby or entertainment industries. Reddit also played a bigger role than usual, appearing in 24% of results, while Yelp was less prominent at 48%. Additionally, several platforms didn't show up even once in this category, despite appearing in other ones. Those include BBB (Better Business Bureau), Wheree, NextDoor, and Dun & Bradstreet.

One standout was Record Store Day, which happened to show up exceptionally often for music stores (80% of them). While it's best known for promoting a global event that celebrates indie record shops, its site also happens to serve as a directory of pledged stores, making it highly relevant to AI. It's a perfect example of a niche site winning LLM optimization battles against more widely recognized names.

Home & garden improvement – Credibility sources matter more

This is one of the categories where AI really took a shine to MapQuest – this time it covered 84% of the picked companies. BBB being relevant for queries about 40% e-commerce businesses is also worth noting, because it reflects how ChatGPT values credibility sources more in industries where reliability and professionalism matter.

Reddit maintained its influence with a 24% appearance rate, as AI was able to find several threads with the researched companies getting recommended. Aside from that, we’d like to mention Houzz as one of the more common sources (16%), since this one is a niche platform dedicated to remodeling and interior design. While Houzz does not rival social media or other more prominent platforms not aimed at specific business types, its presence highlights the potential value of directories specific to industries like furniture or garden stores.

Personal care & lifestyle – Dominance of social and local media

This is where the sources discovered across all queries were the most diverse. That said, many of them were one-offs that happened to reference a specific business. Those could range from online magazine articles to niche business databases and official city websites.

On the other hand we have Yelp, Facebook and Instagram becoming more relevant here than in other categories, along with MapQuest still scoring really high. Those four were seen as sources for information on at least 68% of businesses. Interestingly enough, they were the only ones to even pass the 25% threshold, which underlines how heavily AI leaned on this small group, with social media playing a huge role.

Automotive & vehicle supply – Prestige over social presence

Once more, MapQuest topped the list, appearing in 84% of results. BBB also reached its highest influence here, as it was referred to in 56% of cases.

By contrast, social platforms like Facebook and Instagram had the least impact in this industry – both were used for data on only 32% of e-commerce businesses. At the same time, what we saw more than we expected was Wikipedia itself. While at first glance its 16% score is somewhat modest, it still suggests that automotive businesses could be already seen as prestigious. That leaves them realistically capable of basing GEO on their reputation alone, rather than social media visibility.

What's different between e-commerce and service businesses

With service businesses we noticed that while ChatGPT tends to prioritize the most authoritative websites like Yelp or MapQuest, LLM optimization doesn't fully work in their favor despite their high rankings on Google. While this remains true for e-commerce, the possibility of ChatGPT users inquiring about an online store with different intentions changes things a little.

External sources - usage by ChatGPT across two tested prompts, and appearances in Google's top 20 results without being ChatGPT's point of reference

Firstly, we can notice how much more likely MapQuest is to be referred to, compared to its performance with service businesses. With them it was a source for 39% of AI’s replies, but we could also find 45% other instances of service businesses, the names of which brought MapQuest up on top 20 Google results, but not on ChatGPT. And here, if we count each e-commerce business that was described by AI with MapQuest’s assistance on either of the tested prompts, we will get an outstanding 78% of all e-commerce businesses. Only for 8% of them we found directions on MapQuest that never got mentioned by ChatGPT.

Next, let’s take a look at Yelp – it was cited for 59% of e-commerce stores, while 21% of them had their Yelp profiles omitted. Although the omission rate is among the highest, it’s still less than half of the rate seen on service businesses.

Then, there are Facebook and Instagram, which are still somewhat favored by Google, but not as blatantly as they are with service businesses. In our research AI missed Facebook pages for 58% of those, and Instagram profiles for 37%. With online stores the omission rates are 31% and 26% respectively, which combined with at least 50% usage rate turned social media into a more reliable part of GEO for e-commerce.

What needs to be noted the most is decreased range of regular sources. Even though we saw 17 platforms getting used as points of reference for data on at least 7% of service businesses, with online stores that number decreased to 10 platforms. Many of the sites commonly brought up for service businesses are seldom picked as sources here:

• Reddit – drop from 45% to 16%
• Birdeye – drop from 33% to 4%
• Chamber of Commerce – drop from 33% to 4%
• Zaubee – drop from 16% to 1%
• Yellow Pages – drop from 14% to 5%

There are also three sites frequently ranking on Google, but getting unnoticed by ChatGPT. Those are Crunchbase (present in Google results for 14% of researched e-commerce businesses), Zoominfo (38%), and LinkedIn (56%).

Final takeaway

As it turns out, AI's treatment of online stores is a bit different from how it approaches service businesses. It builds its responses on a smaller, more focused set of sources, and the choice of referred websites depends on user intent.

With general business questions we are more likely to see external sources getting used. MapQuest, Yelp, Facebook and Instagram are the four dominant sites that will help you get optimized for both generative and search engines, at least as long as you take care of the content inside them.

If you want to optimize better for product-related queries instead, prioritize your own website:

• Pay special attention to your home, products and collections pages
• Keep website's content and structure clear – AI needs to move through it as easily as the user
• Don't overcomplicate your pages – AI wants info quality, not quantity

And if you're just planning to launch your own online store, but you're uncertain about your skills in creating compelling, well structured and optimized content in your website, you're in luck. There exist tools like our website builder that make creating professional online stores easy for everyone.

E-commerce businesses may have fewer viable LLM optimization options. However, the options that exist align more closely with SEO. This overlap means that improving your site for generative engines will likely increase its search engine visibility at the same time. And even recognizing this dual benefit could be the key to building a powerful online marketing strategy.

The data and statistics presented in this blog post come from a research study conducted by IKOL in 2025. To learn more about IKOL research methodology and explore other findings, visit: ikol.com/research.