AI search has become measurable since June 2026
Since 2 June 2026 you can see for the first time whether your webshop appears in AI answers. Google Search Console now shows generative-AI statistics: impressions, which pages turn up in AI answers, and in which countries. The rollout started with a portion of site owners in the United Kingdom and then goes worldwide. That sounds technical, but the consequences for your webshop and your Bol listing are practical: AI visibility is no longer an assumption you sense, it is a number you can read off.
That changes the conversation. Until now, GEO, optimising for AI search engines, was about an invisible outcome. You did the work and hoped that ChatGPT, Google AI Mode or Perplexity would name you. Now you see per page whether it works. And the scale at which it matters is large: AI Overviews now has more than 2.5 billion monthly users, AI Mode more than a billion. That is no longer a niche, that is the main entrance.
This article is not a recap of what GEO is. For that we have GEO for webshops and the breakdown for category pages. This is about the fresh step: now that it is measurable, how do you concretely get your webshop and your Bol listing ready to be cited. Which switches do you flip, in what order, and how do you check whether it works.
The pitfall: do not switch on that opt-out
With the news about the AI statistics came a second switch: an opt-out toggle with which you can exclude your pages from AI Mode and AI Overviews. The temptation is understandable. AI answers catch clicks, so why not pull yourself out of them and force the traffic back to your own page?
Because then you disappear from the place where more than 2.5 billion people search. The opt-out does not take you out of an annoying feature, it takes you out of the answer. Whoever opts out chooses to be invisible at the moment a customer asks an AI a buying question. For a webshop or a Bol seller that is not protection, that is a missed sale.
The healthy response is the opposite: stay switched on, measure, and optimise. The Search Console data now tells you which pages already land in AI answers and which do not. That is a worklist. The pages that do not yet appear are precisely the pages where the steps below pay off the most. You no longer have to guess which listing needs attention.
Claims instead of keywords: how an AI reads your listing
An AI does not read your Bol listing or product page the way a human does. A human scans the image and the first line. The model looks for facts it can trust and links those to the customer's question. "The best portable monitor for someone who works a lot on the road" calls for a product that is demonstrably light and compact, not for a listing that repeats the word "portable" twelve times.
That means: write properties that are testable. Not "super-fast charging", but "fully charged in fifty minutes". Not "compact and light", but "weighs just over half a kilo and fits in a laptop bag". The difference is that the second form contains a fact that an AI can link to the buying question and can verify against what customers write in reviews. A claim that rests on nothing does not count for a model, and a listing full of unprovable superlatives reads as less reliable.
Reviews are your strongest ally here. An AI treats your own claims and your reviews as one body of evidence. Claim that a product is quiet, and have ten reviewers confirm it, and "quiet" becomes a fact that the model dares to use. On Bol you have little influence over the page structure, but you do over your bullets, your product attributes and the reviews that come in. Steer towards reviews that address concrete properties, not general satisfaction. How you build a listing as a set of verifiable claims is something we work out at product listing design.
Structured content: make your facts machine-readable
On your own webshop you have a lever you do not have on Bol: structured data through JSON-LD. On a Bol listing you cannot add that schema; there you work with the attributes and bullets Bol gives you. On your own pages you decide it yourself, and that is exactly where the difference sits. JSON-LD delivers your product facts in a format that requires no interpretation. This is the price, this is the stock, this is the weight, this is the average rating. An AI that sees a monitor weighs just over half a kilo can use that fact for "lightest option" without guessing.
Most webshops do have the basics in place. A Product schema often comes standard from the platform, whether that is Shopify or another system. The gap is in the layers above it: the Offer object for price and availability, and the Review or AggregateRating object that makes your ratings readable. It is precisely those two objects that give an AI the facts it needs to recommend with confidence. Set them up completely, and fill in your specifications as structured attributes, not just as a loose sentence in the description.
Besides schema, the form of your text matters. AI systems prefer to grab self-contained, readable units of information: a FAQ block, a comparison table, a direct answer per heading. A FAQ on your product page, with questions customers actually ask and self-contained answers, gives an AI ready-made blocks it can take over literally. Write those questions from your customer-service inbox, not from what seems handy to you. For the broader background of what drives an AI recommendation, see how ChatGPT recommends your product.
Category pages: short, structured, with a direct answer
A common assumption is that AI search calls for long texts. For category pages that is not true. In practice, well-ranking category pages are often surprisingly compact: most stay well under a few hundred words. The gain is not in volume but in structure: a compact direct answer at the top, a clear layout below it, and a self-contained block that an AI can understand in its entirety.
That fits how a model selects. It does not pick your whole site, it picks the most relevant unit that covers a question. A category page that explains in the first two sentences who the category is meant for and how to choose is a usable answer for an AI. A page that immediately shows a product grid without explanation is not.
| Element | What it does for AI search | Common mistake |
|---|---|---|
| Direct answer at the top | Delivers a citable unit in the first 200 words | Page opens with only a product grid |
| Compact length | Keeps the block semantically complete and clear | Long walls of text that dilute the structure |
| Product schema + Offer object | Makes price and availability machine-readable | Schema without Offer or Review object |
| FAQ block | Provides ready-made question-answer units | No FAQ, or generic questions |
| Internal links to subcategories | Guides the AI and the visitor deeper into the range | Dead-end page with no further path |
The full seven-element approach for these kinds of pages is in GEO-ready category pages. The short advice: do not treat your category page as a filter screen, but as a self-contained answer to a broad buying question.
Internal links: the quiet lever that pulls an AI into your site
Internal links are underestimated in AI search. They do two things at once: they guide an AI deeper into your range, and they give your most important pages more weight. In controlled SEO tests, stronger internal linking typically produces a measurable but modest rise in organic traffic, with one honest nuance: the page that gives the link wins, while the gain for the receiving page is not always hard to prove. It is a lever, not a guarantee per page.
On larger webshops the effect can be considerable. Deeper internal links from category pages can noticeably lift second- and third-level categories, sometimes good for tens of thousands of extra visits per month at a large retailer. Treat such figures as direction, not as a promise. The underlying logic does hold. Pages deep in your range, precisely the specific product groups that best match narrow buying questions, often get the fewest links.
Practically: link from your strong pages to your deeper category and product pages with descriptive anchor text, not with "read more". Make sure every important product group is reachable from your homepage in two or three clicks. And link between related articles and products, so that an AI sees the connection between the explanation and the product the explanation sells. You maintain that just as you maintain your SEO: continuously, not as a one-off.
Video as an underestimated channel for AI citations
When you think about AI visibility, you think about text. But video is a large and underestimated channel for citations. After Reddit, YouTube is the most-cited social channel in AI answers, and those citations concentrate sharply: Perplexity and Google AI Overviews together account for the lion's share of the YouTube citation volume, with an enormous lead over any other video source.
Two patterns change how you look at this. First: the vast majority of YouTube citations go to long videos, not to Shorts. An AI cites the explanation, not the fifteen-second hook. Second, and this is the surprising part: views, likes and subscribers do not correlate with getting cited. There is no hard link between reach and citation. A video with few views can be cited more often than a viral clip, as long as that video answers a question thoroughly and clearly.
For a webshop that means something concrete. An explainer video about how to choose, use or compare your product does not have to be a viral success to feed your AI visibility. The value is in the content: does the video answer a buying question exhaustively enough that an AI dares to name it as a source? That is the same test as on your product page, only in video. For the broader shift towards AI-driven buying advice, see agentic commerce on Shopify.
What it delivers: content that carries the traffic
The stakes are not theoretical. Brands that build consistent, structured content over years often see their organic traffic carried above all by their knowledge-base and blog pages, not by their commercial pages. The gain then sits in a handful of explainer pages that answer broad questions exhaustively. Read such examples as direction, not as an average. But the pattern matches what we see: structure and depth, not tricks, carry the traffic.
For your webshop and your Bol listing that translates into a sober conclusion. The pages an AI can read, understand and confidently recommend are the same pages that convince a human. Structured data, testable claims, a compact category structure, strong internal links and explanation in text and video. None of those things is a hack. It is craftsmanship that now, since June 2026, can be measured.
What you can do this month
Do not start with your whole catalogue. Start with your ten most important products and your most important category pages, chosen on revenue and on the buying questions you want to appear in. Run through the same steps per page.
Step one: leave the opt-out for AI Mode and AI Overviews switched off, and open the new AI statistics in Search Console to see which pages already land in AI answers. That is your worklist.
Step two: set up the structured data completely on your own webshop. Product schema with a filled Offer object and AggregateRating, plus full specifications as attributes. On Bol: fill in your product attributes and bullets fully and factually.
Step three: rewrite your most important claims as testable facts, and check whether your reviews confirm those properties. Add a FAQ block with real customer questions and self-contained answers.
Step four: make your category pages compact and structured with a direct answer at the top, and lay internal links from your strong pages to your deeper product groups with descriptive anchor text.
Step five: measure again in Search Console a few weeks later. Are more pages appearing in AI answers? Combine that with your branded-search growth. More direct searches for your brand name is a strong signal that the AI answers are starting to name you.
At Oase Creative we build brand systems that stand on this: a listing or webshop that does not run on atmosphere, but on structure that convinces a human and that an AI can read. Since June 2026 that is no longer a matter of belief. You can read it off.
Frequently asked questions
How do I see whether my webshop appears in AI answers? Since 2 June 2026, Google Search Console shows generative-AI statistics for the first time: impressions, which pages appear in AI answers and in which countries. The rollout started with a portion of site owners in the United Kingdom and then goes worldwide. Do not switch on the opt-out for AI Mode and AI Overviews, because that removes you from those answers instead of letting you measure them.
What is the difference with regular SEO for my Bol listing? Regular SEO optimises for a position in a list. AI search optimises for selection: the model has to be able to understand, compare and confidently name your product. That is about verifiable data, testable claims and structured content, not about repeating keywords. A listing that runs on atmosphere rarely makes it into the answer.
Should I opt out of AI Mode and AI Overviews? No. Google has offered an opt-out toggle since June 2026, but AI Overviews now has more than 2.5 billion monthly users and AI Mode more than a billion. Opting out means you disappear from the place where those people search. The gain is in measuring and optimising, not in staying away.
Does video count for AI search results? Yes, and it is underestimated. After Reddit, YouTube is the most-cited social channel in AI answers, and the vast majority of those citations go to long explainer videos, not to Shorts. Notably, views, likes and subscribers do not predict whether a video gets cited. Content that explains a question beats reach.
How long should a category page be for AI search? Shorter than many people think. Well-ranking category pages are often surprisingly compact; most stay well under a few hundred words. The gain is in structure and a compact direct answer per page, not in long walls of text. Build the page as a self-contained unit of information that an AI can take over in its entirety.
