News & analysis
AI Shopping Needs Readable Pages, Not Just Feeds
A new webinar from Search Engine Journal says ecommerce teams should fix how AI tools read product pages before investing in better descriptions.
By Roomer Editorial · · 4 min read
What does the webinar recommend for getting products found by AI shopping tools?
The session, hosted by seoClarity's Chris Sachs and Aris Vrakas, laid out four connected moves: make product pages readable to bots, enrich content in the language shoppers actually use, keep feeds and pages in sync, and measure what changes.
The core argument: a product can have the right price and features and still get skipped by an AI shopping tool, because the tool never received the information in the first place.
Why do some products never reach AI shopping tools at all?
Access, not quality, is often the problem. Vrakas pointed to JavaScript-heavy pages where shoppers could see details that an automated fetch never picked up.
His advice: check what a crawler actually receives before touching copy. Confirm core product details sit in the HTML rather than loading only after client-side scripts run. Where rendering is the bottleneck, he suggested prerendering or cached HTML as fixes.
The sequencing matters. Better descriptions do nothing for a system that never sees them.
Why don't generic product descriptions work for AI shopping queries?
The presenters used socks as the running example: color and fiber content are listed attributes, but they don't answer every question about warmth, thickness, durability, or intended use, the kind of things a conversational prompt might combine into a single request.
Vrakas said retailers should connect product characteristics directly to the decisions shoppers are trying to make, rather than leaving an AI model to guess a use case from a short spec sheet. He also pointed to attributable review evidence as a way to back up performance claims.
This isn't limited to unusually detailed prompts. Even a broad request for "a good product" can push an AI system toward specific attributes the retailer never surfaced.
Why does feed and page consistency matter for AI visibility?
Because platforms expect the two to say the same thing. Sachs warned against treating the product feed as a side project owned elsewhere in the org. Information added to a feed should also update the page and its structured data.
Vrakas added that when a platform revisits a page, product identity, price, and supporting details all need to line up. An enriched feed doesn't remove the need to make the underlying page make sense on its own.
The presenters called this closing the loop: connecting SEO, merchandising, paid media, and development so improvements in one channel carry over, instead of leaving disconnected versions of the same product information scattered across the stack.
How should teams measure whether any of this worked?
Carefully, and not by treating an AI mention as proof of a sale. The presenters distinguished signals from outcomes, encouraging teams to look at effects across every channel they touched, not just the one they optimized.
Sachs tied this to testing discipline:
On a before-and-after comparison, Vrakas said the approach differs by catalog size. Large catalogs can compare product groups against each other. Smaller sites or broader brand questions need a defined comparison set: improve content around one topic, then track its movement against topics left untouched.
Roomer Editorial's take
Swap "product feed" for "video" and this webinar is describing a problem travel creators already have, whether they've named it or not.
A creator posting a hotel walkthrough in Lisbon has the same two failure modes as the sock retailer. First, the page the video lives on has to be readable by something other than a human scrolling. If the content exists only inside a platform's app, there's nothing for an AI tool or a search crawler to retrieve. A bookable page built from that video, with the hotel, the neighborhood, and the price structured in a way a crawler can actually parse, is the fix.
Second, specificity beats polish. "Great boutique hotel, loved it" tells an AI system nothing. "Walking distance to Alfama, rooftop pool, quiet despite being central" answers the actual questions a traveler is stacking into one prompt. Creators who narrate the specific trade-offs of a property in their videos are doing the enrichment work this webinar describes, without calling it that.
The consistency point applies too. If a creator's own site says one thing about a tour and the booking page behind it says another, that's the same disconnected-feed problem Sachs flagged. On Roomer, a creator's site is built from their videos and connects to bookable pages with rails through Vio, Stay22, GetYourGuide, Viator, and Klook, so the description and the booking live in one place instead of two.
None of this requires waiting for AI checkout to mature, which was the whole point of the session. The structural work, readable pages first, specific language second, is available now.
One action: pull up your most-booked properties or tours and check whether the detail that actually sold them is written down anywhere a crawler can read, not just said out loud in the video.
Source: Search Engine Journal — AI Shopping: Get Your Products Found In Google & ChatGPT
