Expert Opinion: AI Search Visibility | Velstar Webinar Roundup

A brand can rank on page one and still be losing clicks. According to Eugen Bucurescu, Solutions Engineer (AI & Performance) at Velstar, that's not a fluke - it's what happens when AI overviews, carousels and AI-mode answers sit above the results people used to click through to.

Please Note:

Account details may vary. The feature is now part of the subscription

See Pricing

In a live conversation with our very own Catherine Wilding, Eugen walked through what's actually changed in how brands get found, why two brands with near-identical rankings can get wildly different outcomes, and why customer evidence is doing more of the work in deciding who AI systems trust.

Watch the full recording below, or read on for the key takeaways.

Key takeaways from the Velstar webinar

A single ranking no longer tells you whether a brand is actually visible. Eugen has seen brands holding strong top-ten positions while click-through rate quietly collapses, because those clicks are going to Google's AI Overviews and AI mode, plus channels like ChatGPT and Perplexity that traditional rank checkers don't track. It gets stranger still: AI agents often rewrite a shopper's query before searching - Eugen calls this "query fan-out" - so a brand can rank well for the query it's tracking and still be invisible for the query an agent actually ran.

Reviews give AI systems something to work with instead of guessing. Eugen linked AI hallucination directly to a lack of information: when an agent doesn't have enough to go on, it fills the gap - sometimes wrongly. Reviews provide sentiment, specific product detail and an aggregate score that brand-owned marketing copy can't, giving an agent real evidence to recommend a product with confidence rather than guessing.

Brands disappear for two different reasons, and the fix depends on which one it is. Invisibility is a technical problem - a website accidentally blocking AI crawlers, pages that rely on JavaScript an agent can't render, inconsistent heading structure. Misrepresentation is a different problem entirely: outdated information, or negative reviews left unaddressed, that get pulled into an agent's answer and colour how the brand is described. Diagnosing which one you're dealing with determines whether the fix sits with development or with reputation management.

Being visible on your own site isn't enough - off-site presence decides who gets picked. Eugen described a wholesaler client who was outranked in AI recommendations by his own retail customers, despite supplying them directly. The retailers simply had more presence across the web - social media, forums, publications - so agents trusted them more. When several sites can answer the same query, Eugen's point was that the one with the most visible evidence of expertise and existence beyond its own site is the one that gets recommended.

Once you stop chasing rank, you need something else to measure instead. Eugen pointed to three layers working together: presence (does the brand show up in relevant AI answers at all), traffic (do those appearances actually send visitors), and commercial impact (do those visitors convert). He was cautious about leaning too heavily on mentions and citations alone - what marketers call "share of voice," essentially how much of the online conversation a brand accounts for - since these can shift from one query to the next even for the same brand.

Expert quote

"You need to exist and be surfaceable across all of these engines and agents, because each one draws on a different set of sources. The brands that prepare now will be the ones agents can find, trust and transact with." Eugen Bucurescu, Solutions Engineer (AI & Performance) — Velstar

What this means for eCommerce teams:

Stop tracking position as the primary success metric. Track whether you're appearing in AI answers at all, how much traffic those channels send, and what that traffic is worth commercially.
Check whether your content is actually readable by AI agents, not just by Google. Test whether key page elements like FAQs and category descriptions render without JavaScript, since most agents read text only.
Look at your brand's presence beyond your own website - social presence, forum mentions, third-party publications - not just what's on the page.
Audit whether negative reviews have been addressed. An unanswered bad review can shape how an AI system describes your brand, regardless of your average rating.
Treat reviews as a fast, low-effort way to keep fresh, specific, human-language content flowing to AI systems.

Velstar's perspective: presence over position

What made Velstar's session distinctive wasn't a claim that SEO is dead - Eugen was explicit that it isn't. His argument was narrower and more useful: ranking position was always a proxy for something else, and that something else now needs measuring directly rather than assumed from a keyword's position.

The wholesaler example made the stakes concrete - two businesses with comparable products and rankings, but only one recommended, because presence off-site tipped the balance. Eugen's split between on-site presence (can an agent read and understand you) and off-site presence (does the wider web know you exist) gives eCommerce teams a genuinely different lens than a rankings dashboard, and explains why visibility and being recommended aren't the same thing.

Continue exploring AI visibility

This roundup is one part of REVIEWS.io's wider look at what it takes to get recommended by AI. Eugen's full answer is one of five featured across five leading eCommerce agencies in Get recommended by AI - unlock all five agencies' answers.

You might also find these useful:

About Velstar

Velstar is a Shopify Platinum Partner and full-service eCommerce agency with expertise across build, performance, AI and trading, working with ambitious DTC and B2B brands on complex commerce challenges. Speaker: Eugen Bucurescu, Solutions Engineer (AI & Performance).