Earlier this week, Croud was proud to sponsor the third edition of ISBA’s Gen AI Summit, bringing together senior marketers from brands including Currys, L'Oréal, Mars, Monzo, NatWest and Vodafone at Google's London HQ.
The day tracked the AI transformation challenge, from changing consumer behaviour and AI-led discovery to measurement, governance, operating models and skills. Our Managing Director, Integrated Search, Josh Preston-White, shared the latest Croud Consumer Index findings on stage before joining ISBA's GEO Working Group chairs, Elise Funnell from Currys and Ross Edwards from Taylor & Francis, for a panel on earning visibility in AI search. Here are the highlights.
The AI shortlist is shrinking the research phase
Josh opened with the latest edition of theCroud Consumer Index, our proprietary research into how AI search is changing discovery. 74% of AI users now turn to LLMs before choosing a specific product, and because they can run more complex, bespoke searches, the research phase is getting shorter. That's good news for challenger brands that answer questions quickly and accurately, but it means every brand is fighting on several fronts to be seen, found and chosen.
Our research also found that AI users describe what they want through brand, need, use case and style cues, so clear visual and tonal signals matter as much as functional descriptions. This means ensuring we have engaging, platform-specific creative which speaks not only to humans but meets LLM needs is vital to driving discoverability. At Croud, BrandCI measures LLM visibility across query depth, fan-out and position, and SEOCI turns those insights into content and site improvements to ensure we’re translating insight into action.
Welcome to the Validation Economy
The time consumers used to spend researching hasn't disappeared. It has moved into what we call the Validation Economy: the space between an AI recommendation and the purchase, where people look for human proof rather than taking the LLM's word for it. 33% of AI users check four or more sources before buying, with 37% looking for real user experiences and honest reviews, 31% wanting confirmation that the product will do what they need, and 21% seeking evidence of the brand's reputation.
Each platform plays a different role. YouTube, used by 41% of AI users, is the go-to for how-to guides and long-form content on bigger purchases. Instagram and TikTok (35%) build visual confidence and lifestyle fit. Reddit (22%) is the one to watch: it appears increasingly often as a source across LLMs, particularly Gemini, and rewards community building over selling.
How much validation matters also depends on the category. Only 18% of AI users say they must check beyond AI before buying fashion, compared with 44% for financial products. For high-consideration categories, Josh's advice was to build a content programme that answers the right questions across the platforms consumers actually use, and to treat GEO as consumer-first rather than optimising purely for LLMs.
From the panel: Joining up paid, organic and the organisation
Three months into UK trials of ChatGPT ads, Josh cautioned against judging them against Google Search. Initial tests have seen them deliver cheaper engagement and lower CPMs but less business value, though performance is improving as automation and new formats arrive. For now, treat them as a mid-layer of discoverability rather than a competitor to traditional search.
He also argued that it has never been easier to get into marketing, and never been harder to stand out. Croud's approach focuses on five areas over a typical 6–12 month roadmap: organisation (working closely with finance to unlock investment), structure (breaking down silos between paid, organic and social teams), execution, creative and joined-up roadmapping.
The wider discussion kept returning to a few themes:
- Strategy before tactics. Start with “visibility of what, and for whom?”, then decide which content models can use freely, which you want to control, and which stays behind a paywall.
- Measurement is still maturing. Definitions of citations, share of voice and brand mentions vary, and traffic is no longer a clean proxy for value. ISBA's new GEO framework project aims to standardise terminology and methodology.
- Authority beats paid placement. In one example discussed, a retailer’s ad appeared alongside an LLM’s answer, but the retailer was never recommended.
- Build for agents and humans. Content structured purely for agents can end up dry; it needs to work for people too.
- Make proofs of concept count. Set a baseline, run pilots in phases and agree commercials upfront.
Underneath it all was a reassuring message: much of what drives AI visibility is good old-fashioned brand marketing, applied at greater scale. You can't claim to be something the reviews contradict.
What brands should do next
Josh left the room with three takeaways:
- Audit your AI presence. Understand where and how you appear in the AI shortlist.
- Build a connected proof system. Keep content structured, accurate and consistent, create work that satisfies people and LLMs alike, and match the effort to your category's validation needs.
- Treat your website as the moment of validation. A citation alone doesn't win the sale; site structure, creative and visibility across the validation layer do.
Read more
Download the fullCroud Consumer Index: The Validation Economy for more on how AI is reshaping the path to purchase and how brands can win trust in the Validation Economy.
Want to talk through what this means for your brand?Get in touch with the Croud team.