More people now discover some brands through ChatGPT, Gemini and Claude than through Google. Here is a practical AEO playbook, based on Scott Dodson of Holded at the AI Summit Barcelona.

More people now discover some brands through ChatGPT, Gemini and Claude than through Google. That was the headline from Scott Dodson, who leads marketing at Spanish accounting software company Holded, speaking at the AI Summit in Barcelona. His argument: answer engine optimisation (AEO) is today's critical growth channel, and the window to get ahead is open right now.
We've turned his talk into a practical playbook. Call it AEO, GEO or AI visibility; the work is the same.
Every growth channel follows the same curve. It starts as a niche for innovators, gets picked up by early adopters, then hits an inflection point and goes mainstream. Almost every one has been triggered by a leap in technology.
Print built Coca-Cola. Radio and television created brands that didn't exist before. Bundling software with Windows was the way to win in the late 80s and early 90s. Zynga reached a user base on Facebook in a month that took World of Warcraft years to build. The brands that moved early on each new channel won out of all proportion to their size.
Scott's point is that AEO is at that same moment. Holded surveys new customers on how they found the product, and AI assistants have already overtaken search engines as a discovery source. More people now subscribe to Holded on the back of an AI recommendation than a Google search.
Holded's position is strong, Scott said: it leads its competitors in Spain on AI mentions, citations, visibility and average position for the queries that matter to it. The only metric where it doesn't lead is sentiment, where it sits third. Yet the growth curve still looks like the early-adopter stage. The mainstream inflection hasn't happened yet, and that's the opportunity for everyone else.
AI wants answers, just as Google has for two decades. Your job is to be the source of those answers. What changes is how you package them.
Go narrow. Unless you're a household name, you won't win "best software in Spain". You might win "best accounting software for freelancers" or "best invoicing tool for small businesses". Pick the specific questions your buyers ask and own them.
Match the length of the question. This was Scott's biggest takeaway. A Google search is typically three or four words. A prompt to an LLM is 15 words or more, and growing. People describe their situation, their constraints and what they've already tried. Your content needs to answer in complete, specific sentences that match those conversational questions, not a list of keywords.
Stop mass-producing AI content. If you're publishing generic AI-generated articles and wondering why nothing shows up in AI answers, this is why. AI answers lean towards sources that are original, credible and specific. Content that says what everyone else says gives them no reason to cite you.
First-party content is everything you publish on channels you own. Scott's two priorities: make more video, and make E-E-A-T (experience, expertise, authoritativeness, trustworthiness) your mantra.
Build a video library. Do what you've done with blog posts for the last ten years, but on video. Holded runs a series answering business jargon questions on camera. AI models are getting better at reading video, and Google already draws heavily on YouTube.
Make it unmistakably human. Generic content doesn't get cited. Content that's unique, curated, comprehensive, fresh, backed by data and well-structured does. In practice that means:
Put your reviews to work. Holded automatically pulls fresh reviews from a third-party review platform into its site, features the five-star ones on the homepage and marks them up in structured data. Peer validation, kept current, in a format machines can read. Wire everything you can into schema.
Great content is wasted if AI can't parse it. Scott boiled the technical side down to three jobs:
If your SEO foundations are solid, you're most of the way there. If they aren't, fix them first.
Here's the hard truth: according to Scott, most of your AEO results come from third-party sources, not your own site. The strategy is "everything everywhere all at once": be present, credibly, on every platform where your buyers talk.
Third-party SEO is a useful starting map, but some of the old rules have changed.
| Old-school SEO | AEO |
|---|---|
| Content built around keywords | Content built around full questions and answers |
| Chasing dofollow backlinks | Earning credible, contextual mentions, linked or not |
| Brand marketing as a separate job | Every brand mention feeds AI visibility |
| PR as a nice-to-have | Good PR as table stakes |
Mentions matter more than links. AI reads the whole conversation around your brand, so an unlinked mention in the right context still counts.
Peer-to-peer conversations carry weight. Communities where real users discuss real products are highly credible. Scott pointed to OpenAI's content deal with Reddit as a sign of how much weight those conversations carry.
Social is becoming search. As AI gets better at understanding video, YouTube, Instagram and TikTok content will increasingly shape what AI says about you.
Every AI platform sources its answers differently, so you can't optimise for all of them equally. Among AI assistants, Scott said ChatGPT is the clear leader, with Gemini second and Claude third. But when you count every place people go for answers, search included, Google still dwarfs everyone. His order of priority: Google, then ChatGPT, then Claude. Ignore the rest until those are covered.
| Priority | Platform | What it seems to reward | Where to focus |
|---|---|---|---|
| 1 | Google (Search, AI Overviews, Gemini) | Its own ecosystem, especially YouTube | Make more YouTube videos |
| 2 | ChatGPT | Credible community conversations | Build a genuine presence on Reddit; Scott expects LinkedIn to matter more over time |
| 3 | Claude | Scholarly and research-grade sources | Where it fits your brand, partner with universities and researchers on studies |
Scott was clear that the LinkedIn point is his own hypothesis, based on Microsoft owning LinkedIn and its close ties with OpenAI. Treat it as a bet worth watching rather than a rule.
Scott closed with two trends pulling in opposite directions.
Agent-led growth. Product-led growth becomes agent-led growth. Picture telling your personal AI agent before bed: "Find me a good accounting tool." Overnight, it compares options, questions the vendors' own website agents and fills in the sign-up forms. In the morning you get a shortlist, ready to go with one click. Scott said Holded and others are already building the infrastructure. The brands agents can find, understand and transact with will win those decisions.
Human-centric marketing. In a world of agents talking to agents, what feels human becomes more valuable. Scott expects experiential, event-based, story-driven and entertaining marketing to grow in demand. Holded already backs Buenas Ideas, a documentary series on Amazon Prime supporting entrepreneurs. He also noted a stronger backlash against AI in the US than in Europe, which makes craft, imperfection and real-life events more valuable, not less.
His parting bet, with no data behind it, just instinct: Substack.
For us, this is the part that matters most. Getting found by AI is the entry ticket. Being the brand people, and their agents, actually choose still comes down to understanding why people buy.
Want to know how your brand shows up in AI answers today? Get in touch with Behaviour Digital and we'll show you where you stand.
This article is based on Scott Dodson's talk at the AI Summit Barcelona. Figures and results quoted are as presented by Holded.
Luke leads strategy across every Behaviour Digital account — building data-driven advertising systems for ambitious brands. When he's not auditing ad accounts, he's writing about what he found in them.
Get a free, tailored growth plan — we'll run our framework on your accounts and show you exactly what we'd do.
Get your free growth plan →