As personal AIs and chatbots transform search behavior, WPP’s chief AI officer, Dr Daniel Hulme, explains why marketers must now optimize for what AI “knows” – not just what people type.
“Reports of my death are greatly exaggerated,” Mark Twain famously wrote to a newspaper that had (rather prematurely) published his obituary.
The same could be said of SEO.
While some loudly decry its demise, the reality is more nuanced. SEO will remain a crucial tool, but its dominance is ending.
A move towards Generative Engine Optimization (GEO) is coming. It will be both a vital transition for established brands and a huge opportunity for newcomers and challengers.
As WPP’s chief AI officer, I see every day how AI is increasingly shaping shopper journeys.
Customers are asking ChatGPT, Gemini, Claude and Perplexity for purchase recommendations. Soon, personal AIs or agents will be empowered to make independent purchasing decisions on consumers’ behalf. With Gartner predicting that 25% of searches will bypass traditional engines by next year, AI discovery is fast becoming survival-critical for brands.
Want to go deeper? Ask The Drum
Why your SEO strategy is only half the story
GEO optimizes content for visibility in AI-generated responses. Unlike SEO’s keyword matching, Large Language Models (LLMs) infer information from patterns in training data. The fundamental shift: LLMs deliver direct answers with links, not links to answers.
When asked “best skincare brand”, AI draws from statistical patterns across millions of training materials. If your brand wasn’t prominently featured within these sources, you effectively don’t exist.
The silver lining is that GEO-driven traffic is typically highly qualified. Users who’ve researched via AI and still visit your site show genuine purchase intent. But they have to find you first.
First, find out what AI already thinks of you
Since GEO is based on how LLMs process and retrieve information, start by auditing what these models “know” about your brand. Ask multiple models about your company, products and competitive positioning. The results will likely surprise you. Variance across the models is huge, and individuals’ search results can also be contextual.
When you query an LLM, it’s not searching a database. It’s activating neural pathways formed during training. Test both branded queries (“Tell me about [your company]”) and unbranded ones (“Best solution for [your category]”). The response differences reveal how strongly your brand connects to your category in the model’s learned representations.
We can draw parallels with how our brains work. Just as we associate “safe cars” or “winter holidays” with specific brands, we need to train LLMs to strongly associate concepts with our brands, increasing the likelihood of retrieval when queried.
My advice: prioritize fixing high-risk associations first. That includes outdated information, consistent competitor recommendations and reputational issues dominating your brand’s space.
Three moves that separate GEO winners from everyone else
Once you’ve addressed the urgent issues, refine your approach. Based on my work with AI systems, here are three key elements that separate GEO leaders from laggards:
Understand training versus retrieval. Base models such as GPT-4 only access training data with specific cut-off dates, while retrieval-augmented systems like Perplexity can access live content. This is the first question to ask of your content strategy: which content should remain gated for lead generation, and which should be opened to ensure discoverability across both types of systems?
Optimize for semantic understanding. LLMs cluster concepts by meaning, not keywords. Use natural language to cover your full concept space, mirroring how your target audience searches. Instead of “enterprise SaaS solutions,” think “what’s the best software to manage my team’s projects?” Teach AI who you are, not just what you sell.
Build for knowledge extraction. While it’s important to maintain SEO best practices, remember that AI systems increasingly extract entity relationships to build their knowledge bases. Create comprehensive, rich content that explains not just what you do, but how you relate to problems, solutions and competitors.
Why waiting is the riskiest strategy
Don’t wait for perfect measurement frameworks or strategic approaches. Even AI experts are still developing best practices. Start today by asking top AI models what they know about your brand. Document responses regularly, especially after model updates. That baseline will be invaluable in six months.
Suggested newsletters for you
Daily Briefing
Daily
Catch up on the most important stories of the day, curated by our editorial team.
Weekly Marketing
Friday
Stay up to date with a curated digest of the most important marketing stories and expert insights from our global team.
The Drum Insider
Once a month
Learn how to pitch to our editors and get published on The Drum.
And remember, GEO isn’t just another marketing channel. It’s a fundamental shift in how brands need to think about discoverability in an AI-mediated world. Training datasets are becoming increasingly curated and quality-focused, and the content published today by you or your competitors influences how your brand is represented in models trained tomorrow.
So, the question isn’t whether GEO will replace SEO and reshape brand discovery. It’s whether your brand will be part of the conversation.