I just got back from Ahrefs Evolve in Singapore. Two days, excellent speakers, and a community that’s clearly trying to figure out what marketing looks like now that AI is rewriting the rules.
I went expecting tactical SEO content. I left thinking about something bigger: the gap between what’s possible with AI right now and what most marketers are actually doing with it. That gap is enormous, and it showed up in almost every talk and conversation I had over the two days.
Here are the four talks that stuck with me the most, and the themes I kept coming back to afterwards.
Alan Chan (Manus AI): the imagination gap
This was my favourite talk of the conference. Alan Chan is Head of GTM at Manus AI, and his central thesis was simple: most people don’t understand what AI can actually do yet.
He calls it the “imagination gap.” People are using ChatGPT to rewrite paragraphs and summarise articles. Meanwhile, Manus is running agents on cloud computers with skills that dynamically update based on output. The agents find work, execute it, iterate on the results, and improve over time.
That resonated with me because I’ve built something similar. My AI assistant (Jarvis) runs 24/7 on a VPS: managing my calendar, doing research, monitoring markets, sending emails, running scheduled jobs, even drafting content for review. The building blocks for this kind of system are already here. Most people just haven’t connected the dots yet.
Alan’s five principles were worth writing down: context is king, obsess with automation at scale, embrace human-AI collaboration, if you can explain it you can build it, and go the extra mile. That fourth one is the key insight. Any workflow you can articulate clearly, you can turn into an agent. Most marketers aren’t thinking this way.
One practical tip he shared: set up a separate email address, subscribe to all your competitors’ newsletters, and feed them into AI for automated competitive analysis. Simple, effective, and exactly the kind of thing the imagination gap prevents people from doing.
Charlotte Ang (Traffic Bees): video is the new top of funnel
Charlotte Ang shared how Traffic Bees built a million-dollar e-commerce brand from scratch. The case study was impressive on its own, but the structural insight was what stayed with me.
Her argument: video is the new top of funnel. SEO is moving down to mid-funnel as a conversion layer. The content pipeline she described was straightforward: write an article, convert it to short-form video, distribute across TikTok, Instagram, and Facebook. Three content formats to focus on: evergreen, authoritative, and trendy. Live streams for building trust and driving sales.
The part I keep thinking about is the moat concept. Charlotte’s example was PickleGo, which runs its own in-person events to build community. That’s a moat you can’t copy by spinning up another AI agent. It’s a real-world investment in relationships and trust, and it’s a useful counterbalance to the “automate everything” instinct.
Her other point that landed: meaningful experiences beat content quantity. Producing more doesn’t help if nobody cares about what you’re producing.
Robert Lai (Kaliber): growth happens before search intent
I spoke to Robert the night before his talk at the networking dinner, and his framework clicked immediately.
People are talking to AI much earlier in their buying journey now. Instead of googling “dentist near me,” they’re asking an AI “why do my gums bleed when I floss?” The funnel starts earlier than most marketers realise.
Robert’s advice: rank for situations, not keywords. Map the buyer, not the keyword. The way to do this is to mine survey data, customer support logs, and user interviews for the questions people ask before they ever open a search engine. Feed those pre-search questions into your content strategy.
This has direct implications for how anyone in SaaS or WordPress should think about content. Keyword research alone isn’t enough anymore. The real opportunity is in understanding what people are thinking and asking before they’re even in buying mode. I’ve written about the shift towards AI-first discoverability before, and Robert’s “pre-funnel” thesis takes it a step further.
Dan Petrovic (DEJAN): a million questions to LLMs
What I loved about Dan’s talk: he brought hard, first-party data that he collected himself. A million queries to Gemini and GPT, with novel insights that nobody else has. This is exactly how we should all be thinking about AI. We have access to the APIs. Running the experiments ourselves is a powerful approach, and it’s something I’ve done at GravityKit too.
Key findings from his research:
- LLMs have massive biases. Certain countries and industries are significantly underrepresented in recommendations.
- Single-token brand names get recommended roughly 8x more than multi-token names. If you’re naming a product, that’s worth knowing.
- Specificity matters more than superlatives when optimising for LLM recommendations. Quantified outcomes and verbs beat adjectives.
- Brand entity mapping is the first step. Understand how LLMs see your brand and its relationships, then work to strengthen those connections.
He also shared a tool, authority.dejan.ai, for checking how LLMs perceive your brand. Worth bookmarking.
The themes I kept coming back to
Beyond the individual talks, three things kept surfacing across the two days.
The AI adoption gap is real, and it’s massive
At one point, the room was asked: “Who has automated more than 50% of their work with AI?” I was the only one who raised my hand. At a marketing conference. In 2026.
Most attendees hadn’t heard of MCP. Many aren’t using agents, automated pipelines, or scheduled AI workflows. Some are running basic N8N automations, but that’s about it. The gap between what’s technically possible today and what the average marketer is doing with it is staggering.
This isn’t a criticism. The tools are moving fast, the learning curve is real, and most people are busy running their businesses. But it does mean there’s a significant window of advantage for anyone willing to invest the time to build these systems now. I wrote about how setting up Claude Code changed what’s possible in a marketing workflow, and everything I saw at the conference reinforced that.
The AI content ouroboros
This one kept nagging at me. If AI can research a topic and synthesise all the important information from the existing internet, why does a new AI-written article on that topic need to exist? Who’s going to read it? Other AIs?
It’s a circular loop. AI reads the web, writes new content, and other AIs read that content to write more content. The only thing that breaks the cycle is content with genuine original insight, first-hand experience, proprietary data, or a unique perspective that doesn’t already exist in the corpus.
Who wins in this world? Companies sitting on proprietary data. Thought leaders with recognisable voices and authentic points of view. People doing original research (like Dan Petrovic querying LLMs a million times).
Who loses? Anyone producing commodity content that AI can already replicate from existing sources. If your article doesn’t contain something new, something only you could have written, it has no reason to exist. That’s an uncomfortable truth for a lot of content teams.
Marketers need to think forward
I heard some version of “nobody could have predicted AI would get this good” multiple times. With respect: yes, you could have. Draw a line from ChatGPT’s launch in November 2022 to where we are now. The trajectory has been consistent and visible the entire time. Extrapolate three years forward. That’s where you need to be building for.
The marketers and companies who will do well are the ones thinking about where the technology is heading, not just where it is today. Auditing your site for AI agent readiness is one example of forward-looking work that pays off as the landscape evolves.
One more thing: WordPress clients are asking about AI
This came up in a conversation rather than a talk. I spoke to an agency owner who said all of his clients are now asking about AI features in WordPress. It’s becoming an important decision factor when choosing platforms, plugins, and agencies. The demand is coming from the client side, not just the tech side.
For anyone building WordPress products, this is worth paying attention to. Your customers are already thinking about it. If your product messaging and roadmap don’t address AI, you’re leaving a gap your competitors will fill.
Would I go again?
Without question. Ahrefs Evolve was well-organised, the speakers were strong, and the networking was genuinely useful. They’re running it again in May 2027 with an expanded two-day format, and I’ve already registered for the pre-sale.
The biggest takeaway I’m walking away with: we’re in a window right now where the gap between early AI adopters and everyone else is massive. That gap won’t last forever. The tools will get easier, the adoption curve will flatten, and the advantage will shrink. The time to build is now.