Building a company at the start of the AI platform shift presents every founder with an armory full of double-edged swords. Every challenge may also be an opportunity.

The emergent field of AI-driven discoverability is one such balancing act. Traditional SEO playbooks are breaking, and entire new business models are at risk. A real evolution in how users and customers are finding information is well underway, as new disciplines are taking shape.

It is against this backdrop that many have begun exploring what a new era of discoverability will look like, and importantly, how companies can be proactive to capitalise on the shift. 

We invited Malte Ubl, CTO of Vercel and former Engineering Director for Google Search, and Jeremy Cabral, Co-founder and COO of Finder to unpack the topic with Airtree Partner Jackie Vullinghs. 

Have a listen to the full conversation, and read on for some further context, and our practical, step-by-step guide on where to get started.

How AI is changing the SEO game

Search engine optimisation (SEO) has long been about keyword intent, search volume, and ranking factors. In a similar vein to the historically opaque page-ranking algorithms, LLMs don’t rely on keywords alone. They are built to operate on meaning, concepts, and citations.

In many ways, this harkens back to the original foundations that SEO was built upon. Reputation and trust at scale, genuinely earned media, and authentic references matter. User intent is paramount.

The core functions of search are where things have shifted the most — a trend that is putting entire business models at risk, as zero-click searches skyrocket. According to analysis by ahrefs, AI Summaries in search have sent click-through rates plummeting by 34.5% for top-ranking pages.

On the other hand, the new crop of LLM-powered AI-assistants (Claude, ChatGPT, Perplexity, Gemini) are opening huge new opportunities to capture this intent. 10% of Vercel’s signups now come directly via ChatGPT, and Jeremy reveals that Finder now gets more bot traffic from LLMs than from Google’s web crawler ‘Googlebot’. “When OpenAI’s o3 model launched, the company’s traffic from LLMs tripled,” he said.

Unlike traditional search engines, most LLMs lack APIs for submission or sitemaps — meaning the old SEO playbook (volume, keywords, backlinks) no longer works on its own. This, even if the principles of AEO and SEO have a substantial overlap.

So how do you make sure you’re capitalising on these opportunities?

Technically speaking: How LLMs generate answers

To engineer a strategy for visibility, it first helps to understand a bit of how LLMs operate. 

“When an LLM tries to answer a question it does two things. First, there is some inherent knowledge in the models themselves,” says Malte. “This knowledge is potentially out of date because it can only be updated when there’s a new model run, which may happen twice a year, max.”

The second piece is tied to RAG (Retrieval-Augmented Generation), a core function of most LLMs that references authoritative knowledge within reach, but outside of its inherent and foundational sources of data.‍

“When the users ask a question, a search identifies interesting documents that talk very broadly about it, and feeds them in, in a hidden way. This prompts the LLM to answer a question with user input augmented by its own search.”

These retrieval pipelines rely on structured, crawlable, and semantically relevant content including individual webpages, high-authority blog posts, forums, social networks, reviews, and API references. 

Being surfaced in these pipelines means your information is more likely to be included in the model’s context window, which increases your chances of visibility and being cited in the model’s response.

How to make your company visible

The PESO model (paid, earned, owned and social) is a valuable framework for the question at hand. If you want to be discovered, you need to play the distribution game strategically. 

Malte calls this frontier content. “We want to be the author and source on certain topics,” he explains. "We want the LLM to understand that we are the source of this information, and that we are the original provider. That’s where we have been focused: making content that is truly valuable and novel, rather than a strategy driven by volume.”

Malte calls this frontier content. “We want to be the author and source on certain topics,” he explains. 'We want the LLM to understand that we are the source of this information, and that we are the original provider.”

“We want to produce a feedback cycle of links and quotes; a propagation of authority.”

Putting aside ongoing copyright debates, the Web 2.0 generation of sites has become a rich source of training data for LLMs, hungry for original and rich training data. 

Software engineering forum StackOverflow has felt some effects in the shifts of user behaviour, while social networks like Reddit and the entire corpus of YouTube’s library are undeniable goldmines. “Every single social media post where someone has a positive sentiment about your brand, and associates it with some concept, is helpful,” Malte says.

Not every LLM has access to every platform or network, however. Partnerships are being forged, just as lawyers are being deployed.“Reddit is only accessible to Google. Twitter? Only xAI. Instagram? Only Meta’s models,” Malte noted. 

Due to OpenAI’s partnership with Microsoft, optimising for the Bing browser has become more relevant than its relatively small market share has to date implied. The list of publishers which ChatGPT has public partnerships with includes the Associated Press, Axel Springer, Financial Times Group, News Corp, The Guardian and more.

This reality and the evolution of media means that the importance of Digital PR has also skyrocketed. Citing a study out of Princeton in 2023, Jeremy notes that including quotes and statistics from credible sources increased visibility by up to 40%. Where you may not yet have the clout or experts in-house, forging collaborations to these ends can be of real benefit.

The playbook for AEO readiness

So let’s get to the practical bit. What are the actions you can take today to ensure you’re well-positioned for visibility in LLMs and other newly-developed answer engines?

1. Set up your measurement framework

1. Leverage tools like Hall (AU), BrandRadar (AU) to track performance in AI and search interfaces. This will help you test how models cite you and where your content and brand name are showing up.
2. Check AI-referrer traffic in your website analytics, filter for LLM-specific user agents like PerplexityBot, ChatGPT, Anthropic, etc.
3. Set up alerts for brand mentions across sites like X, Reddit, Substack, and other UGC-heavy platforms.
4. Monitor changes in retrieval with incremental updates to your content.

2. Refresh your content strategy

1. Write with specific language and self-contained paragraphs that can stand alone. Avoid “context carry” across paragraphs.
2. Publish unique and proprietary information such as case studies, research, or primary insights from your business and field.
3. Write for citation, rather than conversion. Use FAQs, lists, frameworks, numeric comparisons, glossary-style definitions and benchmarks. 
4. Avoid content shortcuts and traps such as Notion-only docs, PDFs, and embeds from external sources such as Airtable.
5. Design content for RAG systems by adding timestamps, and updating your content regularly. Freshness plays a role in both hybrid search and hallucination bias

3. Distribute across LLM-relevant networks

1. Build a muscle for digital PR, and encourage backlinks from blogs, newsletters, and in expert roundups.
2. Build strategic partnerships and relationships with media, publishers and other organisations with sticky audiences.
3. Proactively seed content into platforms like G2, Capterra, Trustpilot, Reddit, and X by encouraging reviews, user stories, or AMA-style posts.

If you’d like to learn more on this rapidly-developing space you can follow Malte on LinkedIn, and Jeremy on LinkedIn or Substack.