How to actually get your brand into AI search results | Guillaume Ang @Psyke

In this episode of saas.unbound, Anna Nadeina talks with Guillaume Ang, co-founder of Psyke, about programmatic SEO, AI search visibility, hallucinations, and the human work that still sits behind effective agentic workflows.

Guillaume has spent the past 17 years building, exiting, investing in, and advising startups. With an engineering background and a business mindset, he has worked across B2B SaaS, grocery delivery, low-code growth consulting, and now AI-powered organic growth.

At Psyke, the mission is to help innovators spread faster by shortening the path from insight to traction. The company began by helping marketers manage the complexity of modern growth channels, then moved deeper into programmatic SEO and generative engine optimization, or GEO: the work of helping Google and large language models understand when, why, and how a brand is relevant.

Why programmatic SEO is not dead

Creating thousands of pages has a bad reputation. For many marketers, it immediately brings up concerns about low-quality content, Google penalties, and an internet flooded with AI-generated slop.

But programmatic SEO itself is not new. Large companies have used it for years. Major websites often publish tens or hundreds of thousands of pages because they serve a wide range of customer needs, locations, products, and use cases. What has traditionally made this approach difficult is not the number of pages. It is the work required to make every page useful, specific, and unique.

According to Guillaume, Psyke is publishing thousands of pages daily and seeing reported returns in the range of six to ten times the spend after several months. The critical distinction is whether those pages create genuine value or simply repeat information that already exists online.

Search engines and AI systems do not need more generic explanations. They need better context. A page that helps a system connect a particular customer situation to a relevant brand can be useful. A page that merely rewrites the same broad advice already available everywhere is noise.

The underlying principle has remained consistent throughout changes to search algorithms: create content that is genuinely valuable for a specific user need. For broader guidance on building helpful, people-first content, Google’s helpful content guidance remains a useful reference.

How AI search systems learn about your brand

AI search changes the mechanics of discovery. Large language models ingest enormous volumes of information, and their knowledge is continually refreshed. Guillaume notes that much of the content cited or discussed by these systems is relatively recent, which means brand visibility is no longer only about maintaining a static set of evergreen pages.

Every search query comes with context. A person asking for a recommendation may have a particular industry, location, business size, workflow, pain point, budget, or technical requirement in mind. If a company’s site only speaks broadly about itself, the search engine or LLM has to infer how that company applies to all those real-world situations.

That is an unnecessarily difficult job for the machine.

The purpose of high-quality programmatic content is to make that connection clearer. Rather than forcing an LLM to abstract a company’s relevance from a handful of product pages and case studies, a brand can publish useful content that directly addresses the contexts in which customers are likely to need it.

In practical terms, this means building pages around real customer intent—not manufacturing pages simply because a keyword tool produced a long list of terms.

The three ingredients of AI-friendly content

The strongest scalable content sits at the intersection of three sources of knowledge:

• Your brand: Your point of view, tone of voice, beliefs, product strengths, and understanding of the problem you solve.
• Your unique data: Information that only your company has access to through customers, operations, product usage, sales conversations, or internal expertise.
• What the LLM already knows: Existing public knowledge, common language, and the search queries people use to describe their needs.

The LLM should not be the primary source of the content’s value. It should help combine a brand’s proprietary knowledge with known market demand. This is what prevents scaled content from becoming generic.

For a marketer who is still building confidence with AI tools, the first stage is relatively accessible. An LLM can perform a broad website health check, identify basic content gaps, and suggest structural improvements. That can get a company meaningfully closer to a stronger organic presence.

The harder work begins after that. Going from “good enough” to genuinely differentiated requires deliberate preparation: identifying the information a company knows that competitors and public models do not.

Where to find unique content your competitors cannot copy

Most companies already hold valuable information that could become the foundation for original content. The challenge is recognizing it, organizing it, and using it responsibly.

Guillaume recommends setting aside regular time to ask a simple question: what does this company know that is genuinely unique?

Some strong sources include:

• Customer support and success conversations: Support tickets reveal the questions customers actually ask, the language they use, and the answers that solve real problems.
• Sales-call recordings: Sales calls show the objections, needs, comparisons, and outcomes that make a company distinct in the market.
• Product data: Product-led companies often have valuable metadata that can be abstracted by industry, location, use case, maturity level, or performance outcome.
• Internal methodologies: Processes, frameworks, benchmarks, and operational lessons can provide a perspective that does not exist in public content.

Once these sources have been organized into a usable knowledge base, they can be combined with keyword and query data from SEO platforms such as Ahrefs or Semrush. The goal is not simply to find search volume. It is to identify where a company’s unique knowledge can answer a real, specific question better than generic content can.

Use chunking to make content easier for LLMs to cite

Being useful to AI search systems also requires content structure. A page should be easy for both people and machines to understand.

One technique Guillaume highlights is “chunking.” Instead of producing long, unfocused blocks of copy, create self-contained sections of roughly 400 words that answer one clear contextual question.

Each chunk should establish:

• What issue is being addressed.
• The context in which the issue appears.
• The unique angle, data point, or experience the company can bring.
• A useful conclusion or recommended next step.

A page can contain several of these chunks. When they are specific and well structured, they are easier for an LLM to retrieve, summarize, and potentially cite in an answer.

Technical foundations still matter too. Clear page structure, relevant markup, and content that is easy to parse all support discoverability. But technical optimization alone cannot compensate for a lack of original insight.

Why hallucinations require a creator-verifier loop

AI can produce convincing content that is wrong. That is true for complex marketing topics and even for straightforward factual questions. As a result, scaling content production without a quality system creates substantial risk.

The right comparison is not an autonomous machine that works perfectly from day one. It is a new employee.

Employees need a frame of reference, feedback, correction, and time to learn. Agents do too. Training them can be frustrating, particularly when expectations suggest that an entire pipeline can be automated overnight. In reality, useful agentic workflows need iteration.

One effective pattern is the creator-verifier model:

1. A creator agent generates the initial content or output.
2. A separate verifier agent evaluates it for accuracy, relevance, quality, and adherence to the brief.
3. The output is revised based on those checks.
4. The loop continues until the work reaches an acceptable level of stability.

The important point is that the verifier should be distinct from the creator. The first agent is incentivized to produce. The second is incentivized to find flaws. This separation creates productive tension, much like bringing together product, sales, finance, and operations leaders with different priorities.

Different agent instructions can also create what looks like different personalities. An agent given conflicting goals may become overly cautious, hesitant, or defensive in its language. That does not mean it has human consciousness. It means the system is trying to resolve contradictory instructions and expressing that conflict in human-like terms.

The more meaningful change is the combination of humans and AI. People who grow up using AI as a daily thinking partner may approach research, creativity, and problem-solving very differently from previous generations.

The pivot from low-code SaaS to AI search visibility

Psyke did not start as an AI search company. In 2020, the team was building marketing software focused on low-code automation. The product gained traction quickly, reaching around 25,000 users within months.

There was a clear path to continue building in that direction. The company had users, feedback, and evidence of demand. But the team also recognized that AI would soon make many of the workflows they were creating much easier to build and replicate.

Rather than optimize for a short-term opportunity, the founders chose to reconsider the company’s long-term ability to deliver durable outcomes. They believed that marketers needed more than another platform: they needed a better way to create sustainable growth.

That led Psyke toward programmatic SEO at scale. Soon after, the team recognized that the same capabilities were highly relevant to GEO and AI search visibility. The pivot was difficult, requiring serious conversations with investors and a willingness to reinvest in a new direction.

The decision ultimately paid off. Psyke generated more revenue in one year after the shift than it had in the previous three years, earning the company the description of “dark horse” within its investor portfolio.

The SEO cost of a rebrand

A pivot is not only a product decision. It can reshape a company’s identity, brand associations, domain strategy, and search presence.

The company changed its name from Fluid to Psyke to better reflect its focus on human psychology, narratives, and the role of marketing in helping ideas resonate. The name also reflects a belief that human understanding should remain central, even as AI becomes more deeply embedded in growth work.

Still, rebrands come with a cost. Search visibility is increasingly connected to entities: the recognizable relationship between a company, its reputation, its offerings, and the topics it is known for. When a company changes what it sells or who it serves, it should expect a period of rebuilding.

Domain changes can be especially disruptive. Psyke operates on psyke.co, even though it now owns the matching .com domain. The company has not activated the .com because a domain migration could create unnecessary search visibility consequences.

For companies navigating a rebrand, Guillaume’s advice is to be intentional and patient:

• Expect a ramp-up period when the company’s offering changes materially.
• Build visibility around long-tail, highly specific queries before trying to compete for the largest category terms.
• Use unique data to establish credibility in a narrow niche.
• Invest in digital PR, social content, review platforms, communities, and other credible sources of brand mentions.
• Clearly explain the company’s evolution on its website so that both customers and AI systems can understand the transition.

LLMs may update their understanding of a brand faster than traditional search engines, but that does not eliminate the need for consistency. A company still has to repeatedly demonstrate what it stands for and where it is credible.

The one thing AI cannot own: accountability

AI can automate tasks, generate drafts, analyze information, and coordinate increasingly sophisticated workflows. But Guillaume draws a firm line around accountability.

If the answer to a missed outcome is, “An agent was supposed to do that,” then no one is truly accountable. Agents do not have the ultimate incentive to protect the company, own the consequences, or answer for a decision in the way a human employee or leader can.

That does not mean AI should be kept away from important work. It means companies should decide who remains responsible for the result.

The more useful question is often not, “Which jobs can AI replace?” It is, “How can someone redesign a workflow with AI in mind?”

For example, a business selling swimming pools could use AI merely to identify leads and automate outreach. Or it could rethink the customer experience entirely: identify properties without pools, visualize a pool in a prospective customer’s backyard, and send a personalized concept. The second approach is far more creative, differentiated, and potentially valuable.

The opportunity is not only automation. It is developing an agentic mindset that reimagines how work can be done.

Why personal influence has become a growth advantage

One of Guillaume’s most important lessons is that trust matters more than ever. When more people can build and launch products quickly, credibility increasingly comes from who is behind the product, who recommends it, and whether the people involved are visible and trusted.

For an engineer who once avoided social media and the idea of talking publicly about himself, building a personal presence required a mindset shift. Guillaume began posting consistently on LinkedIn and found that the impact was meaningful.

His approach was not simply to publish more. It began with finding a deeper reason to be visible: understanding what he wanted to be proud of, what belief drove him, and what he genuinely wanted to change in the world.

That clarity makes it easier to communicate regularly and authentically. A founder does not need to appear online in exactly the same way every day. AI-generated imagery, edited formats, and other production layers can be useful. But the ideas, intent, and conviction still need to be real.

For Guillaume, that conviction is simple: marketing matters, innovation should spread faster, and founders should feel confident enough to share what they believe.