How AI Search Is Changing Marketing: Zero Click SF Session Notes

Changes observed at Profound Zero Click SF (AEO, GEO)

I attended Zero Click SF in San Francisco . Throughout the sessions, I repeatedly confirmed the awareness that AI Search is changing not only search but also the very way marketing is designed. I have summarized the key messages that emerged repeatedly at the event, focusing on the sessions.

The Era of 'Marketing Engineering,' Not Marketing – Profound Session

The keyword that ran through all 12 sessions of Zero Click was “marketing engineering.”.

What is Marketing Engineering?

Profound CTO Josh Blyskalspoke about the potential of the marketing engineering era, stating, “In finance, there were quants, and in data engineering, layers like Airflow were added, leading to rapid advancement in industries, whereas marketing has long relied heavily on human senses and judgment.”

Thanks to AI, we are now entering a stage where marketing can be designed into a system that monitors, diagnoses, executes, and iteratively improves.

Marketing Engineering?
- Designing a system where data automatically finds the answer and drives performance growth
- A process of detecting and judging via systems/agents -> executing -> measuring -> and reflecting the results back into system improvement

Things possible by using an agent

• Monitor dozens of competitors in real-time and automatically provide battle cards to the sales team
• Monitors thousands of web pages daily and automatically updates relevant content
• Sending contextually relevant, personalized outreach to 500 journalists

The person who makes all of this possible is the 'Marketing Engineer.' They are someone who transforms the team's good ideas into a system, automates it, and multiplies the capabilities of the entire team.

The Era of the Marketing Engineer

In the era of marketing engineering, the capabilities of marketers must also evolve. While the focus previously was on channel-specific execution, copywriting skills, and campaign management, the ability to orchestrate various tools, agents, data, content, products, and sales as a unified wholebecoming more important.

In fact, even in the SEO era, writing well alone did not guarantee results. Content, information structure, technical SEO, internal links, conversion flow, and web experience had to work together. In the era of AI Search, this aspect is becoming much more pronounced.

Now, marketers go beyond merely planning and producing content

• Organize the brand context and
• Connecting signals that AI can refer to and
• Predicting what answer will be generated from a given question and
• You must design a structure that enables verification and conversion after the recommendation.

From this perspective, I believe the capabilities of marketers with an engineering mindset will grow within the organization

How Marketing Engineering Works by Mike King (iPullRank, CEO)

AI Search is actually just search

AI search and traditional search use the same technology foundation (vector space model). However, AI search understands the meaning of a query and predicts the next word. It is not that there is a separate category called "AI Search," but rather that search has evolved. This does not render existing SEO knowledge meaningless; rather, it is an opportunity to redefine what a company means.

Understand Query Fanout

In AI search, when a user prompt is entered, the system breaks it down into dozens of subqueries. Each subquery is passed to a language model and a retrieval pipeline. AI search is a raffle. You must secure as many raffle tickets as possible; that is, you must be ranked in as many subqueries as possible.

Content topic examples	Audience Prompt	Persona	Needs	Expected query fan-out format	Query fan-out search term (keyword)
Project management tools	What is the best project management tool for a small remote team?	Non-technical startup founder managing 5–15 remote employees	Quickly compare options and select low-risk tools	Comparison, Price Overview, Feature Evaluation, Recommendations	Trello vs Asana: Remote Team PM Tools, Project Management Software Pricing, Easy-to-Implement PM Tools
health insurance	What is the best health insurance for freelancers?	Self-employed freelancers are cost-sensitive	Understand Guarantee Options and Prevent Overspending	Search, comparison, eligibility requirements, cost analysis	Freelancer Health Insurance Options, PPO vs. HMO, Self-Employed ACA Plan
CRM software	Which CRM is suitable for a small sales team?	Sales manager for an SMB with 5 to 10 sales representatives	Gain pipeline visibility without complex configuration	Tool comparison, feature analysis, recommendations by use case	CRM recommendations for small businesses, HubSpot vs Salesforce, simple CRM for sales teams, CRM price comparison
Electric vehicle (bottom part obscured)	Are electric cars practical in cold regions?	First electric vehicle buyer residing in Northern U.S. and Canada	Resolved uncertainty regarding performance (partially masked)	Empirical analysis, misunderstanding resolution, actual usage scenarios (partially redacted)	Electric vehicle performance in cold weather, winter electric vehicle driving range, cold climate battery, winter electric vehicle driving tips

Limitations of SEO software

Most SEO tools are based on search models from 10 years ago. They rely on simple language frequency calculations. Google has already been using semantic search since 2012, and all new platforms are hybrid. Rank trackers tell you that rankings fluctuate, but they do not know why.

• SEO as an ETL Pipeline: Most SEO tasks are actually a pipeline that extracts, transforms, and loads data.
• If you design this programmatically, you become a marketing engineer.

Content Engineering ≠ AI Spam

There is a difference between mass-producing content with AI (a bad version of Programmatic SEO) and utilizing data models to create meaningful content at scale. The pattern of AI traffic surging and then plummeting is not a problem with the AI ​​itself, but with poor UX.

Principles for Exploring the Future of Search by Daniel Kang (Expedia, Head of AI Search)

Certain Principles Amidst Uncertainty

In the last two years, there have been three headlines declaring that SEO is “dead.” We cannot predict the exact future. However, some outcomes are sufficiently predictable, and strategies can be built upon them.

The fundamental difference between humans and agents

• Humans are emotional. They possess goals.
• The agent is rule-based. It owns the execution.

Understanding this difference will change how you respond to each one.

Three principles to start with right now

1. Prepare agents as customers: You must configure APIs, file types, databases, and protocols so that agents can access them.
2. Discovery and evaluation methods have changed: The very way people find and review information is changing. Approach this from a portfolio perspective rather than trying to be the best in a single channel.
3. Asymmetric Investment: Not every strategy needs to work. Each failure provides information for the next investment. Conduct many experiments quickly to increase your chances of winning.

You don't have to predict the exact future. You just need to build upon certain results. Don't let uncertainty paralyze you.

LinkedIn Becomes a B2B Trust Signal by Dan Morrill (Linkedin, VP Marketing Tech)

The LinkedIn platform itself is a signal of trust

When B2B buyers ask professional questions to AI, LinkedIn is one of the most frequently cited sources. There are three reasons for this.

• Real Identity
• Real Expertise
• Real Conversation

AI systems interpret these three things as trust. This is the reason why LinkedIn citations across chatbots doubled this year.

Three Ways to Increase AI Exposure on LinkedIn

Content-based

• Company Page: Write product, leadership, and industry information clearly in natural language
• Consistent posting 2-3 times a week has a direct correlation with AI visibility
• Approximately 60% of the cited content is long-form. In-depth content such as articles and newsletters is cited
• 95% of the citations are original content. Reshares are not cited

A trustworthy voice

• AI does not look at content alone. It looks at trustworthiness. In B2B, trust lies with people.
• Posts with 10 or more comments have a significantly higher probability of being cited
• Empower employees, executives, and industry influencers. Authentic conversation is more important than brand messages

PLG in the Era of AI Search by Reema Batta (Figma, VP Growth Marketing)

How traditional PLG (Product Led Growth) worked

Figma's Product-Driven Growth (PLG) began with the elimination of friction. Collaboration starts the moment a link is clicked. The moment of activation was a feeling, not a checklist. It was the moment another person's cursor appeared on the canvas. It was a flywheel that designers loved first, and which then attracted PMs, developers, and marketers.

Flywheel: Easy to find → Fast value experience → Become a fan and recommend → Recommendations drive growth

Things that have changed due to AI

Since the introduction of ChatGPT in 2022, users no longer search by keywords. They ask questions. They do not click. They narrow down the context through repeated questions. They make a selection only when the agent makes the judgment.

A point has arisen where two important systems meet.

• Product Experience: You must answer “I want to use this”
• AI System: Must say, “This is the best solution.”

The point where the two systems meet is the moment of discovery.

Three Strategic Responses⭐️

1. Build Comprehensive Educational Content: People search for problems , not products . You must create content that answers the actual questions asked by each persona.

2. The issue of aligning products and features by persona varies from persona to persona. We have created tools tailored to specific use cases, templates to help get started, and features that connect with existing workflows.

3. Build Trust Everywhere. As the consideration journey in AI search is compressed, reputation itself becomes a signal. You must build trust not only on your own website but everywhere.

• Global Brand Campaign + Social Push
• Partnership with major AI model providers
• Global Make-a-thon operation
• Media and Creator Investment

AI Becomes a Major User of the Web by Parag Agrawal (Parallel, CEO)

AI becomes a major user of the web

When I founded Parallel, I believed that “the way AI uses the web will be the next paradigm.” Looking back now, even that imagination was too small.

What I learned from operating Twitter for 11 years is that the amount of service humans use is ultimately limited by their attention. Agents, on the other hand, are different.When an agent searches, reads, and explores the web, the limit is not attention, but the amount of value that can be generated from the web.

Humans perform 1 to 10 searches per day. With an advanced agent workflow, tens of thousands of searches are performed daily.

Agents are different from humans

• Human: Types search terms lazily. Picks the closest one from autocomplete. Uses short, vague keywords.
• Agent: Writes very detailed and technical queries. Expresses what is desired with extreme specificity. The search term distribution itself changes.
• What the agent is influenced by: Verifiable facts, authoritative sources. They do not respond to flashy UI or emotional copy.

It is still true that websites should work well for humans. However, optimization for agents must be designed based on different criteria.

Webflow's AI Search Playbook by Dave Steer (Webflow, CMO)

To win in AI search, consistency and integrated activities across channels are required. However, most organizations have teams split by channel. While agencies write copy they create, engineering deploys code different from it, and marketing optimizes content, the design team goes through three review cycles.

Webflow calls this structure a “Frankenstack.” It refers to the complexity resulting from the entanglement of teams, tools, and processes. In this structure, content cannot be managed consistently from end to end, which in turn makes scaling using AI agents impossible. (*Frankenstack is a portmanteau of “Frankenstein” and “Tech Stack,” referring to a complex and inefficient structure created by forcibly connecting various software, tools, and systems that are incompatible or lack interoperability.)

What is needed is a control tower that designs and operates the web experience.

Webflow's two playbooks

Content Playbook

1. Content Refresh: Update existing content with an agency-based workflow → 5x increase in content refreshes, 42% increase in traffic
2. FAQ: Written in “Answer-privileged” format + eliminated ambiguity with schema markup → 24% increase in impressions, over 300 new citations
3. Webinar Recycling: Agents extract key insights from webinars accumulated on YouTube → Convert into citation-ready articles → Publish to CMS. If scaling is required, apply in bulk to the entire webinar list using Profound Sheets.

Brand Playbook

1. Reviews: AI models utilize reviews from third-party platforms such as G2 as a high-trust signal. Reviews containing specific use cases and real-world language are key. Expressions such as “a better alternative” or “optimal for this use” are directly reflected in the AI ​​response.
2. Social & Video: Publish answer-style content on YouTube and TikTok in the format of “X vs Y,” “Optimal for this purpose,” or “Step-by-step method.” Leave subtitles, explanations, and comments as text.
3. Media : Target and focus on 10 to 20 key cited media outlets. Not only the Wall Street Journal and Forbes, but also industry-specific media are important.

Offsite-GEO Strategy – Reddit Case Study

Reddit is not social media

Many brands bring their Instagram or TikTok strategies directly to Reddit. This is the first mistake.

Social media is centered around 'people you know or people who are well-known.' Reddit operates solely on 'what you say.' Whether you are a Kardashian, a president, or a pope, on Reddit, you start from zero.

Reddit is not social media, a dinner party. The goal of a brand is not to build followers, but to be invited.

How Brands Become Good Customers

• Do Your Homework: What Is Homework? Understanding the Customer
• Bring something valuable: Throw out topics of value that users can discuss
• Make the community the protagonist: The brand is not the protagonist.
• This is not my home: Do not try to own a subreddit.
• Be humble

Practical Operational Principles

• Posting frequency: Maximum 3 times per week. Anything more than that will get you treated like someone who doesn't keep their mouth shut at parties.
• Reply Guy Ban: On Reddit, users who only comment on others' posts without the original text are called 'reply guys.' Merely editing without adding value is counterproductive.
• Target Subreddits: You need to cover approximately 350 related subreddits, not just 3 to 5 core subreddits. It is important to identify the long tail of subreddits that actually affect queries, along with Profound.

Core Attitude: Not “Look at me,” but “I see you.”

Offsite-GEO Strategy – G2 Case Study

The purchasing journey is diverging

G2's Recent Survey Results:

• Half of B2B software buyers now start their research on chatbots rather than Google .
• 54% cite chatbots as the number one tool for creating shortlists.
• Previously, it took 7 to 12 hours to browse BlueLink pages to create a spreadsheet and reduce it to a final three, but now it is finished with a single prompt.

G2 drove 1 million human visitors directly from chatbot traffic in 2025. Organic traffic increased year-over-year.

G2's Persona Building Framework

This is a persona building framework announced by G2. The key point is that personas should not be defined by demographic information (age, job category, etc.), but by the language the person actually usesand the context of their decision-making,so that AI can accurately respond to that persona's queries.

component	Key question	explanation
Job-to-be-done (goal to be achieved)	What are you actually trying to do?	Rather than “I want to learn X” or “I want to fix problem Y,” what the actual tasks in the real world are
Vocabulary (naturally used language)	What words and expressions are used?	“keeping customers” instead of “churn mitigation”, “making the site easy to use” instead of “UX optimization”
Constraints (time, risk, budget, tools, etc.)	What constraints are in place?	Time pressure, risk tolerance, compliance requirements, budget limits, tooling restrictions
Decision Criteria (Evidence to Trust and Act)	What is needed before believing and acting?	What level of evidence and detail is required to trust and put into action?
Success Metrics (Definition of “Good Enough”)	How do you define success?	Management wants emotional reassurance, while engineers want reproducible, concrete figures

conclusion

The recurring message at the recent Zero Click SF was clear: AI Search is not merely changing a single search channel, but is transforming the entire way brands are discovered, evaluated, and selected. So, what implications does this change have for the Korean market and B2B brand marketing?

In the next post, continue to summarize the Elifunt Team's perspective on this question.