Relevance Engineering: What Brands Need to Know

This piece was drafted by the brilliant minds of our content team: Ashley Wallace Jones, Jenn DeRango, Katie Weedman, Dina Magdovitz, and Grant Burfeind

AI optimization is the newest digital marketing buzzword. Actual experts and those of the armchair variety are hyping it across channels. On social media, that hype often takes the form of sparking controversy for engagement. A post declaring that “SEO is dead,” for example, will undoubtedly prompt users to voice their opinions — whether it’s adding value or not. And the more engagement a post gets, the wider its reach.

But tactics like this to reach a larger audience can erode trust and damage a brand’s reputation.

That’s why at PAN, our AI optimization services are grounded in expertise, not hype. And while the industry can’t seem to agree on what to call this new stage in search evolution — answer engine optimization (AEO), generative engine optimization (GEO), or SEO 2.0 — our team of experts agree that it’s not a “this-or-that,” but rather a “this-and-that.”

In the spirit of clarity (and ridding the marketing world of unnecessary acronyms), we approach AI optimization through a relevance engineering lens. This allows us to enhance digital discoverability holistically across paid, earned, shared, and owned media.

Before we really dive in, let’s be clear: It’s the wild west of AI search, and things are moving and changing fast. Platforms responsible for the tech (e.g., ChatGPT, Perplexity, etc.) aren’t sharing guidance or tips for optimization like Google Search has. So, while we feel confident in where we stand today, new versions and algorithm changes are happening with greater frequency, so it’s important to stay on our toes, and keep both short- and long- term developments in our purview.

What Is Relevance Engineering?
Emerging Trends in Relevance Engineering
How Do You Build Discovery and Relevance for Search Engines?
Relevance Engineering Solutions From PAN

What Is Relevance Engineering?

Relevance engineering is the practice of improving digital discovery via any search-enabled platform. This process helps connect people with information, experts, brands, products, and services that are relevant to them. At least, that’s how we approach it at PAN.

Traditionally, for a search engine, relevance refers to whether content on a webpage aligns with the searcher’s query and intent.

AI search relevance takes it even further.

AI Search Relevance & How AI Search Engines Work: The Basics

Understanding AI search relevance isn’t possible without first understanding how AI search engines work. Here’s a breakdown:

Training, Tuning, and Grounding

AI training is similar to onboarding a new employee: You share your company’s policies, products, and services so they understand how things work. In the same way, large language models (LLMs) learn by absorbing massive amounts of online information to build general knowledge, language patterns, and concepts. Which is why it’s so important to keep your brand’s digital footprint updated and accurate at all times.

AI fine-tuning comes next. Often completed by developers, tuning helps adapt pre-trained models to specific tasks or use cases. Think of it like the focused training that comes after new employee orientation to help new hires learn your company’s specific processes and procedures so they can perform their jobs effectively.

Then comes grounding, which connects LLM outputs to external data sources. The information used for initial training offers a snapshot in time of general knowledge, but knowledge and time are always progressing. When browsing is enabled on tools like Perplexity and ChatGPT for example, AI grounding retrieves external sources in real time, which can improve factual accuracy.

Within the context of the onboarding metaphor, you can think of grounding like you would a manager vetting data pulled by an employee for an important pitch or presentation. For AI search engines like Perplexity, Bing Copilot, and Google AI Overviews, grounding can trigger online source citations. These sources are where brands should aim to appear with optimized messaging to more accurately inform LLM perception.

Takeaway: Keep your brand’s online footprint as accurate and up to date as possible and focus on appearing within online source citations.

Natural Language Processing

Natural language processing (NLP) helps AI search engines understand, generate, and harness human language. While today’s leading LLMs often rely more heavily on transformer architecture and embeddings than “classic” NLP pipelines, it can come into play via:

Semantics – You know that moment when you ask someone a question and they get what you’re driving at, even when you didn’t explain things perfectly? That’s the magic of semantic understanding. It’s about grasping the deeper meaning behind the literal words the user inputs.
Patterns and Context – Conversations flow more naturally when people can pick up on subtle cues and context. AI-enabled systems that are good at this can keep track of the discussion, remember earlier details, and understand how each new message connects to the bigger picture. It’s like having a highly attentive, on-demand conversation partner. And it’s why having a deep understanding of target personas is critical for successful relevance engineering.
Entity Recognition – When you inquire about “visiting Paris” or “problems with my iPhone,” AI search engines can recognize that Paris is a city and iPhone is a specific product. This helps create more relevant responses. It’s also how AI search engines recognize people and brands.
Intent – Sometimes what we ask for isn’t exactly what we need. Someone searching for an answer to “How do I cook pasta?” might actually want a complete beginner’s guide to making dinner, not just instructions for boiling water. Intent is about reading between the lines to understand the underlying need or goal. This is how AI search engines surface truly helpful answers for the user. It’s also key to getting your content referenced as an “answer” to a query.to a query.

Takeaway: Keep semantics, your target personas, and intent top-of-mind when crafting any form of content. That way, your entity, or brand, is recognized within meaningful context and surfaced for the right users.

Knowledge Graphs

Knowledge graphs, or semantic networks, help some AI systems, like Google Search, organize structured facts. LLMs may also benefit from similar structured relationships, but they often lean more on embeddings (more about embeddings later).

Each knowledge graph consists of nodes (which represent entities, things, or concepts), edges (the connections between nodes), and labels (which describe the relationships and reasoning rules on those edges).

For example, a node could be a B2B brand-to-demand agency, like PAN, as well as a client brand. The edge and label would be the customer relationship.

Takeaway: Relationships matter most. Make sure your brand shows up among your competitors, is represented positively, and stands out with accurate, strategically crafted messaging that highlights strengths and differentiators.

Vector Databases and Embeddings

Vector databases are collections of data stored as mathematical representations called embeddings. Put simply, they organize information based on similarities, so the system can quickly find things that are related in meaning, not just exact wording. Many AI-enabled systems use these databases and embeddings to make information retrieval faster, better understand language nuances, and determine the most authoritative sources on topics. This can help AI search engines see the relationships between different concepts — like how “rain” connects to “umbrella” or how “hungry” might lead to “restaurant.”

Understanding these relationships is key for your website content, too. When your content clearly shows how ideas and concepts connect, AI systems can better interpret your pages, and users can find the information they need more easily. Leaning into comprehensive, well-researched pillar page content can provide greater context about relationships between concepts. On your brand’s website, you should also:

Build Topic Authority – Think of your content like a neighborhood. Create pillar pages on main topics, then cluster related articles around them. AI search engines prefer websites that thoroughly cover a subject from every angle.
Write Like a Helpful Human – Naturally weave in synonyms, answer common questions, and dive deep into topics. AI search engines reward content that genuinely helps people and adds value for users.
Make It Easy for AI Search Engines to Understand – Use clear headings, bullet points, and structured data. Think of these as road signs that help AI search engines navigate and extract key information from your content.
Connect the Dots – Link related articles together with descriptive anchor text, and don’t be afraid to cite authoritative sources. This shows AI search engines how your content fits into the bigger picture while reinforcing credibility.
Optimize for Real Conversations – People ask AI search engines questions the same way they’d ask a friend — even when using to research B2B products and services. Write content that naturally answers “What’s the best…” or “How do I…” queries.
Stay Current – Like we said before, LLMs and AI search engines keep evolving, and so do user expectations. Regularly update your content so your brand stays relevant and is accurately portrayed.

Takeaway: Be like Mr. Rogers. Create a helpful, easy-to-understand neighborhood of trustworthy, topically related content on your website. And keep that neighborhood current with the latest information about your brand.

Emerging Trends in Relevance Engineering

With a basic understanding of how AI search engines work, you — much like an LLM — can leverage greater context to assess emerging trends in relevance engineering, such as the following:

LLMs.txt

LLMs.txt has been floated as a possible web standard for providing crawlable, AI search engine-friendly content. However, adoption is virtually nonexistent today, and no leading AI search platforms currently rely on it. Robots.txt files and XML sitemaps already serve much the same purpose for web content crawlability. That said, better understanding and even experimenting with LLMs.txt may help content owners better structure their websites for AI accessibility and prepare for future standards that could impact search and content visibility.

Query Fan-out

This has generated a lot of buzz, largely stemming from Google Search announcing its usage in its AI Mode. Looking closer at the implications, we believe there are two key takeaways regarding query fan-out:

Leverage a deep understanding of intent to craft content that anticipates follow-up questions.
Aim for strong subject matter authority by thoroughly covering your main topic and connecting it to related subtopics — similar to how topic clusters are used in traditional SEO.

Content Chunking

Content chunking, or breaking down large pieces of information into smaller, more bite-sized sections, is not a new concept. Content began shifting to a more “snackable” structure in the early 2010s, driven by the explosion of mobile browsing and social media, responding to how people access and consume information. Today, chunking remains highly relevant: because of query fan-out, how knowledge bases work, and how human readers find smaller paragraphs and sections easier to digest. It makes practical sense for both AI systems and human audiences alike.

LLM Perception Match

The theory of LLM perception match is that an LLM builds a perception of your brand, what you offer, and how well you fit the user’s need before it’s ever matched to a query. When you consider how AI search engines use entity recognition, user-provided context, knowledge graphs, vector databases and embedding, the theory — while speculative — seems plausible. While not a proven mechanism, it is a useful lens for thinking about how brands may be surfaced in AI responses — or not.

At PAN, we’ve been conducting AI search visibility audits via Semrush’s Enterprise AI Optimization (AIO) tool. Each audit acts like a Strengths, Weaknesses, Opportunities, and Threats (SWOT) analysis of a brand from the perspective of a potential buyer using AI search engines to find solutions. Insights are often drawn from citations that reference a brand and/or its perceived competitors—which sometimes might not be the ones you’re keeping an eye on.

These audits can reveal a need for refinements to how products and services are messaged, delivered or even priced. We believe this is because LLM sentiment and perception often reflect User Generated Content (UGC) by way of real feedback and reviews from customers, prospects, and community discussions.

How Do You Build Discovery and Relevance for Search Engines?

That brings us to the how — as in how to do relevance engineering. Believe it or not, today’s top strategies focus on the following best practices, which really are a mastery of the basics:

1. Brand Messaging for Discovery

Brand messaging for discovery helps ensure your solutions are perceived as options in comparison to those of your competitors. If LLM perception match does, in fact, build a perception of your brand, what you offer, and how well you fit the user’s need before your brand is matched to a query, then your messaging should be strategically structured to provide that information up front — or you risk the LLMs drawing their own conclusions.

This means you should lean into literalness and specificity within the text of everything from your boilerplate to your website’s home and about pages. For example, rather than leading with taglines or differentiating statements, start your boilerplate with a sentence like: Brand is an ABC entity that offers LMNOP solutions to XYZ people.

ABC Entity – Ideally comprised of at least one adjective and one noun, this descriptor defines what type of organization you are (e.g., data analytics company, marketing agency, managed service provider).
LMNOP Solutions – This descriptor explains what you offer (e.g., SaaS products, digital marketing services, technology consulting services).
XYZ People – This descriptor denotes the ideal user of your solutions (e.g., operations managers, IT leaders, B2B marketing professionals).

While it might seem tempting, don’t stuff this first sentence full of keywords. And whatever you do, do not list all your products and all your ideal customers and all their job titles within all the industries or verticals you serve. To help stay on track, remember that the goal here is to provide universally or widely shared characteristics for high-level competitive categorization.

Remember printed telephone books? Think of these descriptors as yellow page sections that help potential customers discover new businesses. By using descriptors to make your brand discoverable, you can stand out by highlighting unique capabilities and by describing the specific industries you serve.

2. Search “Everywhere” Optimization

Those wondering how to get listed in AI platform responses should also consider where AI search engines source their online citations.

For example, Search Engine Land recently published a data analysis of approximately 8,000 unique citations across 57 diverse queries that revealed:

Model	Preferences	Top Sources	Takeaways
ChatGPT (GPT-4o) The Authority Seeker	Factual, reference-style, rarely cites user-generated content (UGC) and product pages	Wikipedia (27%), Reuters, Financial Times	Build authoritative presence: news, blogs, Wikipedia. Avoid overtly commercial content.
Gemini 2.0 Flash The Balanced Synthesizer	Mix of expert and community input	YouTube (3%), Blogs (39%), News (26%)	Prioritize in-depth content: blogs, news, YouTube.
Perplexity (Sonar) The Expert Curator	Expert reviews, niche authority	Blogs (38%), Review sites (e.g., NerdWallet), Product blogs	Target trusted review/editorial sites. Tailor by vertical.
Google AI Overviews The Broad Aggregator	Deep links from blogs, news, UGC (Reddit, LinkedIn)	Blogs (46%), News (20%), Reddit, Quora, LinkedIn	Maintain presence across blogs, forums, Q&A, and expert platforms. Invest in structured deep pages.

Based on the top sources, relevance engineering should be applied to web content published across paid, earned, shared, and owned media.

Technical SEO: Owned Website Crawlability and Schema Markup

Promoting AI search engine bot crawlability and understanding of your website’s content is essential step for relevance engineering. This can be achieved by:

Determining whether your robots.txt allows or blocks identifiable AI search engine crawlers (e.g., GPTBot).
Implementing Schema markup, a type of structured data that helps search engines better understand the content on a webpage.

On-Page SEO: E-E-A-T, Structure, and Intent

On-page SEO for relevance engineering includes:

Reinforcing E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness) at the page level with web content that demonstrates subject matter expert authorship and features external links out to highly credible sources.
Leveraging heading tags for keyword placements while nestling answers to common queries within web content.
Analyzing intent granularly and applying an advanced attention to detail when crafting web content intended to answer it.

Helpful Tool – B2B AI Overview Tracker
Are the terms you’re targeting affected by zero-click searches? Find out with our free, non-gated tool that tracks the appearance of Google AI Overviews for 5,000+ B2B keywords across industries like SaaS, cybersecurity, healthcare, and financial services.

Off-Page SEO: Online Reputation Management and PR

Because LLM sentiment and perception often stem from User Generated Content (UGC), reviews, and community discussions, online reputation management is paramount. This can include everything from having official Reddit users represent your brand to securing and promoting your awards, so your brand shows up in the “top” and “best-of” lists.

And don’t forget: your brand’s digital image needs to feel reliable and credible. The AI grounding that enables online source citations is typically used to incorporate timely and fresh content that is trustworthy and authoritative, often drawing from sources like earned media websites. And make no mistake, earned media mentions deeply influence AI-generated responses and content. Data from MuckRack shows 95 percent of cited links in AI responses come from non-paid sources, of which 85% are earned media.

Connecting AI outputs with real-world meaning and recent, verifiable sources requires supplying LLMs with specific and relevant information that goes beyond their static training data. Use first-party research, timely thought leadership, and personal anecdotes to help make sure AI-generated outputs reflect real insight.

3. Product/Service Strategy Alignment

The key takeaway here is that relevance engineering takes more than just messaging and SEO to be effective. Your brand’s products and services need to be the right fit for your target personas, and they need to meet or exceed expectations throughout the entirety of the customer lifecycle across every touchpoint.

4. Optimizing Content for Digital Discovery

Before you start optimizing content for robots, make sure you’ve said something worth repeating to humans. Strategic content begins with a story that’s true to your brand and meaningful to your audience. No amount of polish can make a vague message stand out.

At its core, your story should answer one question: how do you help people solve problems? Maybe it’s faster claims processing. Maybe it’s demystifying compliance. Maybe it’s finally giving that overworked IT team a break. Whatever your angle, the message should reflect real needs, not marketing Mad Libs.

But keep in mind: resonance doesn’t happen in a vacuum. It starts with empathy and an understanding of the lived reality of who you’re trying to reach. What do they value? What frustrates them? What keeps them skeptical? You can’t reverse-engineer connection. You have to build it.

That’s especially true in today’s climate. According to our 2024 Brand Experience Report, two out of three customers said they would leave (35 percent) or consider leaving (32 percent) a brand if it relied on AI rather than people for content and storytelling.

Get the Report – Is AI Taking the User Out of the Brand Experience? The Critical Role Humanity Plays in Your Branding
As brands accelerate toward a future brimming with AI innovation, marketers are tasked with scaling their productivity and efficiency (using AI). But in the AI age, what role do humans play? And what role must they play to maintain customer loyalty?

We asked marketers and customers those questions — and more — to better assess how AI alters the relationships between brands and the people they hope to reach.

This is where your brand’s subject matter experts (SMEs), first-party customer data, and tools like SparkToro become your superpower. When you leverage what your brand’s SMEs have learned, seen, and done — and when you use data to define your personas — your stories stop sounding like everyone else’s and start sounding like something worth paying attention to.

Even when discovery starts with a machine, your audience still gets the final vote. And they’re voting with attention, trust, and clicks. They’re also voting via the user-provided context they’re supplying the LLMs.

That means your story needs to start with your audience: the real people with real pressure, real goals, and real reasons to care. Skip the generic value props. Say something specific. Say something true. And say it like a human would (after all, that’s who you’re marketing to).

Then make sure that story shows up everywhere. On your website, in your thought leadership, through media mentions, and across social channels. AI search engines are pulling patterns from the whole online ecosystem. If you don’t show up with a consistent message, you’re giving them permission to fill in the blanks.

Winning brands are being discovered in AI search engines by communicating clearly, repeating differentiators often, and leaving no doubt about who they are, what they offer, and who they serve.

Takeaway: Start with the human. Build your story from real subject matter expertise, persona insights, and business context. Be specific, be consistent, and be unmistakably you.

Relevance Engineering Solutions from PAN

Like traditional SEO, relevance engineering is far from a quick win. It’s a long game.

LLMs spot patterns in content online. If you start shaping those patterns now, your brand is more likely to show up for the right users, in the right way, for years to come.

So why wait? The best time to start is today. From SEO strategy to content creation and media relations, many of the services PAN provides are relevance engineering solutions.

And if you’re curious about how you’re showing up in AI-powered search, our team is here to help! We evaluate visibility across platforms like Google AI Overview, ChatGPT, Perplexity, and more — and share strategic recommendations to help you stay discoverable as AI transforms search.

Ready to get relevant? Click here to chat with our experts.

What are you looking for?