Engineer Your Brand into LLM Memory

Why memory matters for marketers

LLMs store more than conversations. They build memory banks that include past interactions, user profiles, summarized events, and vector nodes representing people and information users engage with regularly.

When your customers ask an LLM a question, it pulls from this stored context to shape the answer. The brands and content embedded in that memory influence what gets recommended. AirOps research into AI citation patterns underpins many of the strategies Ross covers in this lesson.

Ross opens the lesson with a nostalgia exercise (dial-up internet, Blockbuster, burned CDs) to illustrate how memory works in humans. The same principle applies to LLMs: once a brand enters the memory, it shapes future outputs.

Google is now a destination rather than a portal. AI Overviews and AI Mode have permanently changed how traffic flows. Winning requires more than ranking in ten blue links.

E-E-A-T across every platform

Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trust) applies beyond Google. Every LLM benefits from the same signals.

Ross references a SparkToro study showing Google still dominates as the front page of the web. So following Google's quality philosophies and applying them everywhere is the right play.

What each signal looks like in practice

• Experience: First-hand opinions, proprietary data, author pages, in-depth analysis that proves you did the work.
• Expertise: Content created by people with credentials, deep knowledge, and real subject matter authority.
• Authoritativeness: Press mentions, user-generated content on platforms like Reddit, brand mentions on review sites (linked and unlinked).
• Trust: Positive reviews, positive UGC, and a strong overall brand sentiment when someone runs an analysis of your brand.

Actionable steps

• Create content that includes first-hand opinions. Hire reviewers, build an editorial team, or do the testing yourself.
• Use AirOps to get Insights into which URLs are influencing LLMs today. Export cited domains and build relationships with them to get your brand incorporated into high-performing content.
• This is similar to traditional PR outreach or backlink outreach, but with a key difference: the cited domains that matter to LLMs don't always have high domain authority. Some of the most-cited sites today have low traditional authority scores.

Commodity vs. non-commodity content

• Google distinguishes between two types of content. Understanding where your brand falls determines your content investment strategy.

Commodity content

Examples: "10 things to consider when buying running shoes" or "2024 kitchen trends every interior designer needs to see."

Google acknowledges that many brands have used AI to produce large volumes of this content at scale. Google says they don't want more of it.

Ross frames this with nuance: if you're a well-established brand, you probably don't need more commodity content. But if you're a new entrant, commodity content still matters because it addresses search intent fundamentals.

Non-commodity content

This is the content Google explicitly wants: breakdowns, original research, universities, case studies, reports packed with unique data.

Ross encourages you to think like an asset allocator rather than an artist. Evaluate each piece of content like an investment: what's the expected ROI, how risky is it, and does it deserve your budget?

Every brand's content portfolio will look different depending on lifecycle stage, budget, skills, team size, speed, and brand reputation.

On-site content strategy

Writing quality and citation rates

AirOps research found that ChatGPT prefers college-level writing. Content written at a college reading level was cited 21% more on ChatGPT.

Ross clarifies: this doesn't mean writing in MLA or APA format. It means producing well-structured, substantive content that signals quality to the LLMs.

Author attribution

Make sure the LLMs can identify who wrote each piece of content. Put author names and company names behind your content.

Ross cites NerdWallet as a brand that does author attribution well. Google cares about the expertise of the people creating content, especially in sensitive categories like finance.

Peer recommendations

People trust their peers over strangers. In B2B, this effect is even stronger.

Ask yourself how you can inspire your customers to say more and do more on your behalf.

Content types that generate backlinks

AirOps research on backlink generation by content type found that data roundups generate the most links. Journalists, writers, and bloggers link to collections of niche data.

Thought leadership content was the second-lowest link generator. It over-indexed in investment by brands but under-delivered because most thought leadership isn't rooted in unique insight or proprietary perspective.

High-performing content types: how-to content, glossary content, stats roundups, and tools (especially in the vibe-coding era).

Distribution strategy

Why off-site content drives most brand visibility

AirOps research found that roughly 85% of brand mentions in AI search come from third-party content, not brand-owned domains.

Brand-owned domain URLs were cited far less frequently than third-party content. The opportunity is clear: distribute your story across multiple domains.

Ross frames distribution as the biggest remaining differentiator. Getting content in front of the right audiences and into LLM training data requires more effort than creating it.

Reddit strategy

Ross outlines a specific Reddit playbook:

• Create a subreddit aligned with your industry or brand.
• Create content in that subreddit targeting your brand name and key terms.
• Work with your paid team to run ads that grow the subreddit. A larger subreddit signals authority, expertise, and trust to both Google and LLMs.
• Email your list to drive them to the subreddit.
• Build community through regular engagement and comments.
• Continue creating and distributing content.
• Check the SERPs and learn from what's working.

Reddit shows up more frequently for longer prompts. As LLMs transform simple questions into multi-word queries using stored memory, Reddit content gets cited more often.

Your buyers don't need to be active Reddit users for Reddit to influence their decisions. If Reddit discussions get indexed by Google and enter LLM training data, the brands discussed there get recommended when buyers ask AI for advice.

Across SaaS B2B, a small number of key subreddits consistently appeared in LLM citations. Run your own analysis to find which subreddits matter for your prompts and keywords.

YouTube as mispriced attention

YouTube citation rates jumped 84% in AI Mode.

Ross argues YouTube is mispriced attention right now. You can influence high-value prompts and queries by creating YouTube videos, and the window will eventually close.

In AirOps analysis, YouTube was beating three or more companies in almost every category at a significantly higher rate. Some queries showed YouTube beating all five companies in a category.

YouTube playbook

• Identify key topics that matter to your brand.
• Upload non-commodity educational content to your YouTube channel: how-to videos, tutorials, and guides.
• Run an audit using tools like AirOps to understand which YouTube videos you're competing against for high-priority prompts.
• Create Shorts that link back to longer-form high-value videos and target similar queries.
• Run paid ads on YouTube to grow your channel.
• Engage with creators to produce commercial-intent and high-CPC content.
• Embed videos on your website to reinforce E-E-A-T.
• Distribute across platforms. Short-form video is showing up more frequently in LLM citations.

Platform trends

Ross shares platform-level trends from AirOps data:

• LinkedIn is 162 times more likely to be cited in Perplexity. Gemini doesn't favor LinkedIn as much (possibly due to the Google/Microsoft dynamic).
• Medium.com citations have grown for six straight months. Quora has consistently declined.
• Reddit citations are dropping on Perplexity but are flat or rising on Gemini.
• YouTube citation rates jumped 84% in AI Mode.

Category creation: own the language your buyers use

The deepest level of influence is owning the language your buyers type into LLMs. If your audience uses terminology that only your brand uses, your brand is more likely to surface in the response.

Ross gives an example: if someone searches for "revenue intelligence platform" and your brand coined that term, the LLM will likely recommend you because your content dominates that phrase.

Most LLM users don't type simple queries like "what's the best CRM." They provide detailed, long-tail prompts with context about their company, role, needs, and frustrations. This is where category language matters most.

Ross describes experimenting with "send to AI" plugins on brand websites. These let users send blog posts directly to ChatGPT, Claude, Grok, or Perplexity. This builds familiarity: the LLM learns that a user engages with your content frequently, which may influence future recommendations.

Responsible influence: earning memory the right way

Ross closes with a reminder that the goal is to earn memory, not game it.

• Distribute content across surface areas authentically.
• Create content good enough that LLMs want to cite it.
• Build credibility directly in your niche.
• Be transparent with share buttons and give users genuine choice and consent.

The internet should be a place where LLMs find great information. Playing the long game responsibly is how you build lasting brand visibility in AI search.