How Freshpet Earned AI’s Trust: A GEO & AI Visibility Playbook

A familiar brand can rank well in traditional search and still be nearly invisible when a buyer asks an AI assistant for a recommendation.

That was the problem Freshpet encountered as AI Overviews and large language models began reshaping product discovery. In this on-demand webinar, Brittni Ratliff and Cosima Compton of Intero Digital join Steven Elwell of Freshpet to explain how the teams assessed the brand’s AI visibility and built a more deliberate generative engine optimization (GEO) program.

The discussion moved beyond generic advice about “writing for AI.” It covered the strategic, editorial, technical, and measurement choices that help machines understand a brand in context without sacrificing usefulness for human readers.

Here are the main lessons from the session. You can also watch the full webinar for the complete Freshpet GEO playbook and audience Q&A.

Why Strong SEO Wasn’t Enough For AI Visibility

Elwell said Freshpet entered the project with years of content, third-party coverage, and established visibility in the pet food category. He initially assumed that foundation would make the brand relevant to AI-generated results about fresh pet food.

It did not.

The gap showed why marketers cannot treat AI visibility as an automatic extension of rankings or brand awareness. Compton explained that LLMs try to understand industries, concepts, and relationships between entities, not simply reproduce a conventional search results page. A recognized brand may therefore be absent when the available content does not clearly establish how it relates to a user’s question.

Freshpet’s experience also illustrates why the first step should be observation rather than assumption: Ask the questions your customers are likely to ask, record what the systems return, and identify where your brand is missing, misrepresented, or supported by outdated information.

The Four Areas Marketers Should Examine

Intero Digital presented its framework and used the webinar to unpack several practical areas that influence AI visibility: relevance, authority, structure, and engagement.

• Relevance: Publish content that directly answers current audience questions and matches the intent behind them. Compton recommended reviewing pages regularly, especially when they contain changing statistics or other time-sensitive details.
• Authority: Give claims enough context and support to be trusted. That includes clear author credentials, citations to reputable sources, relevant industry coverage, directory listings, and third-party mentions that reinforce what the brand is known for.
• Structure: Make important information accessible to both crawlers and readers. Clean HTML, strong page speed, descriptive headings, concise passages, lists, tables, schema, and direct answers can all help systems parse a page.
• Engagement: Pay attention to how customers and experts discuss the brand across reviews, forums, social platforms, and industry publications. Those conversations can expose concerns and language that deserve a well-sourced answer on the brand’s own site.

The framework is most useful when these areas support one another. Compton argued that onsite content, SEO, earned media, and social strategy should reinforce the same priority topics instead of operating as disconnected channels.

Write To Be Useful, Understandable, And Quotable

One of Elwell’s clearest observations was that much of Freshpet’s legacy content had been written like a magazine article. It was designed to draw a person through a narrative from beginning to end. That may serve readers, but the key facts are harder for an LLM to extract when they are buried in long passages.

The answer is not to strip away brand voice or write awkward copy for robots. Instead, the speakers recommended organizing complex ideas into focused sections that can stand on their own. A page can still sound natural while using descriptive headings, direct answers, short paragraphs, and meaningful lists.

Elwell described the shift as writing content that can be quoted, not merely read. For Freshpet, that meant revisiting page templates and formatting rules as well as the words themselves. The team simplified elements of its blog presentation, cleaned up HTML, and made it easier to apply appropriate schema.

The webinar also surfaced an easy-to-miss technical issue: valuable content can exist on a page but remain difficult for crawlers to access. Freshpet found that product reviews and Q&A content delivered through JavaScript were not readily visible to LLMs. The takeaway is to audit the rendered experience, not just the editorial inventory. Check crawler controls, JavaScript dependencies, robots directives, page speed, and whether the most important answers are available in parseable HTML.

Use Relevant Third-Party Authority, Not Just Big-Name Coverage

The speakers drew a useful distinction between having backlinks and building contextual authority.

A mention in a major publication can still be valuable, but an industry-specific source may do more to establish a brand’s relationship to a specialized topic. Elwell noted that a credible pet industry publication can reinforce Freshpet’s expertise in ways a broad national outlet may not.

Ratliff added that the surrounding coverage matters. If every mention uses the brand name as anchor text and points to the homepage, LLMs receive less context about the specific subjects the brand should be associated with. Earned media works harder when the article itself accurately explains the brand’s expertise and supports a relevant onsite resource.

Treat Community Discussion As Research

Reddit and other community platforms matter because customers use them to exchange candid opinions and LLMs can draw from those conversations. But the webinar did not recommend that every brand jump into every thread.

Freshpet takes a listening-first approach. Elwell said pet nutrition can be as emotionally charged as conversations about baby food, so arguing with individual users would not serve the brand. Instead, the team looks for recurring questions or misconceptions that its experts can address with authoritative content elsewhere.

That distinction is important: Community monitoring is not only reputation management. It can become an input for content planning. When the same concern appears repeatedly, marketers can create a clear, evidence-backed resource that customers and AI systems can both find.

Build A Prompt Set And Prioritize The Gaps

Compton outlined a manual starting point for teams that do not yet use a dedicated AI visibility platform:

1. Build a representative set of at least 20 to 30 prompts covering the core questions in your market.
2. Run those prompts across systems such as ChatGPT, Gemini, and Claude.
3. Record whether the brand appears, which sources are cited, and whether the response is accurate.
4. Compare weak areas with the corresponding pages and evaluate relevance, authority, structure, and engagement.
5. Prioritize fixes for topics closest to revenue, conversion, or a major customer objection.

Freshpet’s program is much larger: Elwell said the team was tracking 475 prompts across the major LLMs and grouping them into topic categories. That structure lets the team spot clusters of weak answers, create content for related questions, and involve subject matter experts where their credentials can strengthen the response.

The point is not to copy Freshpet’s prompt count. It is to build a set broad enough to represent the customer journey, then organize it so the findings lead to specific work.

Measure Progress Without Pretending Attribution Is Simple

AI visibility does not map neatly to the old model of ranking, click, and conversion. Elwell explained that Freshpet’s primary site often sends customers to retailers, making direct revenue attribution difficult. Its subscription business is easier to measure because users can complete a transaction in a logged-in ecommerce experience.

The team watches referral traffic from LLM citations, downstream actions, and modeled relationships between AI exposure and direct visits. Elwell cautioned against getting too far ahead of observable clicks with projected attribution. That is a useful standard for any GEO program: Combine directional indicators, visibility trends, and business events, but be explicit about what the data can and cannot prove.

GEO Should Strengthen SEO, Not Replace It

During the Q&A, Compton said teams should not expect GEO improvements to create a negative relationship with traditional SEO. Keyword alignment, headings, crawlability, credible sourcing, and clear site architecture remain valuable. Better-structured pages may support both AI visibility and conventional search performance, even as zero-click behavior changes traffic patterns.

Other answers reinforced a practical approach:

• Update a useful legacy page when LLMs keep citing stale information rather than creating overlapping pages about the same subject.
• For complex topics, use a pillar-and-cluster structure so each subtopic has a clear home and supporting pages connect logically.
• Expect prompt type and freshness to affect whether an LLM relies on trained knowledge or conducts an active search.
• Use AI tools to critique content if helpful, but provide enough business, audience, search, and competitive context for the feedback to be meaningful.

Freshpet’s experience shows that GEO is not a single markup change or content format. It is a coordinated process: learn how the brand appears, make its expertise easier to verify, remove technical barriers, build contextual authority, and measure the outcomes with appropriate caution.

Watch the on-demand session for the speakers’ full framework, Freshpet examples, implementation details, and rapid-fire answers to audience questions.