Beyond the Visibility Mirage: Why High AI Mention Scores Are Misleading B2B Brands

By Ethan Crump
Published by Foundation Labs

In the rapidly evolving digital landscape of 2026, marketing analytics departments have latched onto a seductive new metric: AI visibility. Dashboards now flash impressive percentages, assuring CMOs that their brands are dominating the conversation across major generative search engines and conversational AI platforms.

However, a groundbreaking empirical study conducted by the Foundation Labs team in September 2026 reveals a stark and uncomfortable truth: high AI visibility is a mirage.

When brands measure success purely by how often they appear in AI-generated answers—their "presence score"—they ignore the critical variable that actually drives pipeline and revenue: the final recommendation.

Main Facts: The ClickUp Case Study

To understand the widening chasm between brand visibility and actual buyer conversion in AI search, Foundation Labs conducted a rigorous multi-platform audit. The team selected ClickUp, a heavyweight in the productivity and work management software space, as the subject of an exhaustive text-and-context evaluation.

The findings upend conventional wisdom surrounding Generative Engine Optimization (GEO):

• The Presence Mirage: Across 120 discovery-phase prompts where the brand was not explicitly named, ClickUp appeared 117 times across ChatGPT, Claude, Gemini, Perplexity, Google AI Overview, and Google AI Mode. This translated to a staggering 97.5% presence rate.
• The Recommendation Void: Despite omnipresence in discovery queries, when the research team asked the exact same AI platforms which tool a 100-person company should choose between ClickUp, Asana, and monday.com, ClickUp received zero out of 12 default recommendations.
• Model Drift: The narrative surrounding a single brand shifted wildly depending on the AI surface queried, driven by fragmented source ecosystems that include product-owned documentation, editorial reviews, and deep-dive community threads on Reddit and YouTube.

For modern digital marketers, the takeaway is clear: being present in the AI conversation means very little if the AI is actively steering your prospective customers into the arms of your competitors.

Chronology of the Audit: How Foundation Labs Uncovered the Disconnect

The research methodology was structured to mirror an actual buyer’s journey—moving from broad, unaided exploration down to high-intent, comparative decision-making.

Phase 1: Unaided Discovery (September 2026)

In the first phase of the experiment, Foundation Labs ran 21 distinct discovery prompts twice across six primary AI-driven discovery surfaces: ChatGPT, Claude, Gemini, Perplexity, Google AI Overview, and Google AI Mode.

Across 120 total answers generated where ClickUp’s name was omitted from the prompt, the platform successfully surfaced 117 times. By traditional metrics, any marketing team would pop champagne bottles over a 97.5% presence rate. The algorithms clearly recognized ClickUp as a dominant player in the broader work-management ecosystem.

Phase 2: The Comparative Decision Stress-Test

Next, the researchers escalated the prompts from casual discovery to transactional intent. They asked the platforms to act as objective advisors for a specific buyer persona: a 100-person company deciding between ClickUp, Asana, and monday.com.

The illusion of dominance instantly fractured.

Across 12 targeted comparative runs, ClickUp was completely shut out, failing to secure a single default recommendation. While the models consistently praised ClickUp’s expansive feature set, deep customization, and advanced automation, they simultaneously flagged severe adoption risks.

In one notable ChatGPT evaluation, Asana was declared the winner due to superior adoption rates, portfolio visibility, governance structures, and a significantly lower administrative burden.

Phase 3: Shifting the Goalposts (Intent-Specific Prompts)

Recognizing that generalized questions about "team productivity" exposed ClickUp’s administrative complexity, the researchers adjusted the nature of the inquiries. They tested prompts focused specifically on software consolidation, reducing tech sprawl, and embedded AI capabilities.

The results flipped dramatically. When the prompt narrowed to consolidating project management, docs, chat, and time tracking into a single hub, ClickUp climbed to first place in 11 of 12 answers and secured a top-three spot across all 12. Similarly, when the questions centered on native AI execution agents, Claude hailed ClickUp as "the most aggressive on agents," removing general-purpose competitors from the running and pushing ClickUp to the forefront.

Supporting Data: The Mechanics of Model Drift and Source Mix

The research underscored a phenomenon known as Model Drift—the reality that the same brand, confronting the exact same buyer question, will receive radically different evaluations depending entirely on which AI surface is queried.

Surface-Level Discrepancies

Visibility was far from uniform across the web’s dominant AI engines:

• Google AI Mode ranked ClickUp in the top three for 19 out of 20 discovery runs.
• Claude, by contrast, placed ClickUp in the top three for only 11 out of 20 runs.

Category language also drifted. While ClickUp has heavily invested in branding itself as a "Converged AI Workspace," only ChatGPT and Claude systematically adopted that exact nomenclature. The remaining four platforms defaulted to conventional, legacy industry terms: project management, work management, or work OS.

Why? Because large language models synthesize their terminology from the aggregate web. If the phrase "Converged AI Workspace" lives almost exclusively on ClickUp’s proprietary properties, the models will default to the terms most frequently used by third-party reviewers, journalists, and users across the broader internet.

The Source Ecosystem Divide

The underlying data sources fueling each AI platform explain the massive variance in brand perception:

• ChatGPT relied heavily on product-owned documentation and official web pages.
• Gemini and Perplexity integrated a balanced mix of editorial and web references.
• Google AI Overview and Google AI Mode were heavily saturated with community-driven content, particularly from Reddit and YouTube.

For instance, Google AI Overview cited 24 Reddit links and 23 YouTube links in its answers. Google AI Mode cited 24 Reddit links and 36 YouTube links. Conversely, ChatGPT cited 20 Reddit links and zero YouTube links, while Claude cited neither.

This community-heavy source mix creates massive vulnerability. If real users on Reddit and YouTube complain about ClickUp’s steep learning curve, configuration density, and heavy administration overhead, Google’s AI surfaces ingest those qualitative complaints and penalize the brand during comparative recommendations.

Industry Implications: The $750 Billion Blind Spot

The findings published by Foundation Labs arrive at a critical juncture for enterprise marketing strategies.

According to fresh data from McKinsey & Company cited in the study, only 16% of Fortune 500 brands systematically track their performance in AI search. This digital oversight exists despite half of all consumers already relying on AI search interfaces, with McKinsey projecting that a staggering $750 billion in US revenue will flow through AI-driven discovery and recommendation channels by 2028.

Why Traditional SEO Metrics Fail

Most organizations approach Generative Engine Optimization (GEO) through the outdated lens of traditional SEO: chasing keyword rankings, managing basic meta tags, and counting raw brand mentions.

As the ClickUp teardown proves, a high mention count can easily coexist with a zero-recommendation reality. If a brand optimizes solely for visibility without auditing the sentiment, context, and comparative rationale generated by AI models, they are flying blind into a multi-billion-dollar paradigm shift.

Actionable Framework: How to Run a Model Drift Audit on Your Brand

To help marketing leaders look past vanity visibility metrics and diagnose their actual standing in AI recommendation engines, Foundation Labs outlined a structured, four-step audit framework:

1. Map Your High-Intent Decision Prompts: Stop testing generic brand awareness queries. Identify the exact comparative questions your prospective buyers are asking AI when evaluating solutions against your fiercest competitors (e.g., "Which tool is best for a 100-person company needing X, Y, and Z?").
2. Audit Across Multiple Surfaces Simultaneously: Do not rely on a single AI platform. Run your test prompts across ChatGPT, Claude, Gemini, Perplexity, and Google’s AI surfaces. Document the variances in ranking, tone, and descriptive terminology.
3. Analyze the Sourced Ecosystem: Look at the citation links underpinning the AI-generated answers. Determine whether the models are pulling from your gated product pages, objective editorial reviews, or unmanaged community forums like Reddit and YouTube.
4. Optimize for the Decision, Not Just the Mention: If AI models are steering buyers away due to perceived adoption friction, complexity, or implementation risks, build targeted proof points—case studies, simplified onboarding guides, and community advocacy campaigns—specifically designed to counter those algorithmic objections.

Conclusion

The era of AI visibility as a standalone key performance indicator is officially over. As generative engines increasingly serve as the ultimate gatekeepers between B2B buyers and software vendors, the brands that win will not be those that simply make the most noise in the background, but those that earn the active, unreserved recommendation of the algorithm when the final choice is made.

To read the full, unabridged teardown of Foundation Labs’ research, visit the Foundation Labs Model Drift Report.