The AI Visibility Crisis: Why 73% of B2B Software Comparison Data Comes from Outsiders—and How Brands Are Losing Control of the Narrative

By Foundation Labs
September 2026

Main Facts: The "Versus Layer" Blind Spot

When enterprise software buyers turn to artificial intelligence to compare competing products, they assume the answers are drawn directly from the vendors themselves. A recent comprehensive study by Foundation Labs reveals a startling reality: when asked direct head-to-head purchasing questions about applicant tracking systems (ATS), the four leading software vendors supplied a mere 27% of the visible citation evidence.

The remaining 73% of the data powering AI responses—spanning Google AI Overview, ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode—came from third-party sites, industry listicles, review platforms, and even rival competitors.

This invisible infrastructure of evaluation, review, and category pages is what researchers have dubbed the Versus Layer. As AI surfaces increasingly mediate software selection, brands that fail to populate this layer with their own first-party comparison pages are effectively ceding control of their product positioning to outside publishers, affiliates, and competitors.

Chronology: Unpacking the Foundation Labs ATS Study

To understand how artificial intelligence processes software evaluation, Foundation Labs executed a controlled study in September 2026 focusing on the highly competitive Applicant Tracking Systems (ATS) market.

The Setup and Methodology

Researchers targeted four dominant players in the mid-market recruiting software space: Greenhouse, Ashby, Lever, and Workable. These brands created six possible direct two-vendor matchups.

Even before running a single query, a stark asymmetry in content strategy was uncovered:

• Lever maintained dedicated comparison pages for all three of its head-to-head matchups.
• Ashby and Workable covered two matchups each.
• Greenhouse maintained zero dedicated first-party comparison pages for any of its matchups.

Researchers then constructed a rigorous testing framework consisting of 20 distinct buying prompts. Six queries were direct head-to-head comparisons (e.g., asking AI to contrast Greenhouse and Ashby for a mid-market company with an in-house recruiting team). Four prompts requested alternatives to a named vendor, while the final ten queries addressed neutral category questions regarding scalability, analytics, integrations, candidate experience, and administration.

These prompts were executed twice across six major AI surfaces—ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overview—resulting in 240 unique AI interactions and a dataset tracking 2,695 individual source presences.

How the Evidence Shifted

The study revealed that AI engines dynamically alter their source mix depending on where the buyer sits in the evaluation funnel:

1. Neutral Discovery: When asked broad, category-level questions, AI relied heavily on broad listicles and roundups (accounting for 51.1% of source presence), while dedicated comparison pages captured just 12.6%.
2. Shortlisted Evaluation: Once a specific vendor was introduced into the prompt, comparison pages jumped to 33.8% of source share.
3. Head-to-Head Matchups: On direct two-product questions, dedicated comparison pages dominated, surging to 42.1% of visible citation presence.

Supporting Data: What the Numbers Reveal

The Foundation Labs dataset provides granular insight into how different platforms consume content and why traditional search metrics do not correlate with AI visibility.

The ChatGPT Anomaly

While five of the six tested AI surfaces (Gemini, Google AI Overview, Google AI Mode, Perplexity, and Claude) relied heavily on comparison pages and third-party listicles (ranging from 59.1% to 89.4% of their source mix), ChatGPT operated entirely differently.

ChatGPT leaned heavily into first-party product documentation, help centers, and feature support pages. In the ChatGPT test, product and support pages accounted for 77.3% of visible inline source presence (with 63.8% coming directly from product pages), while comparisons and listicles made up a mere 3.3%.

Traffic Metrics vs. AI Visibility

A striking discovery from the study is that high conventional search engine optimization (SEO) metrics do not guarantee AI dominance.

• Lever’s Greenhouse vs. Ashby guide appeared in 27 AI answers across 14 prompts, despite Ahrefs estimating a mere 8 monthly organic search visits, a URL Rating of 0, and only 2 referring domains.
• Conversely, Ashby’s competing page boasted 20 estimated monthly organic visits and 6 referring domains, yet appeared in significantly fewer AI answers.

This disconnect proves that Large Language Models (LLMs) evaluate content based on structural clarity, extraction ease, and comprehensive coverage of trade-offs rather than traditional backlink authority alone.

Official Responses and Strategic Implications

As generative engine optimization (GEO) becomes a critical discipline for B2B marketers, the implications of the "Versus Layer" are profound.

Ross Simmonds, CEO of Foundation Marketing, highlighted the fragmentation of the ecosystem:

"Every platform is going to be different. Every LLM shows links differently and will pull your content differently."

How Competitors Shape Your Narrative

One of the most alarming findings of the audit was that AI surfaces frequently utilized a competitor’s comparison page to define a brand’s weaknesses. For instance, when Google AI Overview evaluated Greenhouse against Ashby, it cited Ashby’s first-party comparison page to explain Greenhouse’s reporting limitations—noting that deeper analysis might require external BI tools or spreadsheets. Because Greenhouse maintained no dedicated counter-page for the matchup, the narrative was entirely shaped by its rival.

Vendor-by-Vendor Breakdown

• Greenhouse: Despite massive brand equity and its "13 Best ATS Software" article capturing 38 answers on neutral queries, Greenhouse’s total lack of head-to-head comparison pages leaves its flank exposed to competitors like Lever and Ashby.
• Workable: Published dedicated pages for two matchups, but buried critical decision criteria behind demo-oriented messaging. As a result, third-party sites like Venture Harbour out-cited Workable significantly.
• Ashby: Covered its primary matchups but left a blind spot on Ashby vs. Workable, which was easily captured by aggregator sites like G2 and SourcrLab.
• Lever: Achieved full coverage across all three matchups but failed to achieve source dominance, as specialized third-party sites still out-cited Lever in specific head-to-head evaluations.

Actionable Takeaways: Building the Versus Layer

To regain control over how artificial intelligence portrays their products, software brands must systematically construct and optimize their Versus Layer. Foundation Labs recommends a three-step execution framework:

1. Prioritize by Evidence Leakage: Cross-reference buyer search demand with AI citation data. Focus first on matchups where commercial intent is high, first-party coverage is missing, and outside publishers currently dominate.
2. Optimize for Direct Extraction: Structure comparison pages so that AI models can instantly ingest buyer fit, decision criteria, pricing transparency, and structural trade-offs. Avoid locking essential comparison data behind top-of-funnel marketing fluff.
3. Audit Competitor Evidence: When a rival’s page is cited to explain your product’s limitations, analyze their claims. Publish authoritative, first-party documentation that directly addresses those decision points, then re-test the AI queries to reclaim ownership of the narrative.

Conclusion

The era of relying solely on organic keyword rankings is over. As AI engines assume the role of enterprise software advisors, brands that refuse to supply clear, structured evaluation evidence will find their market positioning defined entirely by their competitors.