By the Tech & Data Desk Published September 2026 When enterprise buyers turn to generative search engines and conversational AI to research business software, they expect an objective breakdown of their options. But a comprehensive new study from Foundation Labs reveals an uncomfortable truth for software vendors: ask an AI assistant to compare your product to a competitor, and the answer you get may rely heavily on third-party aggregators, independent review sites—or worse, a direct competitor’s marketing page. The study, which audited six major AI surfaces across 20 distinct buying prompts in the Applicant Tracking Systems (ATS) category, exposes a critical blind spot in modern digital marketing. As artificial intelligence fundamentally alters how business software is discovered and evaluated, companies that fail to control the "Versus Layer"—the web pages that explain product differences and pick category winners—are effectively surrendering the narrative of their own products to rivals and third parties. Main Facts: The Anatomy of AI Sourcing To understand how conversational engines assemble purchase recommendations, Foundation Labs focused on the competitive Applicant Tracking Systems market. Researchers tested four prominent ATS vendors—Greenhouse, Ashby, Lever, and Workable—running 20 buying prompts twice across six distinct AI interfaces: ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overview. The resulting dataset of 240 AI-generated answers yielded 2,695 inline source citations. The data unveiled several surprising mechanics of how AI surfaces evidence: Vendors Are Missing in Action: Across six direct head-to-head ATS matchups, the four featured vendors supplied a meager 27% of the visible source presence. The remaining 73% of evidence came from external sources, including competitor comparison pages, tech venture studios, and review platforms. The Dominance of Comparison Pages: Across all prompts, dedicated comparison pages and listicles accounted for 58.7% of surface-weighted source presence. On direct two-product inquiries, dedicated comparison pages spiked to 42.1%. Competitors Writing Your Narrative: In multiple instances, AI models utilized one vendor’s proprietary comparison page to outline a rival product’s weaknesses, trade-offs, and ideal buyer fit. The ChatGPT Outlier: While five of the six AI surfaces leaned heavily on listicles and comparison pages, ChatGPT proved to be an anomaly, drawing 77.3% of its inline citations from first-party product documentation, support pages, and pricing data. Chronology: How the Audit Uncovered the "Versus Layer" The research began with a foundational premise: the B2B buyer’s journey has shifted away from traditional search engine result pages (SERPs) toward synthesized answers generated by Large Language Models (LLMs). Phase 1: Pre-Audit Ownership Gaps Before running a single prompt, Foundation Labs audited the digital footprints of Greenhouse, Ashby, Lever, and Workable to assess how well-prepared each vendor was for head-to-head AI queries. The structural gaps were immediately apparent. Lever had published dedicated comparison pages for all three of its potential head-to-head matchups. Ashby and Workable maintained two dedicated pages each. Greenhouse, one of the market leaders, had zero dedicated first-party comparison pages for any of its matchups with Ashby, Lever, or Workable. Phase 2: Prompt Execution and Multi-Surface Testing Researchers deployed 20 tailored buying questions into the wild. Six prompts were direct head-to-head matchups (e.g., asking AI to weigh Greenhouse against Ashby for a mid-market company with an in-house recruiting team). Four prompts requested alternatives to a named vendor, while the final 10 served as neutral category queries addressing scalability, analytics, CRM capabilities, and administration. Running these prompts across ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overview twice produced 240 structured answers. Every citation was cataloged to track how content formats migrated as buyers narrowed their shortlists. Phase 3: The Data Aggregation When the results were aggregated, a clear pattern emerged regarding how AI shifts its source material based on the specificity of the query. For neutral, top-of-funnel discovery prompts, listicles dominated, capturing 51.1% of source presence while dedicated comparison pages captured just 12.6%. However, once the query introduced specific vendors, comparison pages surged to 33.8%, eventually peaking at 42.1% on direct head-to-head prompts. Supporting Data: Winners and Losers in the Versus Layer A granular look at the data highlights a paradox: having a high search engine optimization (SEO) footprint or high domain authority did not guarantee that a vendor’s page would be chosen by AI models as primary evidence. Traffic Metrics vs. AI Citations Consider the matchup between Greenhouse and Ashby. Ashby’s first-party comparison page boasted a relatively strong organic footprint—estimated by Ahrefs at 20 monthly organic visits, a URL Rating (UR) of 19, and six referring domains. Yet, Lever’s unranked Greenhouse-versus-Ashby guide (estimated at eight monthly visits, a UR of 0, and two referring domains) appeared in 27 answers across 14 prompts. Ashby’s own page appeared in just 18 answers. Similarly, Venture Harbour—an AI automation venture studio—published a Workable-versus-Greenhouse page that generated modest search metrics (11 monthly organic visits, 4.5 UR) but surfaced in 13 answers across five different AI surfaces. +-----------------------------------+--------------------+------------------------+ | Source / Page | Est. Monthly Visits| AI Surface Appearances | +-----------------------------------+--------------------+------------------------+ | Lever (GH vs. Ashby Guide) | ~8 | 27 answers | | Ashby (GH vs. Ashby Page) | ~20 | 18 answers | | Venture Harbour (Workable vs. GH) | ~11 | 13 answers | +-----------------------------------+--------------------+------------------------+ Platform Splits The study also proved that optimizing for one AI engine does not guarantee success across others. While platforms like Perplexity and Google AI Overview relied heavily on third-party comparison architectures, ChatGPT favored deep first-party product education. For instance, when ChatGPT evaluated Lever and Workable, its citations pointed directly to Lever’s internal CRM and candidate-nurturing pages, alongside Workable’s help documentation. Official Responses and Industry Insights The implications of these findings have sent ripples through the B2B marketing community, forcing leaders to rethink how they structure content for machine consumption. Ross Simmonds, CEO of Foundation Marketing, emphasized the fragmented nature of modern LLM optimization: "Every platform is going to be different. Every LLM shows links differently and will pull your content differently." Simmonds’ assessment underscores the danger of a monolithic SEO strategy. Because tools like Claude, Gemini, and Perplexity ingest external data differently, software vendors must diversify their content assets to cater to both third-party aggregators and direct product documentation indexing. Furthermore, the study documented instances where competitor pages actively shaped buyer perceptions of rival systems. For example, when Google AI Overview assessed Greenhouse’s reporting capabilities, it leaned on Ashby’s comparison page to note that Greenhouse handles standard dashboards well, but might require external BI tools for advanced analytics. Without a counter-balancing first-party page, Greenhouse had no direct mechanism to challenge or contextualize Ashby’s assessment within the AI’s generation pipeline. Implications: Building and Defending the Versus Layer For software marketers, product managers, and SEO strategists, the Foundation Labs study serves as a wake-up call. Winning software categories in the age of generative AI requires a deliberate strategy to reclaim the Versus Layer. Experts recommend a three-step framework for addressing these vulnerabilities: 1. Prioritize Based on Evidence Leakage Vendors must audit their categories to identify where buyer demand overlaps with weak first-party coverage and high third-party encroachment. Using metrics like an "Evidence Leakage Score," software companies should build pages targeting high-volume matchups where competitors are currently defining the narrative. For Greenhouse, this means prioritizing an Ashby comparison page to capture high search demand and neutralize external bias. 2. Optimize Page Architecture for Machine Extraction AI engines do not read comparison pages the way human casual browsers do; they parse structured evaluations to extract verdicts quickly. Content must clearly define target buyer profiles, direct comparison criteria, trade-offs, pricing structures, and limitations early in the document. Pages that bury recommendations behind generic marketing language or gated downloads consistently underperform in AI citation metrics. 3. Audit Competitor Comparison Assets When a rival’s comparison page is frequently cited in AI answers discussing your product, marketing teams must reverse-engineer that asset. By analyzing the specific limitations, trade-offs, and capability gaps cited by competitors, organizations can publish objective, authoritative first-party content that addresses those pain points head-on—ultimately coaxing AI engines back to owned properties for source validation. The Bottom Line As conversational interfaces increasingly mediate the relationship between B2B buyers and enterprise software, neutrality is no longer safe. As the data shows, if you do not supply the evidence for how your product compares, somebody else will. Post navigation The New Era of Co-Authorship: LinkedIn Launches Collaborative Posts The Science of the Click: How Headline Trends Have Evolved Since 2017
