Built for Search, Arriving to Discover

---

Across millions of search interactions on employer career sites, 44% of all searches contain no search terms at all. Candidates arrive, leave the search bar empty  and the site returns every open role  – an undifferentiated list that leaves them with the burden of sorting through the results on their own.  

This is not a UX failure. It reflects a mismatch between how career sites have been designed and how candidates now approach job search. The keyword model assumes candidates arrive with a clear job title or role in mind, but a growing share do not know how to begin or have already begun their search elsewhere. 

Career Sites Were Built for Yesterday’s Candidate. 

The logic behind the keyword search model is sound. Candidates who know what they want – a specific job title, a location, a field – can get to relevant results quickly. The career site becomes a fast, reliable routing mechanism between intent and application. For these candidates, the model works well. 

The behavioral data from our network puts numbers on how that model performs in practice. Across four ways candidates engage with career site search, the friction gap between keyword search and browsing is visible in the session behavior. 

Candidates who rely on keyword search require nearly seven searches per session on average and produce the lowest apply click rates. They are iterating, refining and often failing to find what they are looking for before landing on a result worth clicking. Candidates who browse without entering any terms at all show higher apply rates and fewer searches per session. They are not using search as it’s supposed to function; however,  their conversion rate is higher. 

But for many candidates, that level of clarity is the exception. 

Our research shows that 77% of candidates conduct additional research outside the career site when key details are missing from job descriptions, turning to LinkedIn, Glassdoor and peer networks to fill the gaps. Eighty-eight percent require salary transparency before applying. And 55% decide whether to apply within 30 minutes of starting their search. 

Frontline and hourly candidates experience the same problem under stricter constraints. Forty-three percent make their apply decision in under 15 minutes. When information is missing or difficult to find, they leave. 

Neither group is struggling with job search. They are responding to an experience that does not match how they gather information or make decisions. They do not arrive to the career site ready to search; they arrive needing context. 

The zero-search data directly reflect this. Candidates arriving to the career site and starting their job exploration with no search terms are signaling that they came looking for something the tool was not built to provide. 

AI Has Formalized the Shift From Searching to Asking. 

The zero-search behavior predates AI, but AI has accelerated it, adding a new layer that changes what it means to “arrive” at a career site. 

Our research indicates that 77% reported using AI tools before visiting a career site. Asking an AI assistant about salary, culture or role fit is no longer a novel behavior;  it is often the first step. 

This behavior is reflected more broadly. A majority of Google searches now end without a click. When an AI summary is present, users click through to a website in only 8% of cases, compared to 15% when no summary is present. AI is no longer supplement to the job search journey. For a growing share of candidates, it is the first step in the journey. (Pew Research, 2025; SparkToro / Datos, 2024) .

That pre-visit behavior is now visible in our own network data. ChatGPT alone sent 62,800 sessions to our network of Career Sites over a 12-month period; the single largest AI traffic source and 77% of all AI-referred visits. Perplexity, Microsoft Copilot and others contribute thousands more. AI platforms are a real, measurable, growing traffic channel for employer career sites. 

What really matters, though, is how these candidates behave once they arrive. AI-referred visitors show higher engagement – 66% engagement rate compared to 57% across all sources, and a bounce rate of 34% versus 43%. But they have lower apply rates than the broader population: 24% versus 27% across all traffic. Overall, they explore more but take less action. 

The gap is expectation. AI answers are shaped by employer content, setting a baseline before the visit. When candidates arrive, they are looking to confirm and deepen that understanding. When the site does not provide sufficient detail, whether on compensation, responsibilities or culture, they will disengage rather than apply.  

AI-referral performance makes the same point from the opposite angle. When candidates find what they expect, conversion rates are significantly higher than traditional channels – one company found that ChatGPT referral traffic converted at nearly 16%, compared to 1.8% for Google Organic (Seer Interactive, 2025). The candidate intent signal is being obscured by whether the site meets expectations.  

Employers have long built career sites to perform in search, but now they need them to function as a content source for AI. Those are not the same requirement. 

You can’t control what AI tells candidates about your company. But you control the source it draws from. 

The Experience Candidates Want and the Content AI Needs Are the Same Investment. 

Candidates prefer career sites that provide context before asking them to act. In our AI Career Site Experience Study, 88% of participants chose a discovery-based experience over traditional keyword search, and 84% rated it higher than current career sites (Radancy AI Career Site Experience Study, 2025). That preference reflects how candidates approach job search: They want to understand what is available before narrowing their options. 

Participants responded most positively to comprehensive results – broad job categories, company context and role details presented before individual listings. Rather than starting with filters, they wanted the site to show what choices exist and how to evaluate them. 

This preference aligns with how AI systems process and return information. When models draw from structured, specific content, accuracy improves significantly rising from 16% to 54% in information tasks when using structured sources rather than unstructured content (Data.World, cited in Matharu / Oxford Comma Digital, 2025). This directly affects what candidates learn before they visit a career site and how they interpret what they see when they arrive. 

The same content that helps candidates evaluate roles on-site also shapes what AI systems surface off-site. When that content is incomplete or unstructured, both the on-site experience and the AI-generated representation break down in the same way: Candidates do not get the information they need to make a decision. 

Designing for discovery addresses both conditions. Context-rich, structured content allows candidates to understand opportunities without relying on keyword inputs and ensures AI systems can represent those opportunities accurately before the visit. 

What Talent Acquisition and Digital Experience Leaders Should Be Asking. 

The right question is not whether a career site is working. By the standard it was built to, most employer career sites are working exactly as intended. The question is whether it was built for how candidates actually behave today and whether it is positioned to serve the candidate journey that is increasingly becoming AI-mediated. 

That journey now begins before the site visit. Candidates arrive after consulting AI tools, often without a clear role in mind and with limited patience for incomplete information. The zero-search data, external research behavior and AI referral patterns all point to the same gap. 

The practical test is simple: 

• Does the site surface critical information without requiring candidates to search for it? 
• Does it provide context with listings? 
• Is the content specific enough that an AI system can accurately represent it? 
• Does the site confirm and extend what candidates have already learned? 

Those are discovery questions, not search questions, but most career sites were not built to answer them.