93% of AI-Mode Searches Now End Without a Click: What That Means for Your Careers Page
The numbers landed and they're worse than most employer-brand teams expected: 93% of searches in Google's AI Mode end without a click to an external website, according to Semrush data covering the back half of 2025 into early 2026. Across all search — traditional results and AI summaries combined — roughly 60% of queries now end without a click.
ChatGPT alone accounts for around 20% of search-related traffic globally (12% in the US). The total search pie has actually grown — combined traditional-plus-LLM search usage is up 26% worldwide — but an increasing share of it ends at the answer page. GoodFirms' 2026 practitioner survey found 65% of marketers name AI-driven search as their single biggest challenge this year.
For employer brand teams, this is a quiet crisis. Candidates are still asking the questions — "what's it like to work at [company]", "does [company] pay well", "is [company] hiring" — but the click-through to your careers site is gone. The candidate makes a decision based on the AI summary and either applies, moves on, or skips you entirely. You never see them in analytics.
What zero-click means in practice
A decade ago, a strong careers site gave you a measurable funnel: impressions → clicks → page views → applications. Every layer had a conversion rate you could optimise. Today, two of the middle layers have partially collapsed:
| Stage | 2015 | 2020 | 2026 |
|---|---|---|---|
| Search impression (your content surfaces) | Yes | Yes | Yes, but often rewritten |
| Click-through to your site | Dominant | Declining | Mostly absent in AI Mode |
| Full page read | Dominant | Declining | Rare |
| Application | Stable | Stable | Stable-to-growing |
The first and last stages are intact. The middle two are partially replaced by AI summarisation. Candidates form their opinion of you on the answer page, not on your careers page.
Three things this breaks
1. Analytics blindness. You can't measure zero-click impact directly. Your careers-page sessions will decline without the underlying candidate interest declining. Most employer-brand teams are reading their own traffic graphs and drawing the wrong conclusion.
2. Content ROI inversion. The long-form careers content you wrote to convert visitors is now mostly being read by AI crawlers — not humans. The conversion path has shifted: convert the AI first, then the AI converts the candidate.
3. The employer-brand narrative moves. If the AI summary gets you wrong — out-of-date salary bands, missing benefits, a stale Glassdoor quote pulled above your own careers copy — that's now the dominant representation of you. The candidate won't dig.
What "getting cited" looks like
The new primary goal is citation in AI answers, not traffic to your site. This is called AEO (Answer Engine Optimisation) or GEO (Generative Engine Optimisation) depending on who you're reading. The mechanics are the same:
- Your content is structured so AI can extract it reliably.
- Your data is current, accurate, and verifiable.
- Your site makes source attribution easy.
The concrete upshot: AI needs machine-readable facts, not marketing prose. A careers page that reads like a polished brochure converts badly for AI extraction. A careers page that surfaces salary bands, benefit amounts, interview stages, and team structure as explicit, structured data converts well.
The AI-first careers page: a structural checklist
A careers page optimised for AI citation looks different from one optimised for human conversion. Both are needed, but the AI version requires specific structural moves:
1. Ship structured data (JSON-LD)
Every careers page should emit JobPosting schema for every live role, plus an Organization schema on the careers landing page. Use OpenRole's employer schema generator if you don't have internal tooling.
Required fields for JobPosting: title, description, datePosted, validThrough, employmentType, hiringOrganization, jobLocation, baseSalary (this is where UK employers mostly fail).
2. Publish salary ranges in a parseable format
58% of audited UK employers don't publish salary in any machine-readable format. Include salary as both visible copy and JSON-LD, e.g.:
"baseSalary": {
"@type": "MonetaryAmount",
"currency": "GBP",
"value": {
"@type": "QuantitativeValue",
"minValue": 60000,
"maxValue": 80000,
"unitText": "YEAR"
}
}
This single change typically moves an employer's salary-related AI answers from "hallucinated estimate" to "sourced from official careers page."
3. Add an llms.txt file
A short plain-text briefing at /llms.txt is the emerging convention for telling LLMs what matters on your site. It's what robots.txt was for search crawlers, but for comprehension. Include: key pages, current salary bands, top benefits, recent company news. See the llms.txt guide for the format.
4. Reshape careers copy into answer blocks
Candidates ask AI questions. The AI answers in short, declarative paragraphs. If your careers copy is written as flowing marketing prose, the AI will paraphrase and lose precision. If it's written as answer blocks — "What we pay: [specifics]. Our interview process: [specifics]. Benefits we offer: [list]." — the AI can cite you verbatim.
A test: grep your own careers page for the questions a candidate might ask. If the answer sits inside a long paragraph, rewrite it as its own section with a clear heading.
5. Publish an FAQ as an FAQPage schema block
FAQs are high-yield. AI models lean on FAQPage schema heavily for candidate queries. 20–30 questions and answers covering compensation, interview process, remote policy, manager ratio, team structure, and "a day in the life" is a realistic target.
6. Make your bot access rules explicit
Check robots.txt. Crawlers to explicitly allow: GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Bytespider, OAI-SearchBot. Blocking these (often the default with security plugins) is the single fastest way to become invisible to AI.
7. Version your facts with "current as of" dates
AI models heavily weight content recency. Adding an explicit "Current as of 15 April 2026" line at the top of key pages signals freshness and helps AI decide whether to cite you over older third-party sources.
8. Control the Glassdoor / Comparably drift
Third-party review sites appear in AI answers about you far more often than your own content does — because they're structured, the review text is citable, and the content is regularly updated. You can't delete bad reviews, but you can respond (publicly), publish comparable structured content yourself (employee quotes, team metrics), and raise your own SEO weight on employer-specific queries.
What changes on the metrics side
Your scorecard needs to change with the channel. The old metrics (careers-site sessions, time on page, application form completions) are still valid but partial. Add the new layer:
| Metric | What to measure |
|---|---|
| AI citation share | % of AI answers about you that link to or name your careers page, vs third-party sources |
| Fact accuracy rate | When AI answers with specific salary/benefit/process data, how often does it match your actual data? |
| Hallucination rate | How often does AI invent facts about you (wrong salaries, non-existent programmes)? |
| Coverage breadth | Across ChatGPT / Perplexity / Claude / Gemini, how many have a reasonable answer when asked about you? |
These are the core metrics in OpenRole's AI Visibility Score. If you're not tracking them, you can't manage them.
The deeper shift
There's a bigger point underneath the tactics. Careers pages have spent two decades being optimised for humans scanning them for 45 seconds. The next decade, they're being optimised for AI systems extracting structured facts in milliseconds and representing you to a candidate who never sees your page. That's not a change in tactic; it's a change in audience.
Employer-brand teams that treat "getting cited" as the primary conversion event — and treat traffic as a secondary indicator — will compound. Teams that keep optimising for 2019-era sessions will watch their measurable funnel shrink while wondering why applications are still flowing.
Frequently Asked Questions
Q: What is zero-click search?
A: A search query where the user gets their answer directly from the results page — via an AI summary, a featured snippet, or a knowledge panel — and never clicks through to an external website. In Google's AI Mode specifically, 93% of searches now end without a click.
Q: If candidates aren't clicking through, how do they still apply?
A: They get the "should I consider this company?" question answered by AI, then go directly to job-board listings or search the role title, which takes them to your application form without the careers-page detour. The application step survives; the browsing step collapses.
Q: Is traditional SEO dead?
A: No. Traditional search is still growing in absolute terms — the pie got bigger. But ranking first in traditional results now often means "being the source AI summarises from" rather than "getting the click." You still need SEO fundamentals; you also need AEO on top.
Q: How do I measure AI citation share?
A: Manually, run 20–30 candidate-intent queries about your company across ChatGPT, Perplexity, Gemini, and Claude. Record which sources each AI names or links to. Automated tools like OpenRole's AI visibility audit do this at scale across a larger prompt set.
Q: What's the single highest-leverage change I can make this week?
A: Publish salary ranges on every careers-page role as both visible copy and JSON-LD baseSalary markup. Salary accuracy is the single category where most UK employers lose AI citations to third-party guess-work, and shipping it changes AI output within a crawler cycle (often 1–2 weeks).
Q: Does AI search replace job boards?
A: Not yet, and not simply. AI search changes the pre-application phase — how candidates decide whether to care about your company. Job boards still dominate the application phase. The channel closest to disappearing is the careers-site browsing in between.
Run a free AI visibility audit to see which AI models already cite you, which don't, and what they're currently saying.
Related reading: