Agent Readiness Score: Why 'AI Visibility' Isn't the Right Metric Any More

For the past two years, employer-brand teams have optimised around "AI visibility" — being surfaced, correctly, when a candidate asks an AI about their company. That was the right metric for an era in which AI was a research assistant: the candidate typed the question, read the answer, and decided whether to apply.

That era is already transitioning. In parallel with candidates using AI as a research assistant, AI agents are increasingly acting on candidates' behalf: auto-applying to roles, shortlisting employers, drafting tailored CVs, and in some cases conducting first-round conversations. Candidate-agent platforms (Massive, LazyApply, LoopCV, SimpleApply.ai and the 2026 generation) process dozens to hundreds of applications per candidate per month. On the employer side, sourcing agents and AI screeners do the reverse — actively evaluating employers and candidates as inputs to a shortlist.

In this world, the question shifts. It's no longer "does AI know about you and get you right?" — that's a necessary but insufficient condition. It's "can an AI agent act on what it knows about you, reliably, without a human?" That's a different question with a different answer. We call the answer your Agent Readiness Score.

What changed

Five market shifts have collectively moved the goalposts between "visibility" and "readiness":

1. Agents, not just answers. Auto-apply tools now route real applications. The AI isn't a librarian any more; it's an actor.
2. Machine-readable beats well-written. Marketing prose doesn't parse reliably; structured data does. Agents prefer what they can cite with confidence.
3. Verifiable beats visible. Visibility without verifiability is a fraud risk. Agents weight verifiable signals higher.
4. Cross-model consistency becomes a signal. Divergence between what ChatGPT, Perplexity, Claude, and Gemini say about you is a trust problem — agents read it as noise.
5. Freshness compounds. Dated, consistently-updated records outweigh a single polished page. Agents reward systems that prove they're still current.

Each shift changes what a scoring system should actually measure.

From Visibility to Readiness: what the metric covers

Agent Readiness Score reuses the underlying data OpenRole already captures — what each major AI model says about an employer, across a battery of candidate-intent prompts — and reframes it against five agent-centric dimensions instead of the old eight visibility categories.

Each dimension is scored 0–20, composed into a 0–100 overall score. A 100 means an AI agent can pick up your employer record, extract every fact it needs, cross-verify, and confidently route a candidate to you without further intervention. A 30 means the agent has to fall back to third-party data, guess, or skip you.

What the early distribution looks like

Applying the Agent Readiness framework to the 500 UK employers in OpenRole's 2026 dataset, the early picture is stark:

The top band is dominated by US tech subsidiaries (Google, Microsoft, Stripe), a handful of UK scale-ups that have explicitly invested in AEO, and two FTSE 100 firms with exemplary schema implementations. The bottom band includes several well-known UK brands whose content is gated, behind login walls, or explicitly bot-blocked — invisible at candidate-agent scale despite strong human brand recognition.

This distribution is the real point: human-facing employer brand strength does not predict agent readiness. A company can have a famous brand, a polished careers site, and score 32 on agent readiness. Conversely, a scrappy scale-up with aggressive schema adoption can score 85.

The three filters for improving your score

The highest-leverage moves across all five dimensions reduce to three filters. When you're deciding what to invest in next, ask:

1. Does this accumulate structured employer data?

Every action — publishing salary ranges, adding JobPosting schema, filling out the FAQ, populating Organization markup — should produce a machine-parseable record, not just a marketing artefact. If the only output is prose, the output won't compound.

2. Does it have an API-shaped version, or only a page?

A careers page is one-shot per visit. An API endpoint (your JSON-LD on every page, your public data feeds, an /api/employer.json if you want to go further) can be queried repeatedly and reliably. Agents prefer API-shaped surfaces over visual pages.

3. Does it create verifiable cross-reference?

Salary bands cross-referenced to Glassdoor. Organisation data cross-referenced to Companies House and LinkedIn. Benefits cross-referenced to your own llms.txt. Every cross-reference is a trust signal agents can grade. Single-source claims are weaker than triangulated claims.

These filters are adapted from OpenRole's broader infrastructure thesis — the same three filters apply to every product feature we build.

Reframing the work

Most employer-brand teams are already doing 70% of the work needed for Agent Readiness. They're writing careers content, publishing role descriptions, responding on Glassdoor, updating benefits pages. What the Agent Readiness frame changes is the structural form those outputs take.

Same writing → wrap in FAQPage schema. Same role description → emit JobPosting JSON-LD with full baseSalary. Same benefits page → publish as llms.txt entry + structured list. Same Glassdoor response → also post the same content on your own site with dated Review schema so AI has a canonical version to cite.

The investment isn't more content; it's the same content shaped for a different primary consumer.

What comes after the score

Agent Readiness Score is a 2026 reframe of what the audit product has always been doing under the hood — and a step on the path to employer agent-interoperability. The pattern is familiar: first you measure, then you standardise, then you interoperate. The Glassdoor era taught employers to measure reputation; the Agent Readiness era will teach them to measure agent-usability of their public information.

The next step — which we expect to be the 2027 conversation — is direct agent-to-employer API interfaces. "Can my candidate-agent talk to your employer-agent?" At that point, the Agent Readiness Score becomes an input to agent routing decisions, and the employers in the top band get disproportionate application flow. That's the infrastructure layer our longer-term roadmap describes.

For now, the tactical step is to reframe the metric, measure your score honestly, and prioritise the work in that order.

Frequently Asked Questions

Q: How does Agent Readiness Score differ from AI Visibility Score?

A: AI Visibility Score (0–100) measured whether AI models had reasonable answers about you when asked candidate-intent questions. Agent Readiness Score (0–100) measures whether an AI agent can act on those answers reliably — so adds dimensions around structural parseability, cross-model consistency, verifiability, and freshness. Same underlying prompts; richer scoring surface.

Q: Is Agent Readiness Score a replacement for the AI Visibility Score?

A: The reframe is primarily terminology. OpenRole's audit product already measures the underlying components; the Agent Readiness framing is a clearer articulation of what those components are for in an agent-mediated world. Expect to see both used interchangeably during the transition, with Agent Readiness becoming the default.

Q: Can I improve Agent Readiness without spending on content?

A: Yes — most immediate gains come from structural moves (schema markup, llms.txt, bot-access configuration, publishing existing salary data in JSON-LD). Content investment matters but is often secondary to structural fixes at the point most employers are at.

Q: How often should we re-score?

A: Monthly is a useful cadence for most UK mid-market employers. AI models ingest new content on the order of 1–4 week cycles, and monthly scoring catches drift from model updates, competitor moves, and Glassdoor/Comparably changes. Enterprise teams tracking multiple brands or geographies may want weekly.

Q: What single change moves the score fastest?

A: Publishing machine-readable salary ranges (JSON-LD baseSalary) for every live role. Salary is the single category where most UK employers lose agent-level confidence today. Schema visible on one crawler cycle typically moves the dimension score 3–6 points.

Q: Does Agent Readiness care about candidate agents or employer-side agents?

A: Both. The score is designed so that a candidate-side agent evaluating whether to route to you and an employer-side agent enriching or verifying the record both benefit from the same structural investments. Two-sided usage is explicit in the framework.

---

Run your Agent Readiness audit — the same free tool, now with the reframed scoring surface.

Related reading:

• Glossary: Agent Readiness Score
• The Six Unsolved Layers of AI-Mediated Hiring
• From SEO to GEO: 2026 Employer Playbook
• Schema.org for Careers Pages