Buyers used to open ten tabs. Now they ask a model and read one answer. That answer names a few vendors and moves on.
Which names it picks depends on what the model can find, read and trust on your site — decided before anyone types your URL into anything. Nothing in your analytics shows the shortlist you weren't on.
Each check is something an agent does before it decides you're worth recommending. The score is the same one every PlayHero scan carries as its agent half — this page just shows the checks behind it.
Read by some crawlers, ignored by most. We check it — we don't lean on it.
Who you are and what you sell, in a form a model parses without guessing.
SoftwareApplication schema so an agent can classify the product, not just the company.
Self-contained blocks a model can lift into an answer with the claim intact.
How you show up when a model cites you.
Whether GPTBot, ClaudeBot, PerplexityBot and thirteen others are allowed in — or blocked by an old rule nobody remembers.
Alt text and imagery a model can read, not just see.
Coming · Rendered content — is your copy in the HTML an agent fetches, or only after your scripts run? Measured now, scored when we've seen enough pages.
When a model doesn't name you, it names someone. On Monitoring, Lumen asks the questions your buyers ask across 5 platforms every week, records who comes back, and tells you why it wasn't you.
Nothing on your dashboard tells you the models stopped naming you. Lumen does: flagged, prioritized, fixed by a person, verified by the next run.
You weren't named at all on a query your buyers ask. Flagged critical — to an AI-referred buyer, you're down.
Schema, framing, access or content — the specific change with its evidence. Keep it, fix it, or hand it to engineering. Nothing executes without a person.
When your team ships, the next weekly sweep confirms you're back on the list. The card closes itself — never before the proof.
score = human UX × 0.88 + agent readiness × 0.12Agent readiness is the agent half of every PlayHero scan — the same number your team already tracks, not a separate report.The card at the top is this page, scanned by PlayHero on 2026-10-30 under Monitoring — not a demo account and not an illustration. Agent readiness came back 92 against a benchmark of 73. Lumen findings stay on this receipt when they exist.
Follow its evolution on /proof/lumen →Pay per site, not per seat. Unlimited teammates on every plan.
Enough to prove the instrument is real.
No credit card. No trial clock.
Card not charged until day 14 · cancel any time
Your clients' sites, your name on the report.
Card not charged until day 14 · cancel any time
Above 250 sites, talk to us →
Straight answers, the same ones an AI agent gets when it reads this page.
ChatGPT, Perplexity, Gemini, Grok and Claude. Agent readiness is scored from what they can read; on Brand and Agency, Lumen also asks them your buyers' questions every week and records whether they name you.
No. Readiness is whether a model can use your page. Being named is whether it does — and the second is only measured on Monitoring, by asking. This page's own score shows the gap: readable, not yet cited.
A query is what a buyer types. A run is one query on one platform. A sweep is the weekly pass across all 10 queries and 5 platforms — 50 runs per site.
It's one check of seven, graded on quality when present. Some crawlers read it, most ignore it. We check it and never make it the reason for a score.
No. Lumen reads the published URL and asks the platforms directly. Connect nothing; paste a URL.
Never on its own. A finding becomes a Rallycard that waits for a person to approve, assign or dismiss it. The next run records whether the fix held.