Rungs above a break are locked
Stop your team optimising content nobody can fetch. If a crawler cannot reach the page its citation rate is noise, so everything above the break is locked until you clear it.
AI search visibility
Find what to fix first in AI answers
Free scans hand you a number out of 100 and no next step. This names the one rung that is broken and sends you to the page that fixes it.
Key takeaway
A blended visibility score is unactionable by construction: it averages a crawler that cannot reach you with a page an engine cannot extract, and hands you one number for two unrelated problems. The ladder refuses to average. It names the one rung that is broken and locks the ones above it, because fixing rung four while rung one blocks the crawler is wasted work.
The ladder
One broken rung, named, instead of a score
Every competitor sells a number. A number tells you where you stand and nothing about what to do on Monday. Six rungs run from whether an engine can fetch your page at all through to whether it will recommend you, and the ladder stops at the first one that fails.
Each rung has somewhere to go
Get an answer to "what now" in one click. Crawl access, extractability, citations, sentiment and prompt coverage each link to the page that explains the fix.
The thresholds are printed
Argue with the numbers instead of accepting a grade. Paragraph length, list ratio and heading density are published with a floor and a ceiling, labelled as a documented first pass rather than settled science.
The engines
What is actually measured, and what costs extra
Most vendors advertise an engine count and let you assume it is all included. Here is the split, stated up front, because you will find it out at the invoice otherwise.
Included on every plan
Get the three surfaces every brand asks about without an upgrade. ChatGPT and Perplexity are probed daily, and Google AI Overviews is read per tracked keyword because it is a result page rather than a chat engine.
Priced as add-ons
Add only the engines your buyers actually use. Gemini, Claude, Mistral, DeepSeek and Google AI Mode are priced per tier, and AI Mode is separate from AI Overviews because the two behave and bill differently.
Four prompt types per engine
Win the buying stage you are losing rather than all of them at once. Brand direct, category, comparison and pain point are scored separately, because winning one does not win the others.
The join
Joined to traffic you can bank, and labelled where it cannot be
AI visibility is a leading indicator. We report it as one and refuse to close the loop we cannot close, which is the part that makes the rest of the numbers worth reading.
AI referred sessions, measured server-side
Count the traffic AI actually sends you. Sessions from AI referrers are identified in your own first-party tracking and measured against the same deduplicated conversions as every other channel.
What we will not claim
Take nothing to your board that we cannot support. Nothing here bridges a citation to an order, because the data cannot carry it, and a vendor who bridges it is selling you a correlation with a confident face.
On the same keyword row as SEO
Read the trade between a blue link lost and a citation won on one row. Rankings, AI Overview presence and citation rate sit together instead of being inferred across two tools.
FAQ
Direct answers.
Because a single score averages problems that have nothing to do with each other. A blocked crawler and a badly structured page both drag a blended number down, and the number cannot tell you which one you have. The ladder refuses to average: it names the first rung that fails, locks everything above it, and links to the fix. That is less flattering than a score out of 100 and considerably more useful on a Monday morning.
ChatGPT and Perplexity are probed daily on every plan, and Google AI Overviews is read per tracked keyword rather than probed, because it is a result-page surface rather than a chat engine. Gemini, Claude, Mistral, DeepSeek and Google AI Mode are per-tier add-ons. We state the split rather than advertising a headline engine count, because the count is what everyone advertises and the split is what you actually buy.
No, and we will not present it as though we can. AI visibility sits in the leading-indicator regime: mention rate, share of voice, citation rate, each reported with its noise level. Sessions from AI referrers are real first-party traffic, measured against your deduplicated conversions like any other channel, so that part is countable. What we refuse to do is attribute an order to a citation. Causal claims are reserved for paid media, where a holdout test can carry them.
A free scan is a snapshot with no history, no competitors and no next step. Running one is a reasonable way to find out whether you have a problem at all. What it cannot tell you is whether last month's fix worked, whether a rival took the ground you lost, or which of six possible causes is yours. If a scan is all you need, take the free one.
Yes. The ladder is the same for any business whose buyers ask questions before they buy. Page audits branch on whether the property is ecommerce or lead generation, so purchase-funnel checks are skipped rather than counted as failures. The revenue join is thinner without orders; the crawl access, extractability, citation and sentiment rungs are unchanged.
14-day free trial
Find out which rung is costing you
14-day free trial. AI visibility is on every plan, with ChatGPT and Perplexity probed and Google AI Overviews read per keyword.