Short answer: Proof-Grade signals are directly verifiable facts, such as published pages, live schema, crawlable URLs, approved content, or a scan finding that reads the site directly. Directional signals are real but not fully causal, such as early prompt visibility movement, preliminary SOMV shifts, or changes in branded search. Not Yet Measured means a claim requires more data before it should be treated as evidence. The Public Proof Package uses these three labels so reporting stays honest.
AI search reporting needs more honesty, not more hype.
Brands want to show progress. Vendors want to show impact. Buyers want to know what is actually real. The danger is turning every signal into a guarantee.
This is what the Public Proof Package exists to prevent.
AI Search Visibility needs confidence labels
No credible vendor can guarantee AI citations, rankings, traffic, revenue, customer acquisition, or retention.
Search engines and AI systems are external platforms, and their retrieval, ranking, citation, and recommendation behavior changes over time. That does not mean you should avoid measurement. It means every claim should carry a confidence label, so a verifiable fact is never confused with a hopeful trend.
Proof-Grade means directly verifiable
A Proof-Grade signal is directly verifiable, non-inferred evidence.
You can check it without guessing at causation. Proof-Grade signals include:
- Live public pages
- Published schema
- Crawlable URLs
- Approved content
- Sitemap entries
- Support answers created
- An EchoScan finding that directly reports a prompt output on a specific date
- A MeshScore diagnostic finding that directly reads the site
Proof-Grade does not mean guaranteed business results. It means the evidence exists and can be inspected.
Run the free EntityMesh scan to see which parts of your AI search readiness are proof-grade today.
Directional means useful but not causal proof
A Directional signal is real, but it should not be presented as causation.
Directional signals show movement, and movement matters. But a change that correlates with new infrastructure does not automatically prove the infrastructure caused it. Directional signals include:
- Preliminary SOMV movement
- Prompt visibility changes
- Early branded search movement
- Increased mention frequency
- Page engagement movement
- Reduced repeated questions reported by the team
- Early support ticket pattern changes
Report these carefully. They are worth tracking. They are not proof of outcome.
Not Yet Measured means the data is not ready
A Not Yet Measured claim requires more data before it should be treated as evidence.
This label protects you from claiming something you cannot yet support. Not Yet Measured claims include:
- Revenue attribution
- Retention improvement
- Customer acquisition lift
- Lifetime value impact
- Long-term AI visibility trends
- Causal improvement in citation frequency
Not Yet Measured does not mean false. It means the claim has not been measured enough to present as proof.
Why overclaiming destroys trust
Buyers have seen search vendors overpromise before.
Now they are watching the same pattern move into AI citations, ChatGPT visibility, and AI Overview claims. When a vendor presents a directional signal as a guaranteed outcome, the first missed result destroys the credibility of every other claim. Confidence labels do the opposite: they show the buyer exactly what is known, what is directional, and what still needs measurement. That is more persuasive than inflated certainty.
Run the free EntityMesh scan to separate proof-grade infrastructure gaps from directional visibility signals.
How confidence labels improve EntityMesh reports
Reports should separate build evidence from outcome evidence.
| Signal type | Meaning | Example | How to use it |
|---|---|---|---|
| Proof-Grade | Directly verifiable | A published Support Hub page exists | Report as fact |
| Directional | Suggestive but not causal proof | SOMV improves across a small prompt set | Report cautiously |
| Not Yet Measured | Data not available yet | Revenue impact after a build | Do not claim yet |
This keeps reporting useful without turning every signal into a sales claim.
How confidence labels apply to MeshScore, SOMV, and EchoScan
Each tool produces a different grade of signal.
- MeshScore findings that directly read the site are Proof-Grade for what exists, and diagnostic (not a guarantee) for what will happen.
- EchoScan findings are Proof-Grade for a specific prompt output on a specific date, and Directional when read as a trend.
- SOMV movement across a small prompt set is Directional, because prompt outputs vary and a short window is not a stable trend.
The labels do not weaken the tools. They describe the signal honestly.
What should never be presented as proof?
Never present the following as proof unless there is direct evidence for it:
- Expected future citations
- Predicted ranking gains
- Unverified revenue lift
- Assumed retention improvement
- Inferred customer acquisition impact
- A one-off prompt result treated as a stable trend
- A diagnostic score treated as an outcome guarantee
Good proof makes a claim more believable. Overstated proof makes the whole report less trustworthy. This is the same honesty policy behind Can EntityMesh promise AI citations?.
How to use the Public Proof Package in your own content
You can apply these labels anywhere a claim might be misread: case studies, diagnostic reports, blog posts, Support Hub pages, sales decks, and client updates.
Keep the labels simple:
- Proof-Grade: directly verifiable
- Directional: useful signal, not causal proof
- Not Yet Measured: more data needed
This lets a brand build authority without claiming more than it can defend. For a deeper example of applying labels during an audit, see How to measure Support Hub performance.
Frequently asked questions
What is a Proof-Grade AI Search signal?
A Proof-Grade signal is directly verifiable evidence, such as a published page, live schema, a crawlable URL, approved content, or a scan finding that reads the site directly. It can be checked without inferring causation.
What is a Directional AI Search signal?
A Directional signal is a real but non-causal signal, such as early SOMV movement, prompt visibility changes, or increased mention frequency. It suggests progress but should not be reported as proof of outcome.
What does Not Yet Measured mean?
Not Yet Measured means a claim, such as revenue attribution or retention lift, needs more data before it can be treated as evidence. It does not mean the claim is false.
Why do confidence labels matter?
They separate verifiable facts from directional trends and unmeasured outcomes, which keeps reporting honest and protects credibility when external AI systems behave unpredictably.
How does the Public Proof Package prevent overclaiming?
By forcing every claim into one of three labels, it stops teams from presenting a directional signal or an unmeasured outcome as a guaranteed result.
How does EntityMesh use confidence labels?
EntityMesh can label diagnostic and report findings as Proof-Grade, Directional, or Not Yet Measured, distinguishing what was built, what changed directionally, and what still needs measurement.