The Universal AI Visibility Equation
====================================

SOURCE: AI visibility equation infographic.png
SOURCE ID: libsrc-6e10f2781ce0
ORIGINAL SHA256: 9358771f4a95fc2089ee11767dca77e165ffff25f3deb871a9485d31bbdb2adc
REVIEW: 2026-10-17. The original PNG is authoritative. This reading copy separates manually inspected mathematical transcriptions from raw OCR. It is not a verbatim embedded-text extraction.

VISUAL REVIEW NOTES
- The repeated shorthand AI_VIS(...) = Final Visibility Score and p* = argmax_p AI_VIS(...) restate the master score and argmax; they do not introduce distinct formulas.
- COMPETITION and REDUNDANCY appear in the master expression and quick-view prose, but no independent computational formula for either is shown.
- The image is a project artifact proposing an AI-visibility framework. Named platforms in its output list are claims of intended application, not evidence of adoption or empirical performance.

MANUALLY INSPECTED FORMULAS

Gated AI visibility score
Image region: Upper pictogram equation and lower-right COMPLETE FORMULA (TEXT VERSION) block
AI_Vis(p,q,c,u,t,l) = GATE(p,q) · [ α·FIND + β·SUPPORT + γ·TRUST + δ·ENTITY + ε·USEFULNESS + ζ·CONTEXT + η·FEEDBACK + θ·CORROBORATION − λ·RISK − μ·COMPETITION − ν·REDUNDANCY ]
Legibility: Factor names and coefficient signs are legible; several tiny argument-list glyphs are ambiguous.
Editorial mathematical explanation: A gate multiplies a weighted sum of eight positive scoring factors and three suppression factors. The compact transcription names the visible factors without inventing uncertain tiny argument punctuation.
Editorial note: The upper panel uses pictograms in the bracket. Its separator before the competition pictogram appears as a plus sign while the coefficient label below reads −μ; the complete text block subtracts competition. Preserve this visual inconsistency rather than silently choosing one upper-panel sign.
Editorial note: The complete-text block clearly supplies FIND(p,q), SUPPORT(p,c), TRUST(p), ENTITY(p,q), CONTEXT(q,u,t,l), FEEDBACK(p,q), and REDUNDANCY(p;S). Punctuation in the small USEFULNESS, CORROBORATION, RISK, and COMPETITION argument lists is not reliably legible at the supplied resolution.
Editorial note: No calibrated weights, numeric scale, training data, or measured search-engine/AI ranking function is supplied in this image.

Softmax source-selection probability
Image region: Upper selection/citation strip; repeated in lower-right complete-text block
P(select p | q,c) = exp(AI_VIS(p,q,c,u,t,l)/τ) / Σ_j exp(AI_VIS(p_j,q,c,u,t,l)/τ)
Legibility: Main operation legible; denominator candidate subscript differs or is degraded between repeated renderings.
Editorial mathematical explanation: The candidate score is exponentiated after division by a temperature, and divided by the sum of exponentiated candidate scores.
Editorial note: This is a normalized selection model proposed by the poster. It does not establish that any named external platform uses this rule.
Editorial note: The upper typeset denominator appears to print p rather than p_j although its sum is over j. The lower monospaced version appears to include p_j, with some small adjacent punctuation degraded. This transcription follows the lower candidate-index version and records the upper discrepancy.
Editorial note: A defined candidate set, nonzero temperature, and finite scores are needed to evaluate the expression.

Highest-score source
Image region: Upper Selected Source box; repeated in Output and Complete Formula blocks
p* = argmax_p AI_VIS(p,q,c,u,t,l)
Legibility: Legible
Editorial mathematical explanation: Selects a candidate with the largest visibility score.
Editorial note: Tie handling and candidate eligibility details are not specified in this image.

Eligibility gate
Image region: Column 1: GATE
GATE(p,q) = Crawl · Index · Retrieve
Legibility: Legible
Editorial mathematical explanation: Combines crawl, index and retrieval eligibility as a product.
Editorial note: The column lists robots allowed, server OK, HTML accessible, index allowed, canonical valid, and retrieval threshold passed. No numerical definition of the three gate operands is supplied.

Gate suppression condition
Image region: Bottom of column 1: GATE
If GATE = 0, AI_VIS = 0
Legibility: Legible
Editorial mathematical explanation: A zero gate suppresses the visibility score regardless of the weighted factors.

Findability factor
Image region: Column 2: FIND
FIND(p,q) = BM25 + VectorSim + IntentMatch + EntityOverlap + TopicalRelevance
Legibility: Legible
Editorial mathematical explanation: Combines lexical retrieval, vector similarity, query intent, entity overlap, and topical relevance.
Editorial note: The bullets also name RRF retrieval, but no RRF formula or additional RRF term is displayed.

Claim-support factor
Image region: Column 3: SUPPORT
SUPPORT(p,c) = Entailment + Answerability + FactualDensity + EvidenceSpecificity + ExtractionEase − Contradiction
Legibility: Legible
Editorial mathematical explanation: Adds positive evidence and answerability signals and subtracts contradiction.

Trust factor
Image region: Column 4: TRUST
TRUST(p) = PageRank + HITS + E-E-A-T + Reliability + Reviews + Mentions + AuthorCredibility
Legibility: Legible
Editorial mathematical explanation: Combines authority and reliability labels. The poster does not define common units or transformations for adding these signals.

Entity factor
Image region: Column 5: ENTITY
ENTITY(p,q) = EntitySalience + EntityConsistency + KGScore + SameAsSupport + SchemaClarity
Legibility: Legible
Editorial mathematical explanation: Combines entity prominence, consistency, knowledge-graph score, sameAs support, and schema clarity.

Usefulness factor
Image region: Column 6: USEFULNESS
USEFULNESS(p,q,t) = InfoGain + Completeness + Freshness + OriginalExperience + SpecificExamples + QueryCoverage
Legibility: Factor terms legible; tiny argument punctuation has reduced confidence.
Editorial mathematical explanation: Adds information gain, completeness, recency and experience signals for the candidate query.
Editorial note: The small argument list is visually cramped; p,q,t is the legible reading in the factor column. The complete-text repetition contains degraded punctuation.

Context factor
Image region: Column 7: CONTEXT
CONTEXT(q,u,t,l) = Intent + LocalFit + DeviceFit + LanguageFit + TimeSensitivity + Personalization
Legibility: Legible
Editorial mathematical explanation: Combines the query and user-context fit dimensions listed in the image.

Feedback factor
Image region: Column 8: FEEDBACK
FEEDBACK(p,q) = CTR_adj + DwellTime + LongClick + CompletionRate − ShortClick − PogoSticking
Legibility: Main terms legible; CTR subscript is small and moderately uncertain.
Editorial mathematical explanation: Combines click-through and engagement-style signals with short-click and pogo-sticking penalties.
Editorial note: The CTR subscript appears to be adj; it is small and should remain a qualified transcription. The displayed PogoSticking label spans the lower formula lines.

Cross-source corroboration factor
Image region: Column 9: CORROBORATION
CORROBORATION(p,c) = Σ support(source,c) [operator unclear] Trust(source,…)
Legibility: Partial transcription: operator and trailing Trust arguments unresolved.
Editorial mathematical explanation: The panel links source support and source trust in a summed corroboration calculation.
Editorial note: The operator before Trust and the last argument or closing punctuation of Trust are not reliably legible. They have not been reconstructed from expectation.
Editorial note: Bullets name source agreement, claim confirmation, third-party validation, citation-graph support, and consensus strength.

Risk factor
Image region: Column 10: RISK
RISK(p,c) = Spam + ThinContent + Contradiction + UnsupportedClaims + AccessFriction + Duplicate + Manipulation + LowTrust
Legibility: Legible
Editorial mathematical explanation: Adds negative-content and access-risk labels into the risk quantity that is penalized in the master score.

SOURCE-LOCAL SYMBOL DEFINITIONS
p = Page / website / source (the candidate).
q = User query.
c = Claim the statement needs to support.
u = User context, including location, device, language, personalization and history when used.
t = Time: current time, freshness needs and topic volatility.
l = Location: geographic location, service area and proximity.
S = Selected sources already chosen, used for diversity control.
τ = Temperature controlling selection sharpness versus diversity.
α = Find coefficient.
β = Support coefficient.
γ = Trust coefficient.
δ = Entity coefficient.
ε = Usefulness coefficient.
ζ = Context coefficient.
η = Feedback coefficient.
θ = Corroboration coefficient.
λ = Risk penalty coefficient.
μ = Competition penalty coefficient.
ν = Redundancy penalty coefficient.

RAW OCR READING AID — NOT A VERIFIED EQUATION TRANSCRIPT
The following preserves the complete Tesseract output, including recognition errors, decorative glyphs and layout artifacts. Use the manually inspected records above and the original PNG for mathematical notation.

THE UNIVERSAL Al VISIBILITY EQUATION

The Complete Math Behind Rankings, Citations & Al Answers

AI_ViS @.4.c.u,) = GATE(p,a) *

+(M

“

G+ OrP+-4+ Grae

\+@) @

FIND

SUPPORT

TRUST

ENTITY

USEFULNESS CONTEXT FEEDBACK bo a

RISK

COMPETITION REDUNDANCY

a

B

Y

6

5

G

n

Cc)

-A

—p

-v

ALVIS(0,4,6,U,t))/t

Probability of

e

==

Selected

* = argmax,

=

P(select p | q,c)

Source

Selection / Citation

E, eALMS(e,a.cust.0/e

y

ALvis(p.q,c.u,tl)

O)

@

© VY

GO F\4| 9g

a

ngs | (A

GATE

FIND

SUPPORT

TRUST

ENTITY

USEFULNESS

CONTEXT

FEEDBACK

CORROBORATION

RISK

Eligibility

Findability /

Claim Support /

‘Authority /

Entity / KG

Information Gain /

User / Intent /

Engagement /

Cross-Source

Negative / Spam /

(Required)

Relevance

Answerability,

Quality

Strength

Freshness

Location / Time

Satisfaction

Support

contradiction

a

2 =

GGATE(p,q) =

FIND(p.@) =

SUPPORT(p,<) =

TRUST(p) =

|

ENTITY(p,q) =

USEFULNESS(p,q1)=

CONTEXT(quut) =

FEEDBACK(p.q) =

CCoRROBORATION(p,c)=|

ISK(p,c) =

Crawl Index Retrieve

‘BM25 + VectorSim+

Entallment +

PageRank + HITS +

EntitySalience +

Infoain +

Intent + LocalFit +

CTR DwelTine

‘Spam + ThinContent

IntentMatch +

Answerability +

EEAT+

Completeness +

DeviceFit +

E support(source,<)

Contradiction +

EntityConsistency +

LlongClick +

+ Trust(source,,)

EnttyOverlap +

FactualDensity +

Reliability +

KGScore +

Freshness +

Languagefit +

CompletionRate~

UnsupportedClaims+

TopicalRelevance

EvidenceSpecticty+ | Reviews + Mentions

SameAsSupport +

OrignalExperience +

TimeSensitvity +

Shortlick =

‘AccessFrction +

+ robots alowed

ExtractionEase -

AuthorCredibility

SchemaClarity

SpecificBxamples +

Personalization

Pogosticking

Duplicate +

+ server OK

Contradiction

Manipulation +

+ HIML accessible

QueryCoverage

Lowtrust

+ BM25 /TEDF

+ etallment

+ intent class

+ CTR

+ index allowed

+ PageRank

+ entity salience

+ information gain

+ source agreement

+ spam signals

+ canonical valid

+ vector similarity

+ passage arsverbity

+ HITS authority

+ KG connections

+ completeness

local relevance

+ dwell time

+ claim confirmation

+ thin content

+ retrieval threshold

«= bybridcetrival

+ factual density

BEAT

+ same links

+ freshness

+ device / language

long click

+ 3% party validation

+ contradictions

passed

+ intent match

+ evidence blocks

+ reviews

“NAP consistency

+ original experience

+ time sensitivity

+ task completion

+ citation graph

+ unsupported claims

+ entity overlap

+ direct answers

+ third-party mentions |» organization schema

+ examples / data

+ user location

+ return rate

support

+ acess fictions

+ contraction

+ brand & service

+ duplicates

+ topical relevance

+ author credibility

+ query coverage

+ personalization

+ pogo sticking

+ consensus strength

“+ RRF retrieval

detection

+ source reliability

identity

+ manipulation

If GATE

+ Tow trust

ALVIS

=

HOW IT WORKS

OUTPUTS (WHAT YOU GET)

QUICK VIEW: VALUE vs SUPPRESSION

Gatekeoy

First

Ita page cannot be crawled, indexed

(or Fetrieved, it cannot rank or be cited,

VALUE (Add)

‘SUPPRESSION (Subtract)

QHD auis(p,a.ciu,t)

®

‘Score the Value: The system scores the page across

a

= Final Visibility Score

© fino - cant be foun?

RISK ~ Any spam, thin,

8 positive factors (value).

Contradiction, or access issues?

@ Subtract the Suppression: Risk, competition, and

‘SUPPORT - Does it support

redundancy reduce the score.

the claim?

Higher Score = Higher chance of

COMPETITION

Isit better

@ Compare to Others: Your page is compared against all

‘other candidates,

+ Page-One Rankings (Google / Bing)

© TRUST -Isitcreive?

than the other options?

© ENTITY siceay ndesood?

@© Probability Selection: A softmax function converts

+ Al Overview Citations (Google)

REDUNDANCY ~ Does it add

‘scores into selection probabil.

© USEFULNESS - Is it helpful,

something new?

+ Copilot Citations (Bing)

‘unique & current?

The Winner: The highest scoring (most useful, trusted,

+ ChatGPT Citations

and relevant) page wins

© CONTEXT - Does it ft the

+ Gemini Citations

se, intent & location?

FEEDBACK ~ Do users find it

‘+ Claude Citations

INPUTS (WHAT THE SYSTEM USES)

satistying?

Page / Website / Source (the candidate)

+ Perplexity Citations

CORROBORATION - Do other

COMPLETE FORMULA

+ Any Al Grout

1 / Answer Use

trusted sources agree?

(TEXT VERSION)

User Query (what is being asked)

Claim (the statement the Al needs to support)

AL_VIS(p,q,¢,u,t,1) =

GATE(p.q) *

User Context (location, device, language,

Pp’

= argmax, AI_VIS(.

‘a*FIND(p,q) + B*SUPPORT(p,c)+ y*TRUST(p) +

personalization, history if used)

B*ENTITY(psq) + E*USEFULNESS(pyqy t) +

Time (current time, freshness needs, topic volatility)

= The most likely page to be

ACONTEXT(q,ust,1) + n*FEEDBACK(p,q) +

ranked, cited, or used.

6*CORROBORATION (psc)

~ APRISK(p ec) -

Location (geographic location, service area, proximity)

*COMPETITION(p,q,"c) - v*REDUNDANCY (pS)

Selected Sources (already chosen sources

o!

for diversity contol)

P(select pla.c) =

exp(AL_VIS(p,a.c,u,t,1)/t)

Temperature (controls selection sharpness vs diversity)

Ej exp(ALVIS(p 3... 3 €.u, t,1)/t)

pt

argmax

AL_VIS(P.9,¢,U, ty)

=

='

_

—

SYMBOL LEGEND

)

AT A GLANCE FLOW

DOWNLOAD .TXT

SN

of

AI_VIS_EQUATION_COMPLETE.tet

es

OR 4

eer

Foday

Tonge

‘ophy

Tat

ny

eens

ghd

Your complete equation, definitions,

B27 Qa

oF

inputs, outputs, and explanations

Crawl

Retrieve

Score with

Compare

Select /Cite

ina reusable text file

=

aN

&®

& Index

Candidates

ALVIS

& Rank

Peron

Cae

Fert

hart up

Hanae

Waring

moe

vera

Top Sources

contention

competion Reddy

i