Source identity
| Field | Recorded value |
|---|---|
| Source type | project equation-poster image |
| Original record | AI visibility equation infographic.png |
| Date | Jul 11, 2026 |
| Path or archive member | Equation posters/AI visibility equation infographic.png |
| Original SHA-256 | 9358771f4a95fc2089ee11767dca77e165ffff25f3deb871a9485d31bbdb2adc |
| Reading-copy SHA-256 | 314410710ca83c6d50a86036f0514ac710e824f3894be2f710658cabf650be38 |
The original equation poster
Recovered source transcript
Download complete source text ↓809 lines · 14.7 KB
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
Mathematical records from this source
0102030405060708091011121314
Gated AI visibility score
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…
Softmax source-selection probability
The candidate score is exponentiated after division by a temperature, and divided by the sum of exponentiated candidate scores.
Choose a source maximizing the proposed visibility score
The argmax operator returns the candidate source p whose AI_VIS score is largest for the specified query, claim, user context, time, and location. The starred p* denotes the…
Eligibility gate
Combines crawl, index and retrieval eligibility as a product.
Gate suppression condition
A zero gate suppresses the visibility score regardless of the weighted factors.
Findability factor
Combines lexical retrieval, vector similarity, query intent, entity overlap, and topical relevance.
Claim-support factor
Adds positive evidence and answerability signals and subtracts contradiction.
Trust factor
Combines authority and reliability labels. The poster does not define common units or transformations for adding these signals.
Entity factor
Combines entity prominence, consistency, knowledge-graph score, sameAs support, and schema clarity.
Usefulness factor
Adds information gain, completeness, recency and experience signals for the candidate query.
Context factor
Combines the query and user-context fit dimensions listed in the image.
Feedback factor
Combines click-through and engagement-style signals with short-click and pogo-sticking penalties.
Cross-source corroboration factor
The panel links source support and source trust in a summed corroboration calculation.
Risk factor
Adds negative-content and access-risk labels into the risk quantity that is penalized in the master score.