We start from what you can verify
URLs, public company records, product pages, game listings, and the text you paste into Decode. We ground checks in inputs you can see — not opaque private complaint pipelines.
Data sources & methodology
How ShouldEye
actually works
ShouldEye is an AI search engine and scam checker. EyeQ asks 70+ AI models across 13 frontier families, cross-checks their answers, and ships one clear verdict — for websites, companies, people, games, products, messages, and scams.
70+AI models on callacross every EyeQ turn
13Frontier familiesnamed on this page
8Check surfaceswebsite to decode
1Synthesized answernever one model alone
OnPrivacy-firstno sale of personal data
EyeQEyeQ orchestrationroutes, fans out, reconciles
Trust commitments
Four commitments, every check
These apply whether you ask EyeQ a freeform question or run a dedicated check. The sections below show which models and surfaces sit behind each commitment.
We never sit between you and the operator
ShouldEye does not proxy payments, deposits, checkouts, or account actions. We analyze and advise. You stay in control of where you click next.
We do not sell your activity
Site telemetry is scoped to your ShouldEye account and product experience. We do not sell personal data to operators, advertisers, or data brokers.
No single model gets the final word
EyeQ fans questions across multiple frontier families and synthesizes the result. Disagreement is useful signal — the product is built to cross-check, not to rubber-stamp one engine.
The AI Advisory Board
70+ models. One verdict.
Every EyeQ answer is orchestrated across frontier AI families — not a single chatbot. The registry below names all 13 families in production. Variant counts come from the live model catalogue; roles describe how each family typically contributes.
GPTOpenAI
Broad reasoning, coding, and professional writing. Strong default for complex synthesis.
13variants in catalogue
ClaudeAnthropic
Careful analysis, long-form nuance, and instruction-following under ambiguity.
8variants in catalogue
GeminiMultimodal research — documents, screenshots, and tool-heavy workflows.
6variants in catalogue
GrokxAI
Direct tone and fast agentic sweeps when speed and breadth both matter.
5variants in catalogue
DeepSeekDeepSeek
Cost-efficient depth for high-volume text analysis and structured reasoning.
4variants in catalogue
KimiMoonshot AI
Long-horizon multimodal work and agentic coding across large contexts.
5variants in catalogue
GLM
FrontierZhipu AI
Strong bilingual and generalist coverage alongside the frontier pack.
6variants in catalogue
MiniMax
FrontierMiniMax
Additional frontier capacity for diversity in council-style runs.
2variants in catalogue
Qwen
FrontierAlibaba
Open-weight strength for multilingual and technical checks.
7variants in catalogue
NVIDIA
FrontierNVIDIA
Infrastructure-grade models for specialized inference paths.
3variants in catalogue
Llama
FrontierMeta
Open-weight baselines that keep the board from over-fitting to one vendor.
5variants in catalogue
Muse
FrontierdataSourcesPage.board.engines.muse.vendor
dataSourcesPage.board.engines.muse.role
2variants in catalogue
Mistral
FrontierMistral AI
Efficient European frontier models for fast, capable general turns.
5variants in catalogue
EyeQ orchestration
Your prompt (or check) is classified by intent — website, company, person, game, product, decode, scam, or freeform Ask EyeQ. EyeQ selects the right mode and fans the question to the active model lineup for that turn.
Cross-model synthesis
Models answer in parallel. EyeQ compares them, weighs agreement and dissent, and produces one ShouldEye verdict. No single family is allowed to be the unchallenged final word.
What you can see
On EyeQ turns you can inspect which models participated, compare their answers, and share or save the result. The goal is a receipt you can act on — not a black-box score.
How a check runs
From question to verdict, in five stages
Whether you paste a URL, pick a person, open Decode, or ask EyeQ anything, the path is the same shape: gather context, fan out to models, reconcile, synthesize, and give you something you can use.
01
Ask / Check
You start with a clear input: a website, company, person, game, product, conversation paste, scam claim, or freeform EyeQ question.
you decide
- Website URL
- Company or brand name
- Person search
- Game title
- Decode paste
02
Context
EyeQ gathers the check-specific context — public pages, product signals, message text, or search cues — so models are answering the same grounded question.
seconds
- Page / listing context
- Public signals
- Message focus
- Lane-specific prompts
- Attachment context
03
Multi-model analysis
70+ AI models across the active families review the same brief. Diversity is the point: different vendors catch different risks.
parallel
- GPT family
- Claude family
- Gemini family
- Grok / DeepSeek / more
04
Synthesized verdict
EyeQ reconciles the answers into one ShouldEye result — clear language, risk callouts, and next steps — not 13 raw chat windows.
one answer
- Agreement & dissent
- Risk framing
- Actionable next steps
- Shareable result
05
Save & share
Keep the thread, export what you need, or continue asking follow-ups. Your account history is for you — not a marketplace of personal data.
optional
- Conversation history
- Share flows
- PDF where available
- Follow-up questions
Check surfaces
Eight ways to verify before you trust
Each surface is a product entry point into the same EyeQ methodology. Sources and model stacks below describe what that surface typically uses — not a guarantee that every turn fans out to every family.
Ask EyeQ
Freeform AI search with a multi-model board
Ask EyeQ anything. The home composer is the general advisory surface: route by intent, fan out across frontier families, and get one synthesized answer.
AskPrimary entry
13Families available
70+Models available
Typical families on call
GPT Claude Gemini Grok DeepSeek Kimi + more
What we look at (3)
- Your question and any staged context (text, links, images where supported).
- Public web and knowledge signals EyeQ pulls for that turn.
- The active model lineup you selected (or the default Auto / council path).
How it usually runs
On demand. Each send is a fresh orchestration for that prompt — not a background scrape of your browsing history.
What we protect
Account-scoped history stays in your ShouldEye account. We do not sell personal prompts to third parties.
What we do not need
Operator passwords, payment credentials, or anything that would let us act as you on another site.
Methodology aligned to the live model registry
Open EyeQWebsite checks
Verify a site before you click, buy, or login
Website and claim checks look at the URL, public reputation signals, and model judgment before you treat a page as safe.
URLInput
RiskRisk focus
VerdictOutput
Typical families on call
GPT Claude Gemini DeepSeek
Where to run this check
What we look at (3)
- The URL or claim you submit.
- Public page context and reputation cues available for that domain.
- Cross-model risk framing (phishing, impersonation, pressure tactics).
How it usually runs
On demand when you paste a URL or ask EyeQ about a site.
What we protect
We do not need your login cookies or passwords to run a website check.
What we do not need
Your bank credentials, 2FA codes, or live session tokens from the site under review.
Methodology aligned to current check lanes
Check a websiteCompany checks
See what AI and public signals say about a brand
Company checks combine directory context with multi-model analysis so you can compare transparency, trust signals, and risk before you commit.
DirDirectory
AIAI analysis
CompareCompare
Typical families on call
GPT Claude Gemini Grok
Where to run this check
What we look at (3)
- Company directory and profile context on ShouldEye.
- Public brand and corporate signals models can cite.
- Side-by-side AI answers when you run a compare-style turn.
How it usually runs
On demand from directory profiles or the home composer.
What we protect
Public company facts stay public. Personal employee PII is not a target dataset.
What we do not need
Private CRM exports, internal email directories, or paid-only corporate dumps.
Methodology aligned to company directory surfaces
Browse companiesPeople checks
AI people search with careful framing
People search helps you verify public identity signals before you trust a profile, outreach, or claim about a person.
NameQuery
PublicSignals
CarefulGuardrail
Typical families on call
GPT Claude Gemini
Where to run this check
What we look at (3)
- The name or query you provide.
- Public web identity signals models can retrieve for that query.
- Safety framing so results stay investigative, not doxxing tooling.
How it usually runs
On demand from the people-search surface or EyeQ.
What we protect
We encourage public-source framing. Private contact lists are not ingested as a product feed.
What we do not need
Your address book, private DMs, or scraped private social graphs.
Methodology aligned to AI people search
Open people searchGame checks
Verify games before you spend or download
Game checks review titles, studios, and public player-facing signals so you can spot risk, quality issues, or trust gaps early.
TitleInput
StudioStudio
VerdictOutput
Typical families on call
GPT Claude Gemini DeepSeek
Where to run this check
What we look at (3)
- Game title or studio you submit.
- Public listings and directory context.
- Multi-model read on reputation, risk, and player-facing claims.
How it usually runs
On demand from gaming surfaces or EyeQ.
What we protect
No need for your platform account passwords to run a game check.
What we do not need
Your wallet balances, spin history, or operator account identity.
Methodology aligned to gaming lanes
Open gaming checksProduct checks
Compare products with multi-model judgment
Product checks help you verify specs, claims, and trade-offs before you buy — grounded in the product page or SKU you care about.
SKUInput
SpecsClaims
CompareCompare
Typical families on call
Claude GPT Gemini
Where to run this check
What we look at (3)
- Product name, SKU, or listing you provide.
- Public product pages and manufacturer claims.
- Cross-model comparison of strengths, risks, and open questions.
How it usually runs
On demand from products or EyeQ.
What we protect
Purchase history outside your ShouldEye account is not required.
What we do not need
Payment card data or retailer checkout sessions.
Methodology aligned to product checks
Open product checksDecode
Read the room before you reply
Decode analyzes conversations and screenshots you paste — dating, work, support, scams, and more — with multi-model judgment on intent and risk.
PasteInput
FocusFocus
ReportOutput
Typical families on call
Claude GPT Gemini Grok
What we look at (3)
- Text or screenshots you choose to paste.
- Category / focus you select in Decode.
- Parallel model reads of tone, pressure, and scam patterns.
How it usually runs
On demand when you complete Decode intake and generate a report.
What we protect
You control what text enters the check. Redact names if you prefer.
What we do not need
Continuous monitoring of your messaging apps without your paste.
Methodology aligned to Decode
Open DecodeScam checks
Detect scams and hidden risk patterns
Scam checks look for phishing, impersonation, pressure tactics, and other fraud patterns across claims, domains, and messages.
ClaimInput
PatternPatterns
VerdictOutput
Typical families on call
GPT Claude Gemini DeepSeek
Where to run this check
What we look at (3)
- The claim, domain, or message you submit.
- Public scam and reputation cues models can cite.
- Cross-model pattern matching for known fraud shapes.
How it usually runs
On demand from scam surfaces or EyeQ.
What we protect
Anonymous pastes can stay session-scoped to the check you run.
What we do not need
Operator account takeovers or credentials that would let us impersonate you.
Methodology aligned to scam / rug check lanes
Run a scam checkPrivacy & product posture
What procurement and privacy teams usually ask
Honest product posture — not invented live-ops telemetry. If you need a formal questionnaire filled, contact us and we will answer from current policy, not marketing fiction.
Product footprint
- Consumer web app at shouldeye.com
- EyeQ chat as the primary verification surface
- Dedicated check lanes for websites, companies, people, games, products, decode, and scams
- Business portal available separately for B2B buyers
Privacy commitments
- Privacy Policy and Terms published on-site
- No sale of personal data as a product practice
- Cookie / consent controls where the custom consent UI applies
- Account-scoped history for signed-in users
- Security and privacy questions: use /contact
How capacity works
- 70+ AI models available through the live catalogue
- 13 frontier families named on this page
- Plan tiers gate model access and council size
- Checks run on demand — not a silent scrape of your life
- Results you can share, save, or continue in EyeQ
70+ models
13 families
Privacy-first
EyeQ orchestration
No data sale
Want to see the methodology in action?
Open EyeQ and ask anything — or run a website, company, person, game, product, decode, or scam check. You will see multi-model analysis turn into one ShouldEye verdict.