VYROX AiSEM

AI Native Structural Equation Modeling Software

AI Does the SEM. You Can Check Every Number.

VYROX AiSEM is free AI native software for CB-SEM and PLS-SEM. Its AI Copilot drafts the model, picks the methods, runs, reviews and writes up the study, while AiSEM's own engine computes, verifies and cites every result. Free, including for business.

Download Free See the Methods
Service Quality, Satisfaction and LoyaltyReal Output, Simulated Data
# reflective constructs
Quality      =~ Q1 + Q2 + Q3
Satisfaction =~ S1 + S2 + S3
Loyalty      =~ L1 + L2 + L3
# structural paths
Satisfaction ~  Quality
Loyalty      ~  Quality + Satisfaction

Auto Mode chose consistent PLS (PLSc), the path scheme and 10,000 bootstrap resamples with a fixed seed.

0Third-party code libraries
124Cited sources behind the rules
0.05 s10,000 bootstrap resamples on the demo model, 12-thread Mac
FreeFree for personal, academic and business use

What Can the AI Copilot Do for an SEM Study?

AiSEM's AI Copilot can run the whole study with no or minimal human input, and AiSEM's statistics engine checks each step. The AI proposes; the engine decides and computes.

Autopilot Available

  • Reads your column summaries and drafts the model
  • Chooses CB-SEM or PLS-SEM, estimator and options
  • Runs, reviews against published guidelines, improves
  • Writes a plain summary, APA results and limitations
  • Full Auto, or Ask Before Each Change
  • Ask the AI any question about your results

Checked, Not Trusted Built In

  • Every AI model is parsed and matched to your columns
  • Rejected models go back to the AI to fix
  • Every setting is checked against the allowed choices
  • Every number in AI text is matched to the report
  • Full activity log of what the AI did and why
  • Never adds paths only to improve fit

Your AI, Your Key Free Options

  • ChatGPT, Claude, Gemini, DeepSeek, Kimi, GLM
  • Grok, Meta Muse, Alibaba Qwen, Mistral, Groq
  • OpenRouter free models and Gemini free tier
  • Private local models with Ollama or LM Studio
  • Keys stay on your computer, masked on screen
  • You choose what data the AI sees
A researcher reviews survey questionnaires beside a laptop showing charts

Built for Theses, Journal Papers and Business Research

From a master's thesis to a customer study, every result comes with the reason and the source a reviewer will ask for.

  • Postgraduate students and supervisors
  • Academic researchers and journal authors
  • Market research, HR and customer analytics teams
  • Consultants who report findings to clients

What Are PLS-SEM and CB-SEM?

Both are structural equation modeling (SEM) methods for testing how unobserved concepts, such as trust or satisfaction, relate to each other.

PLS-SEM

PLS-SEM (partial least squares structural equation modeling) estimates relationships between constructs by building weighted composites of their indicators, maximising the variance explained in the outcome constructs.

  • Prediction and explaining outcomes
  • Formative constructs and complex models
  • Smaller samples and non-normal data

CB-SEM

CB-SEM (covariance-based structural equation modeling) estimates a model by reproducing the observed covariance matrix, usually with maximum likelihood, and tests how well a theory fits the data.

  • Confirming an established theory
  • Global fit: chi-square, CFI, TLI, RMSEA, SRMR
  • Common-factor measurement models
PLS-SEM vs CB-SEM at a Glance
QuestionPLS-SEMCB-SEM
Main goalPrediction, explained varianceTheory confirmation, model fit
Constructs treated asComposites (PLSc for common factors)Common factors
EstimatorIterative least squaresMaximum likelihood and variants
Sample sizeWorks with smaller samplesUsually needs larger samples
In AiSEMAvailable NowAvailable Now

VYROX AiSEM vs Other SEM Software

34 statistical features across 19 PLS-SEM and CB-SEM programs, checked against the official vendor pages and manuals we could read. AiSEM's column shows what ships today. The AI Copilot is not part of this table.

Feature comparison of PLS-SEM and CB-SEM software
Feature VYROX AiSEMFree, Incl. Business SmartPLSPaid; Limited Free Student Tier IBM SPSS AmosPaid MplusPaid LISRELPaid Stata sem/gsemPaid EQSPaid WarpPLSPaid; Free Trial ADANCOFree for Individuals Only XLSTAT PLS-PMPaid lavaan (R)Free (GPL) seminr (R)Free (GPL) cSEM (R)Free (GPL) OpenMx (R)Free (Apache) semopy (Python)Free (MIT) JASPFree (AGPL) jamovi + modulesFree GSCA ProFree OnyxFree
Licence and Platform
Free of Charge Available Partial No No No No No Partial Partial No Yes Yes Yes Yes Yes Yes Yes Yes Yes
Free for Business Use Available Not Confirmed No No No No No No No No Yes Yes Yes Yes Yes Yes Partial Not Confirmed Not Confirmed
Windows Available Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
macOS Available Yes No Yes Partial Yes Partial No Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes
Linux Available Yes No Yes Not Confirmed Yes Partial No No No Partial Yes Yes Yes Yes Yes Yes No Yes
Graphical Interface Available Yes Yes Partial Yes Yes Yes Yes Not Confirmed Yes No No No No No Yes Yes Yes Yes
Draw Path Diagram to Build Model Available Yes Yes Partial Not Confirmed Yes Yes Yes Not Confirmed Yes No No No No No No No Yes Yes
Text Model Syntax Available Not Confirmed Partial Yes Yes Yes Yes No Not Confirmed No Yes Yes Yes Partial Yes Yes Yes No No
PLS-SEM
PLS-SEM Algorithm Available Yes No No No No No Yes Yes Yes No Yes Yes No No Yes Partial No No
Consistent PLS (PLSc) Available Yes No No No No No Partial Not Confirmed Not Confirmed No Yes Yes No No Partial Partial No No
HTMT Available Yes No No No No No Yes Not Confirmed Not Confirmed No Yes Yes No No Partial Partial No No
PLSpredict Available Yes No No No No No Partial Not Confirmed Not Confirmed No Yes Yes No No Not Confirmed Partial No No
Multigroup PLS (PLS-MGA) Available Yes No No No No No Yes Not Confirmed Yes No Partial Yes No No Not Confirmed Partial No No
MICOM Available Yes No No No No No Partial Not Confirmed Not Confirmed No No Yes No No Not Confirmed Not Confirmed No No
FIMIX-PLS Available Yes No No No No No No Not Confirmed Not Confirmed No No No No No Not Confirmed Not Confirmed No No
IPMA Available Yes No No No No No No Not Confirmed Not Confirmed No No Yes No No Not Confirmed Not Confirmed No No
Necessary Condition Analysis (NCA) Available Yes No No No No No No Not Confirmed Not Confirmed No No No No No Not Confirmed Not Confirmed No No
Bootstrap Confidence Intervals Available Not Confirmed Yes Yes Not Confirmed Yes Yes Yes Not Confirmed Yes Yes Yes Yes Yes Not Confirmed Yes Yes Not Confirmed Not Confirmed
CB-SEM
Maximum Likelihood Available Yes Yes Not Confirmed Not Confirmed Yes Yes No Not Confirmed No Yes Yes No Yes Yes Not Confirmed Yes No Yes
Robust ML (Satorra-Bentler, MLR) Available Not Confirmed No Not Confirmed Not Confirmed Yes Yes No Not Confirmed No Yes Yes No Partial Partial Not Confirmed Yes No Not Confirmed
Ordinal Data (Polychoric, WLSMV, DWLS) Available Not Confirmed Partial Not Confirmed Partial Partial Yes Not Confirmed Not Confirmed No Yes Partial Yes Yes Partial Not Confirmed Yes Partial Not Confirmed
FIML for Missing Data Available Not Confirmed Yes Partial Not Confirmed Yes Partial Partial Not Confirmed No Yes Partial No Yes Yes Not Confirmed Yes Partial Not Confirmed
Fit Indices (CFI, TLI, RMSEA, SRMR) Available Partial Partial Not Confirmed Not Confirmed Yes Not Confirmed Partial Not Confirmed No Not Confirmed Partial Yes Yes Partial Partial Yes Partial Not Confirmed
Modification Indices Available Not Confirmed Yes Yes Not Confirmed Yes Partial No Not Confirmed No Yes Not Confirmed No Yes Not Confirmed Yes Yes Not Confirmed Not Confirmed
Bollen-Stine Bootstrap Available Not Confirmed Yes Not Confirmed Not Confirmed No Yes No Not Confirmed No Yes Not Confirmed No No Not Confirmed Not Confirmed Not Confirmed Not Confirmed Not Confirmed
Bayesian SEM Available No Yes Yes Not Confirmed No Not Confirmed No Not Confirmed No No No No No No No No No Not Confirmed
Multilevel SEM Partial Not Confirmed No Yes Partial Yes Yes Partial Not Confirmed No Yes No No Partial Partial Not Confirmed Partial Yes Not Confirmed
Mixture and Latent Class Models Available Partial Partial Yes Not Confirmed Yes Not Confirmed No Not Confirmed No No No No Yes No Not Confirmed No Not Confirmed Not Confirmed
Measurement Invariance Testing Available Yes Yes Yes Not Confirmed Yes Partial Partial Not Confirmed No Yes No Partial Partial Not Confirmed Yes Partial Partial Partial
Mediation and Indirect Effects Available Yes Yes Yes Partial Yes Partial Yes Not Confirmed Not Confirmed Yes Yes Yes Partial Not Confirmed Yes Yes Yes Not Confirmed
Monte Carlo Simulation and Power Available Partial Not Confirmed Yes Not Confirmed Not Confirmed Yes Partial Not Confirmed Not Confirmed Partial No Partial Yes Partial Not Confirmed No Yes Partial
Automation and Reporting
Chooses the Estimator from the Data Available Not Confirmed No Not Confirmed Not Confirmed No Not Confirmed Not Confirmed Not Confirmed No Partial No No No No Not Confirmed Partial No No
Cites a Published Source for Every Rule Available Not Confirmed No No Not Confirmed No Not Confirmed Partial Not Confirmed No No No No No No Not Confirmed No No No
Writes the Methods Paragraph Available Not Confirmed No No Not Confirmed No Not Confirmed Not Confirmed Not Confirmed No No No No No No Not Confirmed No No No
Confirmed Feature Score (Yes 1, Partial 0.5) 33.5 19 13 13.5 5 16.5 14 13 3.5 6 16.5 16 18 15.5 10.5 13.5 18.5 10 8

Checked 13 November 2026 against each vendor's official website, manuals, CRAN or PyPI pages, using the version shown (or the latest at that date). "Not confirmed" means the official pages we could read do not say either way; it is not a claim the feature is missing. Paid tiers and add-ons count as Partial. Product names are trademarks of their owners; VYROX is not affiliated with or endorsed by them. Today AiSEM scores 33.5 of 34, the highest confirmed score in this table; the roadmap targets all 34, and this table is updated with every release.

Every Decision Explained, Every Number Sourced

Auto Mode Does the Method Choices

It screens your data, picks the estimator, runs the assessment checklist and tells you in plain words what passed, what failed and what to do.

Screen DataChoose EstimatorBootstrapCheck ThresholdsWrite Report

A Paper Behind Every Rule

Each threshold and default names its source, from Wold (1982) to Hair et al. (2019).

Uses Every CPU Core

Bootstrapping runs in parallel with a fixed seed, so the same data always gives the same intervals.

Statistics Run Locally

The statistics run on your own computer. No account. The AI Copilot sends only what you choose to the provider you pick, or nothing at all with Ollama or LM Studio.

Report Ready to Cite

APA-style tables, a methods paragraph and the reference list for exactly the methods used.

Methods in AiSEM

What ships today and what is planned. Each method is built from its original paper.

PLS-SEM Available

  • PLS path algorithm: path, factor, centroid
  • Reflective (Mode A) and formative (Mode B)
  • Consistent PLS (PLSc) with safe fallback
  • Alpha, rho_A, rho_c, AVE
  • HTMT and Fornell-Larcker
  • VIF, f², R², adjusted R²
  • Bootstrap percentile intervals
  • Minimum sample size check
  • PLSpredict with LM benchmark
  • Multigroup (PLS-MGA) and MICOM
  • Importance-performance map (IPMA)
  • FIMIX-PLS hidden segments
  • Necessary condition analysis (NCA)

CB-SEM Available

  • Maximum likelihood (ML)
  • Robust ML (Satorra-Bentler), GLS, ULS, ADF
  • Ordinal Likert items: polychoric correlations and WLSMV
  • FIML for missing data
  • Chi-square, CFI, TLI, RMSEA with 90% CI, SRMR
  • Modification indices and expected change
  • Bollen-Stine bootstrap, Mardia tests
  • Mediation for CB-SEM and PLS-SEM
  • Monte Carlo power analysis

Advanced Available

  • Bayesian SEM, 4 parallel chains, R-hat, ESS, PPP, DIC
  • Two-level SEM with ICCs, within and between models
  • Latent class and latent profile analysis with BLRT

Next Planned

  • Moderation and moderated mediation
  • CB-SEM multigroup invariance
  • CVPAT, blindfolding Q², higher-order constructs
  • Endogeneity tests, growth models

How to Run a PLS-SEM Analysis

Load Your Data

Open the AiSEM app and drop in your survey as CSV or TSV, one column per item. You see a preview and the missing-value count at once.

Draw the Model

Press Suggest Model to group items by name, then click to connect the paths. The diagram writes the model text for you.

Trust   =~ T1 + T2 + T3
Loyalty =~ L1 + L2
Loyalty ~  Trust

Read the Report

Press Run Analysis. The report with checks, tables, a methods paragraph and references opens in the app. Prefer scripts? The same engine runs from the command line.

aisem analyze --data survey.csv \
  --model model.txt

Questions About PLS-SEM, CB-SEM and AiSEM

What is PLS-SEM?

PLS-SEM (partial least squares structural equation modeling) is a variance-based method that estimates relationships between constructs by building weighted composites of their indicators. It suits prediction, formative constructs, complex models and smaller samples.

What is CB-SEM?

CB-SEM (covariance-based structural equation modeling) estimates a model by reproducing the observed covariance matrix, usually with maximum likelihood. It suits confirming an established theory and gives global fit indices such as chi-square, CFI, TLI and RMSEA.

What is the difference between PLS-SEM and CB-SEM?

CB-SEM tests how well a theory reproduces the covariances in the data and treats constructs as common factors. PLS-SEM maximises explained variance in the outcome constructs and treats constructs as composites. Use CB-SEM to confirm theory; use PLS-SEM for prediction, formative measurement or exploratory models.

Is AiSEM free for commercial use?

Yes. AiSEM is free to download and use for personal, academic and business work. You may share the unmodified download free of charge, but you may not sell or modify it.

Does AiSEM support CB-SEM now?

Yes. Since version 0.2 AiSEM estimates CB-SEM by maximum likelihood, robust ML (Satorra-Bentler) or FIML for missing data, chosen automatically from the data, and reports chi-square, CFI, TLI, RMSEA with its 90% CI, SRMR, modification indices, the Bollen-Stine test and indirect effects.

Does AiSEM send my data anywhere?

The statistics always run on your own computer. Only when you use the AI Copilot does AiSEM send what you choose (column names, column summaries, or summaries plus the first 15 rows) to the AI provider you picked. With Ollama or LM Studio on your computer, nothing leaves it.

What does AI native mean in AiSEM?

AI does as much of the study as it safely can: it reads your columns, drafts the model, chooses the methods, runs the analysis, reviews the results against published guidelines, improves the model and writes up the findings. AiSEM's statistics engine checks every AI suggestion before use and computes every number, and every number in AI-written text is verified against the report.

Which AI models can AiSEM use?

Your own API key for OpenAI (ChatGPT), Anthropic (Claude), Google Gemini, DeepSeek, Moonshot (Kimi), Z.ai or Zhipu (GLM), xAI (Grok), Meta (Muse), Alibaba Cloud Model Studio (Qwen), Mistral, Groq or OpenRouter, any OpenAI-compatible service, or a free model on your own computer through Ollama or LM Studio. Free options include the Gemini, GLM Flash, Mistral and Groq free tiers and OpenRouter's free models.

How does AiSEM choose between PLS and PLSc?

When a model has reflective constructs, AiSEM uses consistent PLS (Dijkstra and Henseler, 2015), which corrects the bias of ordinary PLS for common-factor constructs. If PLSc gives an inadmissible result, it falls back to ordinary PLS and explains why in the report.

How large a sample does PLS-SEM need?

AiSEM applies the inverse square root method (Kock and Hadaya, 2018): to detect the smallest path coefficient with 80% power at 5% significance, you need more than (2.486 divided by that coefficient) squared cases. The report states the number for your model.

Download VYROX AiSEM Free

Version 0.7.0-alpha.1 includes the AiSEM app with the AI Copilot: double-click it and it opens in your browser, with Autopilot, a visual path diagram editor, guided steps and saved projects. The statistics run on your own computer.

Download Free Quick Start Guide
Windows 64-bit, .zip
macOS Apple Silicon and Intel
Linux x86-64, .tar.gz