AI4LD

Methodology

Evidence, Not Hype

AI4LD Tool Intelligence exists so learning design professionals can evaluate AI tools against learning science — not vendor claims — before committing time, budget, or learner data to them. It is free and ungated because that decision shouldn't require a sales call.

How we evaluate a tool

Five steps, applied to every tool in the directory before it gets a score.

1

Official documentation first

Every evaluation starts with the vendor’s own docs, pricing pages, and security/compliance disclosures — not marketing copy or third-party summaries.

2

Second-source validation

Vendor claims are checked against at least one independent source — review sites, analyst coverage, customer case studies, or public compliance registries — before they inform a score.

3

Recency capture

AI tools ship fast. Each entry notes when it was last evaluated, and tools with functionality still in beta or with unpublished pricing are flagged as such rather than scored optimistically.

4

Corporate-practice focus

Scoring is grounded in how the tool actually gets used inside L&D and instructional design workflows — not its full feature set, and not its fit for engineering or marketing teams.

5

Comparison views

Every tool is scored on the same 8 criteria so it can be placed directly next to any other tool in the directory — no criteria are added or dropped to flatter a particular vendor.

Vendor claims vs. user evidence

Each tool's evidence field distinguishes what a vendor asserts about itself from what is independently verifiable — certifications, published customer usage, third-party coverage. Where only vendor claims exist, that is stated plainly rather than presented as confirmed fact. The ai4ldNote on each entry gives our own read on fit and confidence, including where the underlying evidence is thin.

Limits, risks, privacy, and governance

  • Scores reflect a point in time. AI product capabilities, pricing, and compliance posture change quickly — verify current details directly with the vendor before acting on any score here.
  • The AI recommendation on the Recommend page is generated by Claude and grounded only in the tools and scores in this directory — it can still be wrong, and it always names concrete things to verify before you commit.
  • This tool collects no account information, requires no login, and sets no email gate. Recommendation requests are processed to generate a response and are not used to build a profile of you.
  • Governance and accessibility fields describe what the vendor discloses publicly. Absence of a claim (e.g. no SOC 2 mentioned) is noted as a gap, not assumed to mean the control doesn't exist — always confirm directly with the vendor for any procurement or compliance decision.

Where the 8 criteria come from

The 8 criteria aren't arbitrary. Each one traces back to a real evaluation framework — for software quality, accessibility, or instructional design — rather than being invented for this site. The specific source for each is listed with it below. One exception: Cost Value has no academic equivalent, so it's labeled plainly as our own procurement judgment call, not a research citation.

The 8-criteria rubric

Every tool in the directory is scored 1-5 on each of these criteria. The full level descriptors are below so a score of, say, 3 on Analytics means the same thing no matter which tool you're looking at.

LXD Workflow Fit

How well the tool integrates into learning experience design workflows

Grounded in: Task-Technology Fit (Goodhue & Thompson, 1995, MIS Quarterly 19(2)) — the correspondence between task requirements and technology functionality, not general acceptance (TAM/UTAUT).

  1. 1Minimal integration with ID/LXD workflows
  2. 2Basic features, significant manual work required
  3. 3Moderate integration, some workflow alignment
  4. 4Good workflow integration, reduces design friction
  5. 5Excellent fit, significantly streamlines LXD work

Quality & Control

Capabilities for ensuring content quality and accuracy

Grounded in: Blend: ISO/IEC 25010:2023 (Functional Suitability, Reliability) and NIST AI RMF 1.0 ("valid and reliable," "accountable and transparent").

  1. 1No version control, limited editing
  2. 2Basic editing, minimal review features
  3. 3Decent editing and versioning capabilities
  4. 4Strong review loops and version management
  5. 5Comprehensive QC with provenance tracking

Personalization

Ability to deliver personalized learning experiences

Grounded in: Edu-GenAI Rubric (Cherner & Donnelly, 2026, Education Sciences 16(5), 706, doi.org/10.3390/educsci16050706), which names personalization as a distinct evaluation dimension.

  1. 1No personalization capabilities
  2. 2Basic recommendations only
  3. 3Moderate personalization features
  4. 4Good adaptive capabilities
  5. 5True adaptive learning with mastery-based paths

Analytics

Depth and usefulness of analytics and reporting

Grounded in: The learning-analytics field (Siemens & Long, 2011, "Penetrating the Fog," EDUCAUSE Review 46(5)) and SoLAR's standing definition of learning analytics.

  1. 1Minimal or no analytics
  2. 2Basic completion tracking only
  3. 3Standard reporting and dashboards
  4. 4Advanced analytics with useful insights
  5. 5Deep analytics with predictive capabilities

Enterprise Ready

Security, compliance, and scalability for enterprise use

Grounded in: Blend: ISO/IEC 25010:2023 (Security, Reliability, Flexibility) and NIST AI RMF 1.0 ("secure and resilient," "accountable and transparent").

  1. 1Not suitable for enterprise use
  2. 2Basic security, limited admin controls
  3. 3Moderate enterprise features
  4. 4Strong security, SSO, compliance support
  5. 5Full enterprise: SOC2, audit trails, SLA

Accessibility

Support for accessibility standards and diverse learners

Grounded in: W3C Web Content Accessibility Guidelines (WCAG) 2.1/2.2 — the four POUR principles (Perceivable, Operable, Understandable, Robust).

  1. 1No accessibility features
  2. 2Basic accessibility only
  3. 3Partial WCAG compliance
  4. 4Good accessibility support
  5. 5WCAG 2.1 AA+ compliant, fully inclusive

Interoperability

Integration capabilities with other systems and standards

Grounded in: ISO/IEC 25010:2023 Compatibility characteristic, plus the ADL Initiative's SCORM and xAPI (Experience API) specifications.

  1. 1Closed system, no integrations
  2. 2Limited integration options
  3. 3Basic API, some connectors
  4. 4Good standards support (SCORM, xAPI)
  5. 5Extensive APIs, full standards compliance

Cost Value

Value relative to pricing (higher score = better value)

Grounded in: No ID/HPT or software-quality construct covers this — it's AI4LD's own procurement dimension, borrowing the term (not a specific study) from Gartner's Total Cost of Ownership methodology.

  1. 1Very high cost, significant lock-in
  2. 2Premium pricing, implementation complex
  3. 3Moderate cost and complexity
  4. 4Good value, reasonable implementation
  5. 5Excellent value, low lock-in risk