Methodology
How the ratings are produced, what each scale means and where you should be cautious.
Sources
Facts were collected in September 2026 from providers’ own certification pages and exam guides wherever possible. Where a provider does not publish a detail, or its pages could not be read, third-party sources were used and are marked as such. Each certification page lists its sources and the details that could not be confirmed.
Coverage ratings
Coverage ratings reflect the share of the provider’s published exam outline given to a topic. Providers publish weights by domain, not by topic, so a topic that cuts across several domains is an estimate.
| 0 | Not covered | |
| 1 | Touched on (under 10%) | |
| 2 | Some (10 to 20%) | |
| 3 | Substantial (20 to 35%) | |
| 4 | Core focus (over 35%) |
- GenAI apps
- Building on foundation models: prompting, retrieval-augmented generation, model selection and evaluating output.
- Agents
- Agents, tool use, orchestration, MCP and multi-agent systems.
- Classical ML
- Training, tuning and evaluating machine learning models.
- Operations
- Deployment, pipelines, monitoring and running AI in production.
- Data
- Data preparation and data engineering for AI.
- Responsible AI
- Responsible AI, governance, security and compliance.
Signal ratings
Signal ratings describe the credential, not the person holding it. They are intended to help a hiring manager decide how much weight a credential can bear.
Practical assessment
How far the assessment makes the candidate do the work rather than recall it.
| 0 | Completion or attendance only | |
| 1 | Knowledge quiz, unproctored or open book | |
| 2 | Scenario-based multiple choice | |
| 3 | Includes interactive, code or case-study items | |
| 4 | Performance-based labs or a practical build |
Identity assurance
How confident you can be that the named person did the assessment themselves.
| 0 | No identity check | |
| 1 | Minimal or undocumented checks | |
| 2 | Some monitoring | |
| 3 | Proctored, arrangements vary by provider | |
| 4 | Proctored with identity verification |
Depth
The level the credential is pitched at.
| 0 | Not rated | |
| 1 | Foundational | |
| 2 | Associate | |
| 3 | Professional | |
| 4 | Specialty or expert |
Experience expected
How much prior experience the provider recommends or requires.
| 0 | None | |
| 1 | Under 6 months | |
| 2 | About 1 year | |
| 3 | 2 to 3 years | |
| 4 | 5 years or more |
Portability
How well the knowledge transfers beyond one vendor’s products.
| 0 | Product knowledge only | |
| 1 | Mostly vendor-specific | |
| 2 | Mixed | |
| 3 | Mostly transferable | |
| 4 | Vendor-neutral |
Verifiability
How easily an employer can confirm the credential is real and current.
| 0 | Cannot be verified | |
| 1 | Hard to verify | |
| 2 | Certificate with an ID | |
| 3 | Verifiable on request | |
| 4 | Public badge or registry |
Limits
- Ratings are editorial judgements made against a fixed rubric. They are not measurements.
- A high rating is not a recommendation. A narrow credential can be the right choice for a specific role.
- Some providers do not disclose exam formats or passing scores. Where a rating rests on an inference, the certification page says so.
- This field changes quickly. Several exams were launched, rewritten or retired during 2026. Always check the provider’s page.