AI certifications in 2026: what you can verify before you pay
In this article (10 sections)
- What changed on official pages this month
- The five-point check
- 1. Find the issuer's own page
- 2. Confirm the exam code and version
- 3. Look for retirement or replacement notices
- 4. Check validity and renewal
- 5. Check how it is verified
- What a certification proves, and what it doesn't
- Deciding whether to take one
- Certificates from training programmes
Related programme: the Generative & Agentic AI course (3 months).
Before paying for an AI certification, verify five things on the issuer's own website: that the issuer is who you think it is, that the exam version you will study for is current, whether a retirement or replacement notice exists, how long the credential is valid, and how an employer can verify it. AI certifications change faster than most — during the week this article was written, several official pages showed exam versions ending or being replaced.
What changed on official pages this month
These observations come from the official pages on 27 September 2026. They will go out of date; that is the point of this section. Use it as an example of what to look for, then check the pages yourself.
| Certification (as named by the issuer) | What the official page showed on 27 Sep 2026 |
|---|---|
| AWS Certified Machine Learning Engineer – Associate | The English MLA-C01 exam listed as available until 28 September 2026; an updated MLA-C02 beta starting delivery on 29 September 2026, adding generative and agentic AI, foundation models and responsible AI. The page states a certification earned on either version stays valid for its three-year term. |
| AWS Certified AI Practitioner | Active. The page noted that the Italian and German language versions will be retired after 15 October 2026. The same page lists an AWS Certified AI Business Strategist credential, described as focused on business judgement rather than AWS service knowledge. |
| Microsoft Certified: Azure AI Engineer Associate (exam AI-102) | A banner stating that the certification and its renewal assessment are retired. Check Microsoft Learn for what, if anything, replaces it before studying older AI-102 material. |
| Google Cloud Professional Machine Learning Engineer | Active. Two-hour exam of 50–60 questions; recommended three or more years of industry experience; updated to reflect the transition from Vertex AI to Gemini Enterprise Agent Platform. The page says it does not directly assess coding skill. |
Two lessons from this table. First, a course or video series that prepared people for the old version of an exam may now be teaching the wrong syllabus. Second, a certification can be retired while employers still see it on resumes — it does not become worthless, but you should not start studying for it now.
The five-point check
Run this for any certification, from a cloud vendor or anyone else.
1. Find the issuer's own page
Search on the vendor's domain (aws.amazon.com, learn.microsoft.com, cloud.google.com, or the relevant organisation). Training companies often use certification names in their course titles; the course is not the certification. If you cannot find the credential on the issuer's own site, treat it as a training-provider certificate, not an industry certification.
2. Confirm the exam code and version
Note the exam code and read the current exam guide linked from the official page. Compare its domain list with the syllabus of any prep course you are considering. If they differ, the course is out of date.
3. Look for retirement or replacement notices
These usually appear as a banner or a dated notice near the top of the page. If a version is ending soon, decide whether to sit the old version before its end date or prepare for the new one.
4. Check validity and renewal
Find how long the credential lasts and how renewal works. Budget for renewal if you intend to keep it.
5. Check how it is verified
Most major vendors issue digital credentials that an employer can check online. Make sure yours will be verifiable with a link or ID, and keep that link on your resume.
What a certification proves, and what it doesn't
| A certification can show | It cannot show |
|---|---|
| You studied a vendor's services to a defined standard | That you can build a working system |
| You passed a proctored exam on a known syllabus | How you handle an unclear real problem |
| Familiarity with that vendor's terminology | Judgement about when not to use a service |
This is why most AI hiring processes include a technical round or project discussion regardless of certifications. A certificate plus one well-evidenced project is a stronger combination than several certificates alone. See three LLM portfolio pieces with acceptance criteria for project ideas.
Deciding whether to take one
Ask yourself three questions:
- 1Do the job descriptions I want name this vendor? If most of your target roles mention one cloud, that vendor's certification is more relevant than another's.
- 2Would studying for it teach me something I need anyway? A syllabus that covers deployment, monitoring and security on a cloud you'll use is useful learning even without the badge.
- 3Is the version current for at least the next few months? Avoid starting on a version that ends before you are ready to sit it.
Certificates from training programmes
A certificate of completion from a course, including ours, shows that you completed that course. It is not a vendor certification and should be listed separately on your resume, under education or training. The same applies to a NeuraPath certificate: it records completion of our programme and is not endorsed by any cloud vendor or university. The Generative & Agentic AI add-on focuses on building and evaluating working systems, which you can pair with a vendor exam if your target jobs ask for one. The cloud skills matrix for AI engineers helps decide which cloud skills to prove first.
Frequently asked questions
Which AI certification is best for a job in India?
There is no single best one. A vendor certification is most useful when the jobs you are applying for use that vendor's cloud or tools. Read ten job descriptions for your target role, note which platforms they name, and check the official certification page for that platform before deciding.
Do AI certifications expire?
Many vendor certifications have a validity period and a renewal process, and exams are revised or retired as products change. Check the official page for the validity term and any retirement notice before you book.
Pankit Kumar has 10 years in Data Science & AI, building and shipping production systems in regulated pharma and clinical environments. He is a freelance trainer at Boston Institute of Analytics, AnalytixLabs and Scaler, and has taught this material to thousands of working professionals.
This article is part of our Generative & Agentic AI programme — 3 months. Add practical GenAI, retrieval and agent-building skills to your existing toolkit.
Explore Generative & Agentic AI