AI Learning Outcome Analyzer
Measure what students actually learned, fast and clearly
NVIDIA: Nemotron 3 Super
Balanced Nemotron for demanding everyday work
NEW
FREE
Your prompt will appear here…
Your beautifully formatted article will appear here once you generate.
No history yet
Your generations will appear here. Sign in to save them permanently.
Were the outcomes achieved, or did the course simply happen? Every unit states what learners will be able to do by the end of it, and remarkably few of those statements are ever checked against what learners could actually do.
Outcomes are claims. Analysing them means testing the claim against the evidence you have.
Short answer: The AI Learning Outcome Analyzer is a free tool that analyses whether stated learning outcomes were achieved. You supply the outcomes and the evidence, and it reports which are supported, which are partly met and which the evidence cannot speak to at all.
What is AI Learning Outcome Analyzer?
It analyses outcomes against evidence. The subject is not the teaching and not the learners individually, but the specific claims a course makes about what people will be able to do afterwards.
Why Use AI Learning Outcome Analyzer?
Because outcome statements are usually reviewed by being reread rather than by being tested.
| How outcomes get reviewed | What that establishes | What analysis establishes |
|---|---|---|
| Rereading the statements | That they sound reasonable | Whether evidence supports them |
| Pass rates | That assessment was passed | Whether assessment tested the outcome |
| Learner satisfaction | That the course was liked | Nothing about capability |
| Assuming coverage equals achievement | That it was taught | Whether it was learned |
How Does AI Learning Outcome Analyzer Work?
All tools on the site sit on one working surface, with analysis behind this button.
- Prompt box. The box is labelled for whatever you want analysed, pasted or described. Give it the outcomes and the evidence together.
- Model selector. The engine is set here, with MSB AI, Google Gemini, xAI Grok AI among others.
- Advanced options. Ten controls behind the accordion, described below.
- Generate. Outcomes, evidence and settings pass through prompt engineering written for analysis.
- Result card. The analysis lands with a word count shown beneath.
- Export row. The export row offers DOC, TXT and HTML.
- Activity history. Earlier analyses stay available with copy, listen, reuse, download and open result, so successive cohorts can be compared.
Key Features
Gaps as a focus
Analysis Focus includes Gaps, which finds the outcomes your evidence cannot speak to at all.
Evidence weighed
Ask for it and each outcome is judged on the strength of its evidence rather than on whether it was taught.
Scorecard output
A scorecard puts outcomes against verdicts, which is what a course review needs on the table.
Unsupported claims flagged
Flag Risks names the outcomes that would not survive scrutiny, which is better found internally.
Rigour on a slider
Rigor from 1 to 100 separates a quick review from something you could defend in a validation meeting.
Advanced Options Guide
Ten controls. Analysis Focus and Rigor together decide how hard the outcomes get tested.
| Option | What it controls | Setting for outcome analysis |
|---|---|---|
| Analysis Focus | Overview, Strengths & Weaknesses, Opportunities, Risks, Trends, Gaps, Comparison or Recommendations | Gaps, then Recommendations for the next cohort |
| Analysis Depth | Quick, Standard, Deep or Comprehensive | Deep for a formal review |
| Output Format | Summary, Detailed Report, Bullet Points, Table, Scorecard or SWOT | Scorecard, outcomes against verdicts |
| Priority Lens | Accuracy, Impact, Risk, Cost, Speed, Quality, Growth or Clarity | Accuracy, since this is about whether a claim holds |
| Extract Key Findings | Pulls the essentials out | On |
| Flag Risks | Names unsupported outcomes | On |
| Give Recommendations | Suggests what would fix each gap | On |
| Age-Appropriate Language | Adjusts wording | Off for a staff document |
| Rigor | Slider from 1 to 100 | Around 75 |
| Custom Instructions | Free text up to 1000 characters | Ask it to distinguish taught, assessed and demonstrated for every outcome |
Example Inputs
Farah is reviewing a unit whose pass rate is high and whose employers are unimpressed. She opens the AI Learning Outcome Analyzer with the outcomes and the evidence side by side.
Stated outcomes for the unit:
1. Explain the main legal duties of an employer.
2. Identify a hazard in a workplace scenario.
3. Complete a risk assessment for a real workplace.
4. Communicate findings to a non specialist manager.
Evidence available:
Assessment was a 40 question multiple choice paper plus a
written risk assessment on a supplied scenario. Pass rate
94%. No observed practical activity. No presentation or
verbal component. Employer feedback says new starters can
recite duties and cannot spot obvious hazards on site.
Analysis Focus = Gaps
Analysis Depth = Deep
Output Format = Scorecard
Priority Lens = Accuracy
Extract Key Findings = On
Flag Risks = On
Give Recommendations = On
Age-Appropriate Language = Off
Rigor = 75
Custom Instructions = For each outcome say whether it was
taught, assessed and demonstrated, as three separate
judgements. Flag any outcome the evidence cannot support at
all and say what evidence would be needed.
Example Outputs
The three way split was the finding. Outcome one was taught, assessed and demonstrated. Outcome three was taught and assessed on a supplied scenario, which is weaker evidence than a real workplace but not nothing. Outcomes two and four were taught and never assessed in any form that could demonstrate them.
That explains the employer feedback exactly. A high pass rate on a paper that never asks somebody to look at a site cannot support a claim about spotting hazards on a site, and the fix is an observed activity rather than a better multiple choice paper.
Pro tip Ask for taught, assessed and demonstrated as three separate judgements per outcome. Almost every problem shows up in the gap between the second and the third, and a single verdict per outcome hides it completely.
Note Outcome analysis and programme evaluation overlap. If the concern is whether teaching produced change across a cohort rather than whether specific claims hold, the AI Learning Assessment Tool is aimed at that question.
Caution An analysis is only as good as the evidence described, and describing your assessment generously is the standard failure here. Say what the assessment actually required somebody to do, not what it was intended to measure, because those two sentences frequently differ.
Tips & Common Mistakes
- ✅ Paste the outcome statements verbatim
- ✅ Describe what the assessment physically required
- ✅ Ask for taught, assessed and demonstrated separately
- ✅ Include any external feedback you have
- ✅ Ask what evidence would close each gap
| Common mistake | What it produces | The fix |
|---|---|---|
| Assuming coverage is achievement | Outcomes marked met because they were taught | Separate taught from demonstrated |
| Describing intent, not the assessment | An analysis of what you meant to test | Say what candidates actually had to do |
| One verdict per outcome | The assessment gap stays hidden | Three judgements per outcome |
| Ignoring employer or placement feedback | Internal evidence only | Include external signals |
What works well
- Separates taught from assessed from demonstrated
- Finds outcomes the assessment cannot support
- Says what evidence would close each gap
- Free to use, with no account needed
What to watch for
- Generous descriptions of assessment produce generous findings
- It cannot see your assessment papers unless you describe them
- External feedback is hearsay unless you bring it
- An analysis does not redesign the assessment
AIToolsay is a free platform with a large library of AI tools, and this analyser sits with the learning analytics group. Every tool takes one job, comes with its own controls, and runs on prompt engineering written for it, which is why an outcome analyser tests claims while a curriculum tool builds coverage. Nothing installs and no account is needed. The engine list includes MSB AI, OpenAI ChatGPT, Google Gemini, Anthropic Claude AI, DeepSeek and more. Related tools are listed on the AIToolsay homepage.
Frequently Asked Questions
Is the AI Learning Outcome Analyzer free?
Yes, and the page works without an account.
What evidence should I supply?
What the assessment required, the results, and any external feedback. Describe the assessment in terms of what candidates physically had to do.
Why split taught, assessed and demonstrated?
Because most outcome problems live between assessed and demonstrated. A paper can assess an outcome without ever demonstrating it.
Can a high pass rate hide a failed outcome?
Routinely. A pass rate tells you about the assessment, and if the assessment does not test the outcome the two are unrelated.
Is this useful before validation?
Yes, and it is better to find an unsupported outcome yourself. Set Rigor high and ask specifically what would not survive scrutiny.
What do I do with a flagged outcome?
Change the assessment or change the claim. Both are legitimate, and leaving a claim that cannot be evidenced is not.
An outcome nobody has tested is a claim, and courses accumulate those quietly. Splitting taught from assessed from demonstrated takes an afternoon and tends to explain whatever external feedback has been puzzling you.
So open the AI Learning Outcome Analyzer, paste the outcomes exactly as written, describe what your assessment really required, and ask for three judgements on each one. Thanks for reading, and I hope fewer of them are unsupported than Farah's were. If it becomes part of your review cycle, the AIToolsay community is open to you, our social accounts announce every new tool, push notifications reach you first, and the newsletter carries guides in this style.
Let AI Speak.