Post-study work-pass rules tighten.
“Which finance master’s still lead to a work pass after graduation?”
72 → 58Answers now open with work-pass outcomes. Yours were never published.
Detection Engine
Find out whether AI mentions your brand when people ask the questions that matter to your business. Compare sampled answers, inspect their sources and choose what to improve first.
Part of JefurryAxis, the AI SEO, GEO and AEO product by Jeffery Asia.
Personas built from your own material
With your permission, JefurryAxis reads your de-identified material and your public brand footprint. Its agent mines the audience tags that matter — latent ones included — and combines them into the buyers most likely to ask AI about you.
Your material
- Product & service brochures
- Business & sales material
- Campaign & social content
- Real customer questions
- Public brand entity
De-identified, used only with your permission
JefurryAxis Agent
ReadExtractFilter
Best-fit personas
P1Working professional moving into financeFit 94%
P2Quant hopeful · Singapore88%
P3Promotion seeker · Jakarta81%
- 01Read your material
- 02Extract tags in four dimensions
- 03Rank core and latent audiences
- 04Combine them into best-fit personas
A real test starts with a real question.
We never test “does AI know our brand.” We test the question a real buyer asks in a real week — with your name nowhere in it.
Put your brand in the question and the model will talk about your brand. You will have measured your own prompt.
How one question gets built
P
A working professional moving into finance
ECan’t stop working, and paying for it themselves
?“Which finance master’s can I do while working, and what does it really cost?”
Illustrative. Your profiles and questions are built from your own material.
One Axis run, from square to archive
Five steps, in one sitting. Thirty brand-free questions, six models, every answer kept.
One question set · six models · multiple independent rounds · every answer archived
Persona × environmentLocate one cell first
E1E2E3E4E5P1P2P3P4P5Selected: P1 × E2
One cell = one audience contextQuestion set QS P1·E2The set fixed to that cell
Q1Awarenesswhat existsQ2Comparisonhow they differQ3Rankingwhich is bestOne set per context, never changed
Dispatch to six modelsSame set, six models, in parallel
ChatGPT
Claude
Gemini
DeepSeek
Qwen
Doubao3 international · 3 Chinese, two tracks
Repeat for n roundsEach model repeats n rounds
Repeat for n rounds
Round 1Round 2···Round nRound 1Round 2···Round nRound 1Round 2···Round nRound 1Round 2···Round nRound 1Round 2···Round nRound 1Round 2···Round nOne round is random, n rounds show variance
Collect and archiveEvery answer is archived
Answer sample library
QuestionModelRoundAnswerQ1Round 1Round 2Round nQ2Round 1Round 2Round nQ3Round 1Round 2Round nA comparable, repeatable sample
5 buyer types × 2 situations × 3 questions = the thirty. A re-test runs exactly the same thirty, so this month and last month are the same measurement taken twice.
Why ask the same thing twice?The same model answers differently each time. Sampling turns a lucky screenshot into a range, and every score we report sits in the middle of that range.
Why the two tracks never share a number.International and Chinese models read different corpora and trust different sources. One brand scored 83.0 on one track and 32.6 on the other. The average, 57.8, describes nobody.
Four measurements, then one score
Four questions about one answer, counted the same way every day — which is what makes any two readings comparable.
MRMention rate
Are we in the room at all?
Named in 6 of 10 answers
Read across every sampled answer, not the best one.
OFOccurrence frequency
Mentioned once, or discussed throughout?
Twice inside one answer
Scaled to the length of the answer, so a rambling one cannot flatter itself.
ARAverage rank
First on the list, or ninth?
Third of ten
Not being listed at all counts as last. Otherwise absence would improve your average.
BCRBrand content ratio
How much of the answer is actually about us?
22% of what was written
Sentences about you, over every sentence in the answer.
Why the four are not simply added up
They point in different directions — a low rank is good news, a low mention rate is not — and they are counted in different units. So each is first turned the same way round onto one scale, then weighted. Mention rate carries the most, because nothing compensates for not being there.
The weights are a modelling choice, frozen before anything is reported. A scale you are allowed to retune after seeing the result is not a scale. The full definition travels with every report.
The method, the question set and the test conditions travel with every report. A score without its basis is not a measurement.
The radar: read it one line at a time
Five corners, one per buyer profile. The value is in what you lay over it.
Then all of them at once
Five lines is the real shape of your category. Two corners get circled; everything else waits.
Your brandA competitor
- AttackA rival leads a corner you need.
- DefendYou lead, and the gap is closing.
- MonitorEverywhere else. Spend nothing on it.
Illustrative shapes. A corner can look strong on one track and collapse on the other — which is why the two tracks never share a chart.
Who they are holds still.
What they ask does not.
In B = f(P, Et), the person is the constant and the environment is the variable. A new rule, a story trending on social, a rival’s price cut — the buyer is the same person, but the question they type is not. That moving half is what JefurryInsights ↗, Jeffery Asia’s public-opinion analysis and industry intelligence platform, watches.
A rival’s cancelled-classes story trends on RedNote.
“Which part-time programmes actually run the classes they promise?”
71 → 76The question moved toward you — because your timetable was already public.
Fees rise across the region.
“Which finance master’s are still worth it at this year’s fees?”
75 → 68A new square to win. The brief goes out before the next run.
01JefurryInsights spots the signal
Brand sentiment on social and in AI-readable sources; policy, industry research, competitor moves.
02E is rewritten
The situation your buyer is in this week, updated for each persona it touches.
03New questions join the run
Questions for the new situation go into the next run, beside the thirty you already track — so the old trend stays comparable.
04Re-run, and explain
When the score moves, the report says which signal moved it.
The tracked thirty never change, so the trend stays honest. What changes is how early you see the next question coming. Open JefurryInsights ↗
What a run does not tell you
A sample of AI answers under stated conditions. Not the model’s reasoning, not its training data, not every answer your customers see.
-
Keep the baseline
Questions, profile, language, model, mode, time. All recorded.
-
Compare like with like
Same setup after changes. Variation reported, not smoothed away.
-
Visibility is not revenue
Visits, enquiries and sales live in your CRM. We will not claim credit for them.
Questions about this engine
Which AI platforms can I test?
We support ChatGPT, Claude, Gemini, DeepSeek, Qwen, Doubao, Kimi, Hunyuan, Grok and Perplexity. The models available in a run depend on your plan and test setup. Tencent Yuanbao, Xiaohongshu and Google AI Overviews can be added through team-assisted visibility scoring.
Why use questions that do not name my brand?
They show whether your brand appears when someone is exploring a need or comparing options. Brand-named questions are useful for checking accuracy, and should be reported separately.
Is the score a percentage of the entire AI market?
No. VS is a composite score for the recorded test set. It is not market share, a probability of a sale, or a guarantee that another user will see the same answer.
What should I do after the diagnosis?
Use the findings to prioritise missing information, incorrect claims and weak supporting evidence. The Optimization Engine turns those findings into a content plan.
Be the AI
Recommendation
You already measure rankings, traffic and ad spend closely. AI recommendations deserve the same clarity: how often you are named, where, and who appears beside you. One run gives you that picture.