AI Readiness Index Dashboard
Not a ranking. Not a single score. This dashboard shows what our open data can already support, segment by segment — with every figure carrying its sample size and source. The full ranked edition, built on the published methodology, is still in progress.
STATUS: DATA EDITION — SEGMENT DASHBOARDS ONLY · NO COMPOSITE SCORE · NO RANKING
Three segments, one table
Policy adoption is the only dimension measured in all three segments. Teaching is measured only in Central Asia. Anything marked N/A was never measured there — not zero, not estimated.
| Segment | n | Publishes AI guidance | Teaches AI |
|---|---|---|---|
| US Carnegie R1 | 187 universities | 178 (95.2%) | N/A — not measured |
| Global policy audit | 186 records | 144 (77%) | N/A — not measured |
| Central Asia | 31 universities | 6 (19%) | 18 (58%) |
Sources: US R1 AI Policy Audit 2026 · Global University AI Policy Audit 2026 · Central Asia AI Education Audit 2026. All datasets CC-BY-4.0.
Near-saturation — with one holdout
Of 187 Carnegie R1 universities, 178 (95.2%) publish AI guidance on their own domains, 8 are unverifiable from public sources, and exactly 1 is a confirmed NO: the University of Wyoming. Excluding the 8 unverifiable cases, 178 of 179 universities with a definitive verdict (99.4%) publish guidance.
The 8 unverifiable cases — Brown, Clemson, UT Health San Antonio, UT Southwestern, UMass-Boston, Miami, UNH, and Weill Cornell Medicine — each show real AI activity (committee reports, program-level rules, intranet-gated pages) but no adopted, public, university-wide policy on their own domain. Every verdict traces to one page per university.
Adoption falls as rank falls
Across 186 audit records, 144 (77%) publish AI guidance, 24 do not, and 18 are unverifiable. The sample splits show a clear gradient — and a cliff outside the QS sample:
| Sample | n | Publish guidance | Confirmed no | Unverifiable |
|---|---|---|---|---|
| QS top 50 | 50 | 46 (92%) | 1 | 3 |
| QS ranks 51–100 | 49 | 43 (88%) | 0 | 6 |
| QS ranks 101–150 | 50 | 38 (76%) | 6 | 6 |
| Global South flagships | 37 | 17 (46%) | 17 | 3 |
Sample bands sum to 186 records: 149 QS-band records plus 37 Global South flagships.
The gap between the top 50 (92%) and Global South flagships (46%) is 46 points. Of the 37 flagships, 20 lack published guidance — 17 confirmed with no policy and 3 unverifiable.
Teaching three times faster than governing
Of 31 leading universities across 5 Central Asian countries, 6 (19%) publish a verifiable public AI usage policy on their official domain, while 18 (58%) offer a verifiable AI-named degree program or AI course portfolio. Institutions are teaching AI roughly three times faster than they govern its use.
The other 25 (81%) are UNVERIFIABLE, not confirmed policy-free: no policy could be confirmed on the public web, often because sites were down, login-walled, or crawler-blocked. That is a finding about transparency as much as about adoption — and it is the highest unverifiable rate of any region measured. Zero campus-wide AI bans were found; every verified policy follows the global norm of delegated permission plus mandatory disclosure.
NOTE: No per-country breakdown is published. The policy and program files do not share a clean join key, so per-country rates would be unreliable. Totals only.
Three panels that frame the segments
These panels are context, not readiness inputs. They are labeled separately so nothing below is mistaken for a dimension score.
The price is usually estimated
The tuition audit covers 740 online programs at 537 US institutions. 587 listings (79%) required computed estimates — the real total price was not findable on official sites — and only 153 (21%) show a verified total price. By level: bachelor's 299, MBA 228, MCS 108, BSCS 67, DBA 38.
Five operator-owned ACLAS rows (MBA $89, BBA $159, DBA $199, BSCS $199, MCS $299) are flagged as such, all with verified prices. Because estimates dominate, tuition is context here, not a readiness input.
Most universities clear most rungs
The global recognition census checks 16,822 universities against four signals — swot, Wikidata, ROR, OpenAlex. 6,817 score 4/4 (41%), 5,801 score 3/4 (34%), 1,109 score 2/4 (7%), and 3,095 score 1/4 (18%).
Among the earned signals, swot domain coverage is scarcest at 47%, below ROR (75%) and OpenAlex (75%). Wikidata is the census frame, so it reads 100% by construction.
Provisioned AI can mislead
Readiness is not only policy. A 0.5B RAG tutor tested on 30 questions across 3 retrieval configs (90 generations, hand-graded) hallucinates on unanswerable questions at rates that depend heavily on retrieval settings:
| Config | Chunk | top_k | Hallucination (unanswerable) | Refusal | Citation accuracy (answerable) |
|---|---|---|---|---|---|
| A | 500 | 3 | 7/10 (70%) | 3/10 | 4/10 (40%) |
| B | 1000 | 3 | 5/10 (50%) | 4/10 | 5/10 (50%) |
| C | 500 | 5 | 3/10 (30%) | 5/10 | 6/10 (60%) |
Retrieval was not the bottleneck: the correct chunk was in the top-k for all 10 answerable questions. The failures are in generation. Raising top_k from 3 to 5 is the cheapest win available — it more than halves hallucination and lifts citation accuracy from 40% to 60%.
What each figure measures — and does not
| Dimension | Measures | Does NOT measure |
|---|---|---|
| Policy adoption | Whether an adopted AI policy exists and is public on the institution's official domain | Content depth, quality, or enforcement of the policy |
| Teaching (Central Asia only) | Whether a verified AI-named degree program or AI course portfolio exists on the official domain | Program quality, enrollment, or graduate outcomes |
| Tuition (context) | Listed or computed total program price per official site, with estimate basis per row | Value for money, or what students actually pay after aid |
| Recognition (context) | Presence in swot, Wikidata, ROR, and OpenAlex | Academic quality or legitimacy beyond database presence |
| RAG reliability (context) | Hallucination, refusal, and citation rates of one 0.5B tutor on 30 fixed questions | Performance of larger models, other prompts, or reranked retrieval |
This dashboard does not rank universities, does not publish a composite score, and does not compare segments as a league table: the samples and instruments differ, so a cross-segment ranking would be misleading. For the full methodology — weights, rubric, sensitivity analysis, and ranking rules — see the Readiness Index methodology preprint.
Read these before quoting a number
- Small Central Asia n. 31 universities cannot carry fine-grained claims; totals only, no per-country rates.
- Tuition is mostly estimated. 79% of listings required computed estimates; treat tuition figures as audit findings, not price guarantees.
- R1 means US only. The 95.2% figure describes Carnegie R1 universities, not the world.
- The global audit is not a world census. It covers QS-band universities plus 37 Global South flagships — a structured sample, not every institution.
- Audits are 2026 snapshots. Publication is not enforcement, and policies change after audit dates.
- RAG results are one small model. A single 0.5B model, one prompt template, no reranker. Larger models will differ; the direction (more context, fewer hallucinations) is the finding, not the exact rate.
Dashboard FAQ
01Why is there no single readiness score?
Because the segments use different samples and different instruments. Collapsing them into one number would hide weak dimensions — the methodology preprint states the sub-scores are the product and any composite is a convenience. The ranked edition comes later, when teaching and infrastructure are measurable at the same scale.
02Why is teaching marked N/A for R1 and the global audit?
Those audits measured policy publication only. Teaching was measured only in the Central Asia audit, which checked for verified AI-named degree programs and course portfolios. We do not impute teaching rates for segments that were never measured on it.
03Why is there no Central Asia breakdown by country?
The policy file and the program file do not share a clean join key — school names differ across the two files — so per-country rates would be unreliable. We publish the 31-school totals only.
04When does the ranked edition ship?
The methodology preprint states the roadmap: a ranked edition follows once teaching and infrastructure dimensions are measurable at the same scale as policy. This dashboard ships the segment data first, so the ranked edition starts from published numbers, not new claims.
From dashboards to evidence
Every figure above traces to an open dataset: the Global University AI Policy Audit 2026, the US R1 AI Policy Audit 2026, the Central Asia AI Education Audit 2026, Online Degree Tuition 2026, Academic Database Recognition 2026, and the RAG Hallucination Benchmark. The full synthesis is the State of AI in Higher Education 2026, the per-university verdicts are in the university directory, and the short answers are in Key Findings.