The Caribbean Examinations Council put preliminary CSEC and CAPE results online at noon today, and for most families the number that matters is a single grade on a screen. The number that should worry education ministries is a different one: Jamaica's national CSEC mathematics pass rate sat at 44 percent last year, while individual schools ranged from Mile Gully High's 9 percent to Old Harbour High's 26 percent, a gap CXC's exam calendar cannot explain and a new AI cheating crackdown will not close.
Six weeks before today's results, CXC announced its biggest overhaul of School-Based Assessments in years, phasing out take-home projects in non-practical subjects because generative AI was making it impossible to trust that submitted work reflected a student's own thinking. That reform addresses a real problem. It also happens to be the easy half of the AI story. The harder half, the one that actually decides whether a teenager in Mile Gully finishes school with a usable mathematics qualification, is still sitting unbuilt.
The Number Behind Today's Headlines
Results day is the same ritual every August. Students refresh a login page, principals brace for the school-by-school breakdown, and newspapers run the national averages by mid-morning. This year, ahead of the release, the Gleaner spoke with principals across parishes who were still measuring their chances against last year's figures, because 2025's numbers were the clearest picture available of where the system actually stood.
That picture was uneven in ways a single national average hides. St Andrew Technical High School sent 220 students into the May-June sitting. Old Harbour High, with roughly 590 students across CSEC and CAPE combined, posted a mathematics pass rate of 26 percent, well under the 44 percent national figure. Mile Gully High, where about 70 students sat mathematics, posted 9 percent. St Elizabeth Technical High, which lost close to a month of instruction to Hurricane Melissa, still landed near the national average at roughly 43 percent.
Principal Lynton Weir put a number on the pattern behind the numbers: newly trained mathematics teachers arrive, perform well for a year or two, and then migrate for better pay abroad. He recalled a year, 2016, when his own school hit 69 percent in mathematics, a figure well out of reach of what most schools are managing now. Christopher Tyme pointed to declining student engagement and attendance. Keith Wellington was blunter about the subject itself: mathematics simply requires more practice time than most other CSEC subjects, and schools losing teachers mid-cycle cannot deliver that practice consistently.
A Region-Wide Pattern, Not a Jamaica Problem
Jamaica gets the most detailed school-level coverage because its press reports it closely, but the underlying numbers are regional. CXC's own results showed a 39 percent CSEC mathematics pass rate across the 16 countries and territories it serves in 2025, up from 36 percent in 2024. English A came in at 76 percent regionally in 2024. Jamaica's 44 percent in mathematics sits above that regional average, which is not a comforting fact. It means the same spread CXC results revealed in Jamaica, from single-digit school pass rates to figures well above the mean, is close to certain in Trinidad and Tobago, Guyana, Barbados, and the smaller territories sitting on the same syllabus, the same exam calendar, and the same teacher migration pressure.
CXC is headquartered in Bridgetown, Barbados, and sets the same mathematics curriculum for a student in Kingston, a student in Georgetown, and a student on a small island where the entire secondary school might have one qualified mathematics teacher for every grade level. A national average flatters that last student's chances. A school-by-school breakdown, the kind Jamaican newspapers already publish and other territories mostly do not, is what actually shows where intervention needs to land.
The Core Argument
CXC cannot rewrite the mathematics syllabus every time a school loses its teacher. It can build the tools that let one remaining teacher, or no teacher at all for a stretch of term, keep every student in a class moving through the specific skill they are missing. That tool does not exist at any scale across the region yet, and the gap between a 9 percent school and a 44 percent average is the clearest evidence of why it should.
The Other AI Story CXC Just Told
On 31 July 2026, CXC confirmed that traditional take-home SBAs in non-practical subjects are being phased out. In their place comes Paper 32, a supervised written assessment where a research topic is released roughly a month before the exam and students bring approved reference notes into a controlled setting rather than submitting coursework built at home over weeks. Schools can choose between the old SBA format and Paper 32 from the 2027 sitting. Paper 32 becomes the default for CAPE non-practical subjects from 2027 onward, and supervised assessment for private candidates in the practical sciences becomes mandatory by 2028. CXC also added increased moderation, random school visits, and interviews with students, teachers, and principals at assessment centres.
The trigger was straightforward. CXC's own statement cited growing concern that AI-generated content and plagiarism were undermining confidence that a submitted SBA reflected the candidate who signed it. That concern predates generative AI in one sense, SBAs have long been vulnerable to being copied, sold, or effectively written by a teacher trying to protect a school's average, but large language models turned a manageable integrity problem into one CXC judged needed a structural fix rather than another memo.
For this year's cycle, the interim rule already applies. Any candidate who used an AI tool while producing an SBA had to complete a Disclosure Form describing where and how, backed by an Originality Report supporting the claim that the final work is genuinely theirs. CXC Director of Operations Dr Nicole Manning has said human judgement stays central to how those disclosures are reviewed, and work found to be wholly or substantially AI-generated is treated as academic dishonesty. No AI use, no form required.
Why Banning AI Only Solves Half the Problem
Paper 32 and the disclosure regime are reasonable responses to a genuine integrity risk, and CXC deserves credit for moving on it within a single admissions cycle rather than letting the problem fester for years the way some of the mathematics pass-rate coverage suggests the underlying teaching gap has. But the reform is defensive by design. It exists to stop something from happening, students submitting work that is not theirs, rather than to make something better happen, a student in a school with one overstretched mathematics teacher actually closing a skill gap before results day arrives.
Those are different problems with different fixes. A 9 percent pass rate at Mile Gully High is not primarily an AI cheating story. It is a story about teacher migration, practice time, and a syllabus that assumes consistent instruction a small rural school cannot always deliver. CXC's SBA overhaul, useful as it is, does nothing for that student. The tools that would are AI tools too, just aimed at learning rather than policing.
What AI-Powered Learning Support Could Actually Do
None of what follows is speculative technology. Every category below runs somewhere already, in school systems facing versions of the same teacher shortage and attainment gap the CXC results expose every August.
Adaptive Practice at the Point of Need
Software that diagnoses exactly which mathematics skill a student is missing, rather than cycling the whole class through a fixed syllabus pace, has produced measurable gains in developing-country school systems that face teacher shortages similar to Jamaica's. A student who has mastered fractions but not yet ratios gets ratio practice, not a repeat of a fractions lesson they already understand. That is the kind of differentiation Keith Wellington's point about practice time demands, and it is exactly what one teacher covering three or four classes cannot deliver by hand.
Predicting Failure Before It Happens
A model trained on attendance patterns, in-course assessment scores, and prior-term performance can flag a student on track to fail CSEC mathematics in October of their final year, not in August when the result is already fixed. Jamaica's schools already generate most of that data through internal assessment records. What is missing is a system that reads it continuously and surfaces the students slipping, rather than leaving that judgement to a teacher's memory across 30 or more students per class.
Filling the Gap Teacher Migration Leaves
Lynton Weir's account of teachers leaving within a year or two of arriving is not a problem AI solves outright. It is a problem AI can soften. An AI teaching assistant that generates differentiated practice sets, marks routine exercises, and runs supplementary drilling sessions gives a single remaining mathematics teacher the reach of a small department, buying schools time while recruitment and retention policy tries to catch up with the salaries pulling teachers abroad.
Making the Grade Review Process Fairer and Faster
The 2026 grade review window opens on 1 September and runs to 15 September, at XCD 85 per subject, a real cost for a family already stretched by school fees. Analytics that flag scripts with marking patterns statistically inconsistent with a candidate's overall performance could help CXC target reviews at genuine anomalies rather than leaving every disputed grade to a flat fee and a slow queue, cutting both the cost and the wait for students whose results genuinely warrant a second look.
What Should Happen Next
Three moves would turn today's results into more than a headline that repeats next August.
First, CXC and Jamaica's Ministry of Education should treat the school-by-school breakdown already sitting in this year's results as a targeting list, not a shame list. Mile Gully High and schools like it are already identified. What is missing is funded, adaptive tutoring support aimed specifically at those schools rather than another islandwide curriculum announcement that arrives years after the students it was meant to help have already left the system.
Second, the moderation infrastructure CXC just built to catch AI misuse in SBAs, the audits, the school visits, the closer data tracking, should be repurposed to catch struggling students early in the school year rather than only dishonest submissions at the end of it. Much of that infrastructure is data collection. Data collection aimed at learning gaps is at least as useful as data collection aimed at cheating.
Third, governments across the 16 territories CXC serves should treat AI-assisted teaching support as a genuine, if partial, answer to teacher migration rather than a side conversation from the SBA policy debate. A teacher who leaves cannot be replaced by software. A teacher who stays but is covering three classes alone can be given tools that make that coverage survivable for the students in the room.
Two Stories, One Missing Half
CXC moved fast on the AI story that threatened exam integrity, and the region is better for it. The AI story that could actually move a mathematics pass rate from 9 percent to something closer to 44, or 44 percent to something closer to Lynton Weir's 69 from 2016, has no equivalent policy, no equivalent funding announcement, and no equivalent urgency behind it yet.
Results day will come again next August. Whether the school-by-school breakdown looks any different depends less on how tightly SBAs are policed and more on whether anyone builds the tools that help a student in Mile Gully practise the one skill they are missing, on the day they are missing it, rather than finding out on results day that they were.
Frequently Asked Questions
When were the 2026 CSEC and CAPE results released?
CXC released preliminary results for the May-June 2026 CSEC, CAPE, CCSLC and CTEC examinations on Tuesday, 18 August 2026, from 12 noon Atlantic Standard Time, through the Student Results Portal at studentportal.cxc.org. Candidates across the 16 countries and territories CXC serves logged in with their centre number, candidate number and birth certificate details. The grade review window runs from 1 to 15 September 2026, at XCD 85 per subject.
What was Jamaica's CSEC mathematics pass rate in 2025, and why does it matter for today's results?
Jamaica's national CSEC mathematics pass rate reached 44 percent in 2025, up from 39 percent in 2024. English A rose to 85 percent, a nine-point jump. Only 19.2 percent of Jamaican candidates achieved five or more subjects including maths and English, itself only a 1.2-point improvement year on year. That baseline is what principals across the island were measuring today's 2026 results against.
Why do some Jamaican schools have maths pass rates as low as 9% while the national average is 44%?
Old Harbour High School posted a 26 percent mathematics pass rate in 2025 and Mile Gully High School posted 9 percent, both well below the national average, while St Elizabeth Technical High School sat closer to the average despite losing roughly a month of instruction to Hurricane Melissa. Principals and teachers point to mathematics teachers migrating within a year or two of qualifying, uneven student attendance, and the sheer volume of practice time mathematics demands compared with other subjects.
What changes has CXC made to School-Based Assessments because of AI?
CXC announced on 31 July 2026 that it is phasing out take-home SBAs in non-practical subjects, replacing them with Paper 32, a supervised written assessment where research topics are released a month ahead and students bring approved reference notes into a controlled exam setting. Schools can choose between the old format and Paper 32 from 2027, Paper 32 becomes standard for CAPE non-practical subjects from 2027, and supervised assessment for private practical-science candidates becomes mandatory by 2028.
Does CXC now allow students to use AI in their SBA work?
For the 2026 cycle, candidates who use AI tools in an SBA must complete a Disclosure Form describing where and how, backed by an Originality Report supporting that the final submission reflects their own thinking. CXC Director of Operations Dr Nicole Manning has said human judgement remains central to reviewing those disclosures, and work found to be wholly or substantially AI-generated is treated as academic dishonesty. No AI use means no form is required.
How could AI actually help raise CSEC maths pass rates, rather than just policing cheating?
Adaptive tutoring software that identifies the exact skill a student is missing, rather than re-teaching a whole syllabus, has produced measurable gains in other developing-country school systems and could run cheaply across Jamaica's high schools. Paired with early-warning analytics built on attendance and in-course assessment data, schools could flag a student at risk of failing mathematics in October rather than discovering it in August. Neither tool requires banning AI. Both require funding and building, which nobody in the region has done at scale.
Is the maths pass rate crisis unique to Jamaica or a wider Caribbean problem?
It is regional. CXC's own figures show a 39 percent CSEC mathematics pass rate across its 16 territories in 2025, up from 36 percent in 2024, alongside a 76 percent regional pass rate in English A in 2024. Jamaica's 44 percent sits above that regional average, which means the school-level spread visible in Jamaica, from single digits to well above average, is very likely mirrored across Trinidad and Tobago, Guyana, Barbados, and smaller territories sharing the same syllabus and the same teacher migration pressure.
What should Caribbean governments and CXC do next?
Three things. Fund adaptive maths tutoring pilots in the specific low-performing schools this year's own results already identify, rather than waiting for another islandwide curriculum review. Extend the data infrastructure built to detect AI misuse in SBAs so it also flags struggling students earlier in the school year, not only dishonest submissions at the end of it. And treat AI-assisted teaching support as a partial answer to teacher migration, giving one mathematics teacher covering three classes the tools to reach every student rather than teaching to the middle of the room.