Jamaica's All Media Survey Now Runs on AI: What to Ask

What changes for media buyers

Market Research Services Ltd launched what it calls The All-New All Media Survey on Thursday 1 October 2026. According to the Gleaner, the study now covers social media influencers and content creators alongside radio, television, newspapers and internet use. It adds segments for banked and unbanked Jamaicans, prospective vehicle and home buyers, business owners and farm owners.

The collection window grows from a six-week snapshot to six months. MRSL chief executive Kishka-Kaye Anderson said the point is to capture how people move between platforms and screens across the day. Subscribers to MRSL Classic and The All Media Insight Suite get an interim look at trends after the first three months.

MRSL has partnered with DatanXzen, a US-based technology company with a strong technical presence in India. Dr Kiran Kumar Koka, Group CEO of ASOFT Group of Companies, told the launch the tools automate the preparation and analysis of research data and flag duplicate records, mismatches and missing information. He described the technology as support for human analysts, who would retain responsibility for the work.

The launch report leaves several buying decisions unresolved: the full-results date, subscription prices and the method for measuring and ranking influencers. Ask for those details before committing a budget.

What the Jamaican AI-use evidence shows

AI-use research gives some context for the survey, though it cannot tell us how much of a particular audience's media is machine-made. In July 2026 the Jamaica Observer reported research by StarApple Analytics, in collaboration with Anthropic, on how Jamaicans use generative AI. Content creation and copywriting made up 25.2 per cent of AI conversations in Jamaica, against 22.7 per cent globally. The Observer did not report the sample size or method, so treat the percentages as shares of conversations rather than shares of people.

Anthropic's own index gives the global picture. Its January 2026 report puts arts, design and entertainment tasks at 11 per cent of Claude.ai conversations, against 6 per cent on Anthropic's first-party API, and finds that a 1 per cent rise in GDP per capita goes with a 0.7 per cent rise in Claude usage per capita at country level. Our inference, not Anthropic's: Jamaica's above-average content-creation share suggests creative work is where many people start.

Formal training remains limited. The University of the West Indies' SALISES institute surveyed 1,072 Jamaicans in all 14 parishes between October and December 2025. Six per cent have been formally trained in AI, 62 per cent want training and 81 per cent back moderate or strict oversight. The Observer gives Jamaica an awareness score of 76 out of 100, but reports that much of that awareness is filtered through "science fiction style machines", and quotes research fellow Dr Stephen Johnson on the gap between awareness and understanding.

Most Jamaicans are online, although national usage figures do not establish whether access is reliable or affordable for every audience segment. The UNESCO Jamaica AI Readiness Assessment, launched by the Office of the Prime Minister on 1 April 2026, puts internet penetration at 85 per cent, with more than 196 people from government, academia, business, civil society and youth groups taking part. The World Bank's series is higher: 90.1 per cent of Jamaicans used the internet in 2024, up from 70.8 per cent in 2019. The two figures come from different sources and different years, and on either one most of the audience is online. The UNESCO report also lists the Data Protection Act and the Cybercrimes Act as the existing base for digital governance.

At the Caribbean Broadcasting Union's 57th General Assembly in Guyana on 19 August, themed Caribbean Media and Artificial Intelligence, Guyana's information minister Kwame McCoy urged broadcasters to shape AI rather than only adopt tools made elsewhere. The Observer's account lists data ownership, algorithmic control and the economic value of Caribbean content among his concerns.

The method: Extracted Intelligence, applied to audience data

Adrian Dunkley's Extracted Intelligence describes data about a community's behaviour, language and decisions that leaves as raw material and returns, if at all, as a product the community must buy. The question it asks is who benefits from the intelligence the data produces, which is wider than whether anyone consented.

Apply it to the survey. The raw material is how Jamaicans listen, watch, scroll and buy over six months. The product is an online analytics platform that subscribers filter and explore. That is a legitimate business, and MRSL is a Jamaican firm. The launch report names a US-based partner with a technical presence in India but does not identify where the data is processed or stored. Media buyers should ask MRSL for those details.

A test we suggest, adapted from the method: list what the survey collects and what Jamaican buyers receive in return. If you cannot fill in the second list, you cannot price what you are paying for.

Checking audience claims in Jamaica

Take a Kingston distributor who splits a modest radio and social budget by gut feel and last year's invoice. The new survey promises to rank channels for her segment, perhaps business owners, with six months of data. Her risk is spending against a ranking she cannot audit. Her best defence is her own sales ledger, which already shows which promotion brought customers in.

A Montego Bay hotel marketer has a different question: whether an influencer is worth the fee. The survey adds influencers and content creators to its scope, but how they will be measured is the open question. Until that is answered, the audience numbers for creators are the ones to double-check.

For a community radio station in a rural parish, the concern is how smaller audiences are represented. If AI cleaning flags records as duplicates or mismatches, small-sample areas are where a wrong flag costs most, because there are fewer records to absorb the error. That is our inference from how data cleaning works, and we have not seen MRSL's method note.

The StarApple Analytics and Anthropic work breaks down Jamaican AI use by task. Dunkley told the Observer that Jamaicans appear to use only about a fifth of what the models can do, that people ask AI for business plans but rarely for the harder strategic questions such as competitor analysis and market research, and that users in Trinidad and Tobago build websites and apps more readily. He called for AI literacy in high schools and warned about shadow AI, meaning daily use stuck on low-level tasks. Those are his reported views, and the Observer is the source.

StarApple AI, which describes itself as the Caribbean's first AI company, is based in Kingston and led by Adrian Dunkley, whom it describes as the Caribbean's regional leader in AI. Its Jamaican training work is aimed at the literacy gap SALISES measured. Disclosure: AI Jamaica is supported by StarApple AI, so read that description as the house view. The Observer reports Dunkley as founder of StarApple Analytics and co-founder of Maestro AI Labs.

A Jamaican analyst still needs enough access to trace a result back to its source and explain how the model treated it.

What to do this week: a 14-day evidence test

1. Send five questions in writing to any research vendor, MRSL included, by 23 October 2026: where is the data processed and stored, how are influencers counted and ranked, how many records did the AI flag and who reviewed them, what is the margin of error for my segment, and what do I receive if I stop subscribing.
2. Write your own baseline. List every channel you paid for in the last 90 days, the spend in Jamaican dollars and the number of sales or enquiries you can trace to each. If you cannot trace a channel, mark it unknown.
3. Test it: on 12 October to 26 October, give one channel a unique promotion code or phone number and compare enquiries per $10,000 spent with your baseline. Trust a survey ranking only if the higher-ranked channel also delivers within 20 per cent of that cost per enquiry. This is a test we suggest, and the 20 per cent line is ours.
4. Check your analysts, not only your data. Take 50 rows of your own customer records, plant 5 duplicates, and ask your analyst or an AI tool to find them. Count how many it catches and log every false flag. A tool that misses more than 1 of the 5, or flags more than 2 good rows, should not clean data you pay to rely on.

Where this can go wrong

The first harm is money. An advertiser who shifts a year's budget on a ranking built from data nobody outside the vendor audited carries the loss, and the smallest advertisers feel it most because they have no second source. The guardrail is the written question list above plus your own sales check. The second harm is to small stations and publishers. If the AI treats thin local data as noise, a channel can look smaller than it is, and its income follows. The guardrail is asking for segment-level margins of error. The third is a flood of synthetic content. Jamaicans exceed the global average in using AI for content creation, and the CBU assembly raised disclosure of synthetic content as a standard still to be set. Until it exists, ask whether any creator you pay discloses AI use.

The weakest part of this article is its source base. We have the Gleaner's report of the launch, not MRSL's method note, and the StarApple Analytics figures come to us through the Observer without a stated sample size. A reader with the primary documents should trust those over our summary.

About the author

Lancelot Williams is a Caribbean AI Analyst covering AI adoption, research and policy across the English-speaking Caribbean. Every figure above was checked against the linked source on 12 October 2026.