New editorial guide, published 17 October 2026. The historical sources linked below are preserved separately and are not represented as current product or service information.
A forecast often loses its conditional language as it moves through a news chain. A report projects a future market size; a summary says the market will reach that size; a later post treats the number as if it has already been observed. The digits can remain unchanged throughout that transformation. What changes is the claim. A historical item in this domain’s catalogue discusses an AI-market projection associated with a 2025 UNCTAD report. This guide does not update the forecast or assess investment opportunities. It explains how to read and describe a projection without quietly turning it into a fact about a later outcome.
Identify the kind of future statement
A projection, target, commitment and scheduled event are different things. A projection estimates a possible future quantity using a method and assumptions. A target states an intended result. A commitment records what an actor says it will do, while a schedule identifies a planned time. None automatically establishes completion. Begin by finding the source’s own description. Do not replace projected with expected merely because the second word sounds smoother if the change obscures the method or confidence. The aim is not to make every future sentence timid, but to preserve the category of statement actually supported.
UNCTAD’s April 2025 public summary describes a projection of the global AI market reaching $4.8 trillion by 2033. That is a dated projection, not a measured 2033 result. A careful reference keeps the issuing organisation, source date and target year together. The number alone cannot tell a reader whether the estimate includes the same products, currency basis or scope as another forecast. Those definitions need inspection before comparison. This guide does not infer that the projected path is already being realised or that the forecast remains unchanged in every later publication.
Write down the starting point and destination
Use a simple forecast card: issuing date, baseline period, baseline value, target period, projected value, unit, coverage and method location. If a field is unavailable, mark it unavailable. A headline may contain only the target value while the source explains a narrower market definition. An old forecast can also use a price basis that differs from a later current-price figure. Do not assume that two dollar amounts describe the same economic object merely because they share a currency symbol. The card should expose what must be checked rather than hide it behind a polished chart.
In an invented example, a research group estimates that a defined service market could rise from ten units to twenty over five years. A second group estimates thirty units but includes an additional product category. Their numbers do not yet constitute a disagreement about the same forecast. First align the coverage and dates, or report that they differ. Avoid averaging the two estimates into twenty-five as if the midpoint were a new consensus. That calculation would be your own construction and would require an explanation of why it is meaningful.
Find the assumptions that carry the result
A useful forecast description identifies the assumptions most important to its interpretation. They might concern adoption, prices, capacity or the treatment of new product categories. Read the method rather than inventing a generic list and attributing it to the source. If the public summary does not expose enough method, say that the summary alone is insufficient for an assessment. That is a finding about the available evidence, not proof that the forecast is worthless. It tells the reader what additional material would be needed before making a stronger judgement.
Keep scenarios separate. A low, central and high case may reflect different assumptions rather than a probability distribution with a single implied confidence level. Do not call a central case guaranteed or assume a wide range means every point inside it is equally likely. Where the source provides probabilities, preserve what those probabilities refer to. Where it does not, do not add them. An editor can explain that outcomes depend on assumptions without claiming to have quantified uncertainty that the report itself leaves qualitative or does not discuss in its public summary.
Check the arithmetic without confusing it with validation
Arithmetic can reveal a reporting error even when it cannot establish the forecast’s plausibility. In the invented example, moving from ten to twenty is a doubling, or a one-hundred-percent increase. It is not a two-hundred-percent increase. If a growth rate is reported, check whether it is annual, cumulative or compounded over several periods. Keep the calculation’s inputs visible. A correct calculation confirms a relationship among stated numbers; it does not prove that the forecast’s assumptions are realistic or that the underlying market definition captures the reader’s intended question.
Be careful with comparisons to national economies, company values or spending totals. They can involve different measures and time periods even when every figure is denominated in money. A colourful analogy may be harder to justify than the original forecast statement. If the comparison is not needed to explain the reader’s task, leave it out. If it is included, define both quantities and their dates. Do not treat rhetorical scale as evidence of an event’s inevitability. A large projected number can be important to discuss without being certain to occur.
Track revisions as versions, not embarrassments
A later forecast can change because new observations arrived, definitions changed or assumptions were revised. Preserve the earlier version and describe the reason given for the new one. Do not silently replace a number in an old article and leave its original date suggesting the new estimate was known earlier. Conversely, a revised forecast does not automatically prove dishonesty in the original. The appropriate question is whether the method, assumptions and revisions are accurately documented. Separate a source’s explanation of its revision from your own interpretation of why the estimate changed.
To evaluate a past forecast, compare it with an observed measure that actually matches its scope and target period. If the definitions diverge, explain the limitation rather than declaring a precise forecasting error. The target year may not yet have arrived. In that case, intermediate data can be relevant but cannot establish the final outcome. Avoid grading a long-horizon projection solely by whether one early period followed a smooth path. A meaningful retrospective needs a stated comparison rule, not a judgement selected after seeing which numbers make the forecast look best or worst.
Keep the forecast label wherever the number travels
The headline, summary, chart caption and social description should all retain enough context to identify the figure as projected. A careful body paragraph does not repair a headline that reports a future estimate as an observed fact. If space is tight, reduce decorative language before removing the target year or projection label. Link to the source and identify which version you inspected. A short sentence can remain direct: a named report projected a defined quantity for a named year. It need not become either promotional certainty or an unreadable pile of qualifications.
For the historical AI story, the catalogue records what the saved page discussed and keeps its metadata separate from the capture date. This new guide adds a reading method; it does not supply a new market forecast. Readers can apply the method to other future claims by asking what kind of statement it is, which assumptions support it and what observation would eventually test it. The important distinction survives every edit: a forecast is evidence of an estimate made at a time, not evidence that the estimated future has already happened.
Primary references
Related historical source records
- Global AI race dominated by 100 firms says UN report · captured 9 May 2025