Every chart on AI Chartist begins with a source we can name, a figure we can verify, and a visual format that presents the data honestly. This page explains exactly how we work.
Why this page exists
Data visualisation can mislead as easily as it can clarify. A truncated Y-axis overstates a trend. A cherry-picked date range flatters a narrative. A single survey presented as consensus misrepresents a genuinely contested picture. The methodology page exists to make our working visible — so that readers, researchers, journalists, and anyone considering embedding our charts can assess the quality of what we produce before deciding whether to trust it.
We hold ourselves to a simple standard: if we could not defend every number in a chart to a specialist in that field, we do not publish it.
Data sources
We draw primarily on published research and data from the following categories of source, listed in descending order of our preference.
Tier 1 — Institutional primary research. Peer-reviewed academic papers, reports from major international institutions (IMF, World Economic Forum, World Bank, OECD, ILO), and primary data published by official statistical bodies (Office for National Statistics, Bureau of Labour Statistics, Eurostat, and equivalents). These sources carry the highest methodological standards and are our default starting point for any chart.
Tier 2 — Tier-one consulting and financial research. Published reports from organisations including McKinsey Global Institute, Goldman Sachs Global Investment Research, PwC, Deloitte, and comparable firms that employ rigorous research methodologies, publish their sample sizes and methods, and stand behind their findings publicly. We treat these as reliable but note where a funder relationship may create an incentive to present findings in a particular light.
Tier 3 — Technology company research. Published findings from major technology companies including Anthropic, OpenAI, Google DeepMind, and Microsoft. We use this research where it represents genuine methodological work — Anthropic’s Economic Index and OpenAI’s GPT exposure studies, for example, are substantive academic contributions — but we note the provenance and cross-reference where possible.
Tier 4 — Industry data aggregators. Platforms including Similarweb, Statista, Sensor Tower, and comparable data providers are used for traffic, usage, and market size figures where no primary institutional source exists. We treat these as indicative rather than audited and label them accordingly.
We do not use anonymous blog posts, unattributed press releases, or single-source claims that have not been independently corroborated. Where a compelling statistic circulates widely online but cannot be traced to a verifiable primary source, we do not use it.
How we handle conflicting data
In fast-moving fields like AI, conflicting figures are the norm rather than the exception. Different studies use different methodologies, different sample populations, different definitions of key terms, and different measurement periods. We handle this in one of three ways, depending on the nature and degree of the conflict.
Where studies disagree on magnitude but agree on direction — for example, different estimates of how many jobs may be displaced by AI by 2030 — we use mid-point estimates and note the range. The chart will state the figure used and the source; the methodology note on the chart page will describe the range across sources and explain our choice.
Where studies disagree on direction — a genuine empirical dispute where credible researchers reach opposite conclusions — we do not present one side as settled fact. Instead, we either build a chart that explicitly maps the disagreement, or we note the contested nature of the finding prominently within the chart and its supporting copy.
Where a single high-quality primary source exists and others are clearly derivative of it, we use the primary source and note that subsequent figures in wider circulation are drawn from it.
We do not resolve conflicts by choosing the most dramatic figure, the most recent figure, or the figure that produces the most visually striking chart. We choose the figure that best represents the current state of the evidence.
Synthesised data and estimates
Some charts on AIChartist.UK synthesise findings across multiple research sources rather than plotting a single dataset. The AI Exposure by Salary Band chart, for example, draws on six independent research papers to construct income-band estimates that no single study provides in the exact form displayed.
Where we synthesise data in this way, we do the following without exception. We list every source used, either within the chart or on the chart’s page. We label the output clearly as a synthesised estimate rather than a single-source figure. We describe our synthesis methodology in the chart page’s methodology note. And we are explicit about the confidence level — whether the figures represent a strong cross-study consensus, a balance of evidence, or a directional indication based on limited data.
We do not present synthesised estimates as audited figures, and we do not present directional patterns as precise measurements.
Labelling conventions
Every chart uses consistent labelling to indicate the reliability of the figures displayed.
No qualifier means the figure is drawn directly from a named primary or tier-one secondary source and is presented as published.
“Estimated” means the figure is our synthesis or interpolation across multiple sources, or is drawn from a source that itself uses modelled rather than directly measured data.
“Indicative” means the figure is directionally supported by the evidence but should not be treated as a precise measurement.
“Publishing soon” or “data pending” on upcoming charts means we have identified the research gap and are building the chart but have not yet completed our source verification to the standard required for publication.
Where a chart uses data from multiple years and the most recent available year varies by data point, we note this within the chart or its supporting copy.
Chart update policy
AIChartist.UK publishes charts that are designed to remain relevant over time, not to capture a single news moment. Where the underlying data changes materially — new annual figures, a substantially revised study, a shift in the evidence base — we update the chart and note the revision date.
The publication date displayed on each chart reflects the most recent data revision, not the original publication date. When a chart is updated, we note both dates: the original publication date and the most recent revision.
We do not update charts to chase minor fluctuations in regularly updated data series. A chart showing AI market size by year, for example, is updated annually when the relevant institutional reports are published — not monthly when interim estimates appear.
We maintain a record of significant data revisions. If a correction materially changes the story the chart tells, we note that explicitly on the chart page.
What we do not do
We do not accept payment to alter, suppress, favour, or add data points to any chart. Commercial relationships do not influence editorial content.
We do not publish a statistic because it is striking. A figure that cannot be sourced, verified, and contextualised to our standard does not appear on this site, regardless of how widely it has been cited elsewhere.
We do not use misleading visual techniques: Y-axes on our bar charts begin at zero unless there is a specific and disclosed reason to do otherwise. We do not use dual axes to imply correlations that the data does not support. We do not cherry-pick date ranges to flatten or exaggerate trends.
We do not present AI-generated data as source data. AI tools are used in our production process for drafting, research assistance, and code generation. They are not used as data sources. Every figure that appears in a chart can be traced to a named human-authored source.
Corrections policy
We take accuracy seriously and we make mistakes. If you identify an error in a chart — a wrong figure, a misattributed source, a misleading label, or a calculation error — please contact us via the corrections form on the contact page.
We review all correction requests. Where an error is confirmed, we correct it promptly, note the correction on the chart page, and update the revision date. We do not quietly remove incorrect figures; we note what changed and why. Where a submitted correction turns out not to be an error — because the figure is correct and the source reliable — we respond to explain our reasoning.
We distinguish between errors and disputes. A factual error is a figure that is wrong by reference to its stated source. A dispute is a disagreement about which source or methodology is most appropriate. We welcome both, and we handle them differently.
Embed and attribution standards
All charts on AIChartist.UK are free to embed under the following conditions. The embedded version must include the visible credit line that appears below each chart, including the AIChartist.UK attribution and source citations. The credit line must not be obscured, removed, or modified. The chart must not be presented as the original work of the embedding party.
When a chart is embedded, the data within it remains our responsibility. If we update a chart following a correction or data revision, the embedded version updates automatically. If you embed a chart and subsequently disagree with a data revision we have made, you may contact us to discuss it, but we do not maintain separate versioned embeds for individual publishers.
Commercial licensing — including white-label use, removal of attribution, or custom-branded versions of our charts — is available on request via the contact page.
A note on AI in our production process
In the interest of full transparency: AIChartist.UK uses AI tools, including Claude, in the production of charts, supporting copy, and this website. Specifically, AI is used to assist with code generation for interactive chart builds, first-draft editorial copy that is then edited and fact-checked by a human, and research synthesis to identify relevant sources across a large body of literature.
AI is not used to generate data, invent figures, or substitute for source verification. Every number that appears on this site has been verified against a named primary or secondary source by a human editor before publication. The use of AI in our production process does not diminish the accuracy standard we hold our published output to — it is a production tool, not an editorial one.
Contact
Questions about our methodology, source requests, or correction submissions can be sent via the contact page. We read everything and respond to substantive queries.