Verified timeline
| Date | Event | Source | Certainty |
|---|---|---|---|
| From 1565 onwards | Port books document individual voyages, ports, ships, shipmasters, and cargoes in unbroken 200-year record | Southampton.ac.uk | Reported by outlets |
| Late 16th to mid-18th century | Project charts Britain's growth into an Empire during the reign of Elizabeth I to George III | Southampton.ac.uk | Reported by outlets |
| 1696–97 | European markets accounted for around 80 percent of recorded value of England's overseas trade | Southampton.ac.uk, Yahoo | Reported by outlets |
| 1600s to 1799 | Dutch East India Company captains recorded daily latitude and longitude during global voyages | IBM | Reported by outlets |
| Around 1790 | First apparent inflection point identified in sailing ship speeds | IBM | Reported by outlets |
| 1820s | Second apparent inflection point identified in sailing ship speeds | IBM | Reported by outlets |
Southampton.ac.uk and Yahoo differ on the total timespan: Southampton reports a 200-year unbroken record from 1565, while Yahoo reports 461 years of maritime history.
Why this matters
Southampton.ac.uk reports that historians have created an interactive map charting Britain's merchant shipping from 1565 onward, affecting scholars, citizens, and the general public who can now explore maritime trade data. Yahoo.com notes the project mapped 461 years of Britain's maritime past, revealing how the nation grew into an empire through European connections. According to Southampton.ac.uk, researchers discovered forgotten trading communities and ports, including the vanished port of Ravenspurn, illuminating a "forgotten maritime landscape."
The University of Southampton's findings shift understanding of Britain's economic priorities. Yahoo.com reports that even as the British Empire expanded, between 1696–97 European markets accounted for around 80 percent of England's recorded overseas trade value. Southampton.ac.uk found that of the top 15 countries by maritime connections to British trade, 10 were in Europe, establishing Europe's centrality rather than colonial trade as Britain's primary commercial engine.
IBM.com describes a related initiative where structured AI tools analyzed historical shipping data, identifying apparent inflection points in sailing speeds around 1790 and the 1820s, though this work was framed as exploratory rather than definitive research.
What is established, alleged and stated
Reported by the outlets
-
According to Southampton.ac.uk, the University of Southampton project used AI to process approximately 100,000 pages of handwritten port records from the National Archives in Kew, spanning from 1565 to the mid-18th century.
-
Yahoo.com reports that coal appeared most frequently in customs data, followed by malt, wheat, iron, and wool, with coal recorded in over 101,000 cargo entries.
-
Southampton.ac.uk states researchers identified over 3,000 individual ports at home and abroad, including over 1,000 British ports in a network around the country's coastal and inland waterways.
-
According to Southampton.ac.uk, in 1696–97, European markets accounted for approximately 80 percent of England's recorded overseas trade value.
-
IBM.com reports that Chris Hay, a Distinguished Engineer at IBM, used structured AI tools including Dutch East India Company logbooks to analyze historical maritime data and identified apparent sailing-speed inflection points around 1790 and in the 1820s.
Stated by officials or the organisation concerned
-
Professor Craig Lambert of the University of Southampton told Southampton.ac.uk that "without AI extracting and organising the data, this would have taken years" and that AI "sped up the process enormously."
-
Chris Hay stated to IBM.com that he did not present the experiment as peer-reviewed research or as a definitive answer to why ships got faster over time.
Alleged or disputed
Nothing in the sources.
What we still do not know
-
Yahoo News reports the dataset spans from the Elizabethan Era to just before the Industrial Revolution, while University of Southampton describes an unbroken 200-year record from 1565 onwards—the sources differ on whether these timespans are reconciled or overlap inconsistently.
-
The sources do not specify how extensively manual verification occurred after AI processing or what percentage of extracted data required human correction.
-
The sources do not clarify whether the University of Southampton project and IBM's Chris Hay's Marine Archive Server are the same initiative or separate efforts analyzing different historical maritime datasets.
-
IBM states Hay's findings on sailing-speed inflection points around 1790 and the 1820s were not "peer-reviewed research," leaving their preliminary or conclusive status uncertain.
-
The sources do not explain the handwriting-recognition system's accuracy rate or the precise mechanisms by which conflicting historical spelling variations were resolved without human intervention.
-
The sources do not disclose how many of the 3,000 identified ports were completely lost versus merely declined, or whether defunct ports are counted separately.
What the models agree on
- University of Southampton used AI to process approximately 100,000 pages of handwritten port records from the National Archives in Kew.
- Coal was the most frequently recorded commodity in customs data, appearing in over 101,000 cargo entries.
- In 1696–97, European markets accounted for approximately 80 percent of England's recorded overseas trade value.
Agreement between models does not validate an allegation. It only shows that they read the same reports in the same way.
Where they differ
Every difference is typed. An omission is not treated as an error: it is only called critical when leaving the point out would change how a reader understands the story.
| Divergence | Type | Why it matters | Models concerned |
|---|---|---|---|
| The project covered maritime trade records spanning from 1565 to the mid-18th century. | Critical omission | Establishes the historical scope and timeframe of the data extracted from port records. | Left out by Gemini |
| AI learned to recognize historical handwriting variations and changing spelling conventions to extract the data. | Critical omission | Explains a technical challenge the AI overcame to accurately transcribe historical port records. | Left out by GPT, Gemini, Perplexity and Grok |
| France ranked highest among trading partners with 157 identified maritime sites, followed by the Netherlands with 68 sites. | Neutral omission | Identifies the primary European trading partners based on documented maritime connections. | Left out by GPT, Gemini and Perplexity |
| Researchers identified over 3,000 individual ports at home and abroad, including over 1,000 British ports. | Neutral omission | Shows the extent and breadth of the port network documented in the extracted records. | Left out by GPT and Gemini |
| The South West of Britain dominated merchant shipping activity with 426 recorded maritime locations across the centuries. | Neutral omission | Indicates regional concentration of maritime trade activity within Britain during this period. | Left out by GPT, Gemini, Perplexity and Grok |
| The project revealed lost trading communities including the vanished port of Ravenspurn which sank into the North Sea. | Neutral omission | Demonstrates how the extracted data illuminates previously undocumented or forgotten historical ports. | Left out by Mistral, Perplexity and Grok |
| An interactive map was developed to allow the public to explore the maritime trade data. | Neutral omission | Shows how research findings were made accessible to a broader audience beyond academic researchers. | Left out by Claude and Mistral |
| IBM's Chris Hay used structured tools and historical logbooks including Dutch East India Company records to investigate sailing speeds. | Neutral omission | Identifies a separate but related research contribution analyzing maritime technology changes over time. | Left out by Claude, GPT, Mistral and Grok |
| Researchers identified apparent sailing-speed inflection points around 1790 and in the 1820s. | Neutral omission | Suggests technological advances in maritime transportation at specific historical moments. | Left out by Claude, GPT, Gemini and Grok |
Model reliability scorecard
| Model | Coverage | Critical omissions | Unverified claims | Contradictions | Editorial reliability |
|---|---|---|---|---|---|
| Claude | 75% | 0 | 0 | 0 | Solid |
| GPT | 50% | 1 | 0 | 0 | Mixed |
| Gemini | 50% | 2 | 0 | 0 | Mixed |
| Mistral | 75% | 0 | 1 | 0 | Solid |
| Perplexity | 67% | 1 | 0 | 0 | Solid |
| Grok | 58% | 1 | 0 | 0 | Mixed |
How the rating works. Coverage is the share of the panel's points that a model includes. Each model was asked for 6 to 10 claims, so coverage reflects what it chose to include within that budget. Penalty points: 1 per critical omission, 2 per unverified claim, 2 per contradiction, 1 if coverage is below 60%. Strong: coverage of 80% or more and no penalty. Solid: coverage of 65% or more and at most 2 penalty points. Mixed: at most 4 penalty points. Weak: more than 4. There is no numerical grade.
The most important finding
The main result is not that the models disagree. None of them contradicted another. They differ mostly by omission: 2 critical and 7 neutral across 12 points. The more important signal is reliability: Mistral added 1 claim that the cited reports do not support, and those matter more than any omission.
Reading between the lines
What this suggests is that Yahoo reports the University of Southampton project spans "461 years" while Southampton.ac.uk describes an "unbroken 200-year record" from 1565 onwards. If both refer to the same dataset, one source may be counting differently—perhaps including preliminary or fragmentary records beyond the core consecutive span. This discrepancy matters for assessing how complete the archive truly is.
What this does not establish is whether the AI system achieved full accuracy without human intervention. Southampton.ac.uk states the process "still needed manual checking," but neither source quantifies how much correction was required or what error rates remained after verification.
The key signal appears in IBM's account: Chris Hay explicitly frames his sailing-speed findings as exploratory analysis, not peer-reviewed research. This suggests structured AI tools can compress historical investigation, but independent scholarly validation would be needed before treating the speed inflection points as established historical fact.
What to watch
-
Will other archives adopt this AI methodology to digitize historical records at scale? According to southampton.ac.uk, the Southampton project represents a task "not seen before," suggesting potential replicability across heritage institutions.
-
Can researchers explain the sailing-speed acceleration pattern identified in late 18th and early 19th century records? IBM notes that Chris Hay's analysis identified inflection points around 1790 and the 1820s but explicitly states the findings were not peer-reviewed research.
-
How will public engagement with the interactive map reshape understanding of Britain's maritime past? southampton.ac.uk notes the map allows users to explore data "for different periods, trading hubs, ships, goods, routes and more," indicating potential for discovery.
-
Will the Southampton database and IBM's Marine Archive Server integrate or remain separate initiatives? Both southampton.ac.uk and IBM describe historical maritime data projects, but their relationship remains unclear.
-
Will formal peer-reviewed research validate the Southampton project's dataset and methodology? southampton.ac.uk describes the project's completion but does not indicate publication in academic journals.
Planned review of this page: October 29, 2026.
Method, honestly stated
- Every model received the same reports (southampton.ac.uk, yahoo.com, ibm.com) and the same prompt, and was told to use only those reports.
- Models that ran: Claude, GPT, Gemini, Mistral, Perplexity and Grok. No model was left out.
- Each claim a model made was checked against the reports. A claim counts as verified if its supporting passage appears word for word in a report, or if a separate model judged it a supported paraphrase. Otherwise it is marked unverified.
- A separate model merged equivalent claims into comparable points and classified each difference as contradiction, critical omission, neutral omission, different emphasis or unverified addition. That step is automated and can be wrong.
- The similarity score (87/100) measures how alike the analyses read. It is not an accuracy or confidence score, not agreement on facts and not a measure of truth.
- An automated pipeline wrote and published this page. No person read it before publication.
Verdict
Coverage of the central fact is broadly consistent: 3 of 12 points appear in every analysis. The differences are mostly about completeness and framing, not contradiction. The main risk identified is added claims: Mistral included material the cited reports do not support.
