A good example of AI trickery

I asked Claude AI to check out this map, which when posted on Facebook was queried by two members of the Preston History Facebook group. Its verdict: it’s a fake, probably created by AI.

A fake, AI-generated map of Preston

Here’s the full report from Claude (it took less than a minute to generate):


The short version: the mill is real, but the map almost certainly isnโ€™t a genuine 1893 Ordnance Survey sheet. It looks like a modern or AI-generated โ€œvintage-styleโ€ image built around real names.

What checks out

Spa Mill was a real Preston mill run by Levi Fish. In the 1891 Worrallโ€™s directory, Levi Fish is listed at Spa Mill, Marsh Lane, with 530 looms making satteens and fancy dress goods. The mill survived well into the 20th century. Historic Englandโ€™s aerial collection has a 1930 photograph showing the Spa Cotton Mill, the English Electric Works and the Albert Edward Dock, and another showing the English Electric Works, United Electric Car Works and the Spa Cotton Mill along West Strand. Marsh Lane, Kay Street, Leighton Street and the Lancaster Canal basin are all real streets and features in that part of Preston. I couldnโ€™t find any record linking โ€œRobert Gardner & Coโ€ to Spa Mill. Cotton Mills in Preston 1891 +2

Why the map itself looks fake

  • Grid coordinates. The margin shows โ€œ257000 Eโ€ and โ€œ436500 Nโ€. The National Grid wasnโ€™t introduced until the late 1930s, so no 1893 sheet would carry these. They are also wrong for Preston: the 1930 aerial photos put Spa Mill at about 352573, 429604 (SD526296). rcahms
  • Location. Those same photos put Spa Mill down by Strand Road and the docks area, not hard up against the canal basin as drawn here. Thatโ€™s worth checking against a genuine sheet.
  • Cartographic style. Real OS 25-inch maps donโ€™t use generic labels like โ€œTerraced Housesโ€, โ€œWarehousesโ€, โ€œMeadowโ€ or โ€œNavigable to Garstang & Lancasterโ€. Kay Street also appears twice, running in two different directions.
  • Other details. The building shapes are vague and smudgy rather than precisely surveyed. The red outline and underlined caption are more like a modern annotation than an OS original. The sheet number (โ€œLXXIX.12โ€), the price and the magnetic variation figure also look doubtful to me for Preston, though I havenโ€™t verified those.

How to check it

The National Library of Scotlandโ€™s map site (maps.nls.uk) has the genuine OS 25-inch first and second editions for Lancashire, and the Preston 1:500 town plans of the 1890s, free to view. Laid side by side, the differences should be obvious.

If this has turned up in your Facebook group, it might be worth a gentle note that itโ€™s a modern illustration rather than an original OS sheet, while the mill and Levi Fish are genuinely part of Prestonโ€™s history.

Sources:

  • Cotton Mills in Preston 1891 โ€“ Graceโ€™s Guide
  • EPW034081 โ€“ Spa Cotton Mill, English Electric Works and Albert Edward Dock, 1930 โ€“ Britain from Above
  • EPW034079 โ€“ English Electric Works and Spa Cotton Mill along West Strand, 1930

There are several other mistakes, including, most glaringly, that the orientation of the map is wrong, having the canal run east/west, rather than north/south.


Discover more from preston history

Subscribe to get the latest posts sent to your email.

One thought on “A good example of AI trickery”

  1. A.I. reports false reports yet again. I’ve had similar experiences as regards Google’s A.I. client. I asked about my history of false reports. Doesn’t exist A.I told me. “In order to protect my Privacy”

    Yet as we go forward these falsehoods are only going to go so often unreported.

Leave a Reply