Does GPS damage your brain?

No study has shown that GPS damages your brain. The research everybody reaches for, the London taxi drivers, ran in the opposite direction: it measured what four years of learning the Knowledge does to a hippocampus. One small study has followed the same people while their GPS use rose, and its authors ask readers not to draw strong conclusions from it. The dementia claim attached to all of this is a forecast made in 2012 by a named person who has never presented it as a finding.

Four years in, and the scan changes#

Eleanor Maguire's group published the cross-sectional study in 2000: sixteen London taxi drivers, fifty controls, and posterior hippocampi that were larger in the drivers, with volume tracking the years spent driving. That design cannot separate the job building the brain from a particular brain choosing the job.

The 2011 study closes that gap and is the one that matters. Katherine Woollett and Maguire scanned 79 trainee taxi drivers and 31 controls before and after four years. Some trainees qualified and some did not. In the authors' account, structural change appeared only in those who qualified: no change in the trainees who failed, none in the controls. Same starting cohort, same curriculum, and the grey matter follows the completion.

Two details get lost when the study is retold. The gain was accompanied by costs elsewhere in the memory profile, so what the Knowledge produced was a trade rather than a bonus. And nobody rescanned a qualified driver after Transport for London's examiners had gone and the satnav had arrived. The literature has a beginning and no ending.

The prediction, in the words it was actually made in#

The claim in circulation is that GPS use will produce a measurable rise in dementia. It is usually attached to Vivienne Ming. She has published it herself, and the published version is milder than most restatements of it. On Socos Academy, dated 17 February 2022, dating the forecast to about ten years earlier, she writes:

My prediction: in 20 or so years, we will see a small but substantial increase in dementia. The cause: GPS-based AI navigation systems, like Google Maps, Apple Maps, and Waze.

Versions that have her forecasting a statistically meaningful increase in early-onset dementia are carrying weight she did not put there. This research had one of those versions in its own provenance notes until 5 September 2026, taken from a secondary account rather than from her post, and it has been corrected.

Ming's own framing is careful. She sets out her reasoning, links the studies she is reasoning from, and offers a piece of advice rather than a result: take a different route to work. A forecast presented as a forecast is not the problem here. The problem is the citation chain that arrives three steps later, in which a prediction from 2012 and a set of scans from 2011 have merged into a finding that neither of them contains.

Thirteen people, three years, and no scanner#

One study has looked at the removal side. Louisa Dahmani and Véronique Bohbot tested 50 regular drivers in Montreal, then brought 13 of them back an average of 3.2 years later. Greater lifetime GPS experience was associated with lower use of hippocampus-dependent spatial strategies, poorer map drawing and fewer landmarks noticed. In the follow-up, hours of GPS use since the first session tracked a steeper decline on the spatial memory measure.

Three things about that study should travel with it wherever the finding does. It measured behaviour on two virtual mazes and never put anyone in a scanner, so every statement about the hippocampus is an inference from tasks validated elsewhere. The follow-up was unplanned, which is how a sample of 50 becomes a sample of 13. And the authors write, in the paper itself, that they "caution against any strong conclusions as spurious correlations are possible".

Their argument for the causal reading deserves a hearing. Heavy GPS users in this sample did not report a poorer sense of direction, so the obvious reverse explanation, that people with weak spatial memory reach for the satnav, has no support in the data. That is real evidence. It is 13 people.

The Christmas paper, and the age its drivers died#

In December 2024 the BMJ published, in its Christmas issue, a study of death certificates across 443 occupations in the United States. Taxi drivers and ambulance drivers had the lowest proportion of deaths attributed to Alzheimer's disease of every occupation examined, and the pattern did not appear among bus drivers or pilots, who follow predetermined routes. The authors are direct about what it is: "We view these findings not as conclusive, but as hypothesis generating."

The load-bearing caveat came from Tara Spires-Jones, reviewing the paper for the Science Media Centre. The taxi and ambulance drivers in the sample died at around 64 to 67, against 74 for every other occupation, and Alzheimer's onset is typically after 65. Some of them may simply not have lived long enough. She also notes the sex imbalance, 10 to 22 per cent women among the drivers against 48 per cent elsewhere, on a disease women are more likely to develop. Robert Howard, reviewing it alongside her, drew the practical conclusion: it would be premature to suggest that drivers should turn off their satnavs to prevent dementia.

Acquisition is measured, removal is assumed#

Navigation is the cleanest case in the whole literature on skills and delegation. It has a four-year longitudinal imaging study, a defined curriculum, an examiner and a pass mark, and it still cannot answer the question people want answered.

The asymmetry is the finding. Across skill after skill, the evidence that practice builds capability is strong, replicated and old. The evidence for what happens when the practice is handed to a machine is thin, recent and mostly behavioural. Organisations reason as though the two were symmetrical, and take the strength of the first as licence for the second. Capability debt accumulates in that gap.

Hold one measured exception beside all of this. Across four Polish centres, 19 endoscopists averaging 27.6 years of experience detected adenomas in 28.4 per cent of their unassisted colonoscopies before AI arrived in the department, and 22.4 per cent of their unassisted colonoscopies afterwards. Same doctors, same procedure without the tool, six percentage points apart. That is removal, measured, in professionals, in a clinical setting. One study, one procedure, one country. It remains the closest thing the field has to the experiment nobody has run on the taxi drivers.

Where this research has used the material, and one figure it will not repeat#

This argument has been made in Box of Amazing before, in "The Case for Being Bad at Things" on 18 January 2026, which puts Woollett and Maguire beside the 2020 Scientific Reports study and draws the same contrast between acquisition and loss. Two things in that essay are tightened here.

It gives the Knowledge as 25,000 streets. That figure circulates widely and is not Transport for London's. TfL describes thousands of streets and landmarks within six miles of Charing Cross, and specifies the 320 Blue Book runs. The verifiable number is 320.

It also reports the GPS finding as heavy users declining faster "than those who navigated without assistance". Dahmani and Bohbot had no such comparison group. Their longitudinal result is a correlation within one sample of 13 people between hours of GPS use and change in score. The direction of the finding survives; the comparison does not.

How to argue about this without overclaiming#

If you are making the case for keeping a human capability, the taxi studies are the wrong citation and reaching for them weakens you. They show that a hard thing, done for years, changes the brain of the person doing it. Nobody disputes that. Cite them for what they establish, which is that difficulty is where the capability comes from, and stop there.

If you are hearing the case, the question to ask is which direction the study ran. Acquisition studies are common because they are easy to fund and quick to publish. Removal studies require someone to take a working tool away from a professional and measure what happens, which is expensive, slow and often unethical. The absence of removal evidence tells you about research design rather than about safety.

And for the individual question underneath all this, the answer that survives the evidence is small: use the map sometimes, learn a route occasionally, notice a landmark. Ming's own advice, take a different way to work, is the part of her post that rests on the firmest ground. Nobody has shown it prevents anything. It costs nothing and it keeps the practice alive, which is the only mechanism anyone has actually measured.

Related SuperSkills research#

On the mechanism, cognitive offloading and the Google effect. On whether it comes back, can you regain a skill you have lost and how fast skills decay. On the same shape in figures rather than studies, the most-quoted AI statistics, checked. On what the practice builds, deliberate practice.

Key sources

• Maguire, E. A. et al. (2000). Navigation-related structural change in the hippocampi of taxi drivers. PNAS, 97(8).
• Woollett, K. and Maguire, E. A. (2011). Acquiring 'the Knowledge' of London's layout drives structural brain changes. Current Biology, 21(24).
• Dahmani, L. and Bohbot, V. D. (2020). Habitual use of GPS negatively impacts spatial memory during self-guided navigation. Scientific Reports, 10, 6310.
• Patel, V. R. et al. (2024). Alzheimer's disease mortality among taxi and ambulance drivers. The BMJ, 387, Christmas issue.
• Budzyń, K. et al. (2025). Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy. The Lancet Gastroenterology and Hepatology.
• Transport for London. Learn the Knowledge of London. Read at source 5 September 2026.
• Ming, V. L. (2022). Can GPS cause dementia? Socos Academy, 17 February 2022. Read at source 5 September 2026.

About this definition#

The Knowledge of London is Transport for London's examination and is not a SuperSkills coinage. The reading offered here, that the taxi literature measures acquisition and is routinely cited for removal, is an interpretation by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), and is marked as such rather than as a finding.

About this research#

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.