Checked 9 October 2026
There’s no single figure. Published estimates of the purple pound, the spending power of disabled people and their households, run from £212bn to £446bn a year. On 9 October 2026, both AI engines that answered me gave £274bn, and neither said it was a 2020 estimate.
The figures, with their dates
| Figure | Who published it | Year | Source |
|---|---|---|---|
| £80bn | DWP | released 2004 | GOV.UK, 27 Aug 2014 |
| £212bn | DWP | 2012/13 | GOV.UK, 27 Aug 2014 |
| £249bn | DWP | 2014/15 | DWP note, 2014/15 |
| £274bn | Purple | 2020 | Purple infographic |
| £446bn | Credited by Purple Tuesday to University of Bristol and Money Advice Trust research | Not stated anywhere I could find | Nationwide, 19 Mar 2026 |
These figures aren’t one number growing over time. DWP’s figures add up household income after housing costs, and DWP says its £80bn figure was calculated per person, so it can’t be compared with the later ones. I couldn’t find the method behind £446bn.
The £446bn footnote
Nationwide’s press release footnotes £446bn to a House of Commons Library briefing, CBP-9602. I read the whole PDF (17 November 2025). The House of Commons Library briefing that Nationwide footnotes for £446bn doesn’t contain that figure, or any purple pound total.
The briefing has the inputs: 16.8 million disabled people in 2023/24, 38% of people living in a family with a disabled member, and a median household income of £506 a week after housing costs. It doesn’t add them up into a total. Purple Tuesday, as reported by the BBC, credits £446bn to University of Bristol and Money Advice Trust research. I couldn’t find that report.
What the engines said
Same prompt, pasted as is: How much is the purple pound worth in the UK? Run from the UK on 9 October 2026, between 11:00 and 11:30, twice per engine.
| Engine | Session | Figure | Year given? | Source named |
|---|---|---|---|---|
| ChatGPT | Signed in, temporary chat (unpersonalised) | £274bn, both runs | No | Clearsight Access |
| Gemini (Flash-Lite) | Signed in | £274bn, both runs | No | None shown |
| Google AI Mode | Signed in | No answer in two attempts | – | – |
| Perplexity | Logged out | Asked me to sign up | – | – |
| Claude | – | Not run: my browser tool can’t operate claude.ai | – | – |
Both AI engines that answered gave £274bn, and neither said it was a 2020 figure. ChatGPT did say the figure was dated, and on the second run disputed. It still didn’t give the year, which is the one detail that would let a reader judge it.
What the purple pound can unlock
The money only reaches a business if a disabled customer can tell, before they set off, whether they can use it. Research shared with Nationwide by the Research Institute for Disabled Consumers found that 82% of disabled people named missing access details as the biggest barrier to judging accessibility. Euan’s Guide’s 2025 survey found that 70% return to venues that get access right and 66% recommend them.
This next part is my reasoning, not something I measured. When someone asks an AI engine “is this place step-free?”, the engine can only repeat what someone has published. A page that states its step-free entrance, hearing loop and accessible toilet, with the date it was checked, gives the engine something to cite. A page that says nothing leaves the answer to guesswork.
Two things to do:
- If you quote a purple pound figure, give its year and who made it. “£274bn (Purple, 2020)” can be checked. “£274bn” can’t.
- Publish your access information as text, with the date you checked it. In March 2026, Nationwide committed to publishing access guides for all 605 of its branches.
What I will not claim
- Which figure is right. They measure different things in different years.
- That every user gets the same answer. These are two engines, signed in, on one morning.
- That publishing access information gets a business cited by AI engines. I haven’t measured that.
- Anything about traffic or sales.
Dated call
- Claim: A year from now, AI engines will still quote the purple pound without the year it refers to.
- Called: 9 October 2026.
- Check date: 9 October 2027.
- Test: The same prompt, the same five engines, fresh sessions, run from the UK on one day, twice each.
- Right if: at least half of the engines that answer give a figure with no year attached.
- Wrong if: fewer than half do.
- Baseline: 2 of 2 answering engines on 9 October 2026.
It goes on the predictions page when the ledger goes live. If it’s wrong, it stays up in red.
Data
purple-pound-ai-answers-2026-10-09.csv. It has one row per engine run, including the run that failed and one ChatGPT attempt I excluded because the prompt was sent garbled. Related: AI answers quote prices with no date.
Added 10 October 2026: this call is P10 on the scoreboard. It is marked hit, miss or void there, using the test above unchanged. This post called those outcomes right and wrong.