How this works — and where it falls short

We don't ask you to trust us. We show our working, and every claim links to the source it came from — so you can check it yourself in one tap.

What this is

Show the Working looks at what each UK party has actually said it will do, and asks a simple question for each thing that affects your life: would this policy make it better, worse, or is it genuinely unclear?

We organise that around a deliberately balanced set of topics — not just public services and the environment, but the things every side of politics cares about: the cost of living, your take-home pay and tax, whether the public finances add up, jobs and living standards, crime and safety, your rights, your personal freedom, inequality, and more. The set is built so that whatever a party is trying to achieve, it can register as a benefit somewhere — not only as a cost to someone else's priority. You pick which topics matter to you; we never add them into a single score.

How we judge a policy

For every policy, on every issue:

We start from the party's own words (their manifesto and on-the-record positions).
We gather research from outside sources — official data, independent think tanks, academics — on what the policy would really do.
The verdict is built from individual claims, and every claim quotes the exact source line behind it. Tap "show source" on any claim to see it.
A separate, independent check runs over the verdict to catch overstatements, mis-cited sources, and false balance, before it's published.

The promise: if a claim can't be tied to a real source we gave the system, it doesn't get shown. A made-up figure or a mis-named source can't reach you — that's enforced automatically, not by trust.

What we deliberately DON'T do

No overall party score. We never add it all up into "Party X = 7/10". Which issues matter most is your call, not ours — so we show each issue separately and let you weigh them.
We judge the policy, not the party. Each verdict is made without telling the system who proposed it — the same policy gets the same verdict whoever tables it.
We don't let the choice of topics favour a side. The topics are picked so that every major political tradition's goals can register as a benefit, not only as a cost to someone else's — a lower tax burden, sound public finances and personal freedom sit alongside public services, the environment and equality. We actually found and corrected a tilt towards one side in an earlier version of the topic set.
Some things are your call, not ours. A few issues — immigration and border control above all — have no "better" or "worse" that everyone agrees on; whether tighter or looser is good depends on your values. For those we give no verdict — we show which way a policy moves things and let you decide. (Its knock-on effects, like on wages, housing or the public finances, are still judged under those topics.)
We don't claim to be neutral. We make editorial choices (like which topics to organise around). We tell you what they are so you can disagree with them.

Where it falls short — honestly

The interpretation is ours. The facts each link to a source, but the conclusion we draw from them ("Our reading") is a judgement you might disagree with. That's why we put the evidence above our reading — so you can see the facts and draw your own.

Some things we couldn't verify show as gaps. If we couldn't tie a policy to checkable sources, we say "not yet assessed" rather than guess. A gap is honest — it's not us hiding the hard ones.

It's a snapshot, dated. Parties announce new policies all the time. Each verdict shows when it was generated; newer announcements may not be in yet.

It's AI-generated and checked, not perfect. The analysis is produced by AI with automatic source-checking and an independent review step. That catches fabricated and mis-cited claims, but it can't guarantee every judgement is right. If you spot something wrong, tell us — there's a "report an error" button on every verdict, and corrections are logged in public.

Check us, challenge us

Every verdict opens to its claims and sources. The analysis pipeline and the prompts are public. If a verdict is wrong, report it — that's how it gets fixed, in the open.

← Back to the comparison