AI Transparency Statement
EFFECTIVE 3 AUGUST 2026
AI is used in building and running TrueTotal, and pretending otherwise would be exactly the kind of quiet omission this site exists to oppose. Here is precisely where it is used, and the hard line it is never allowed to cross.
Where AI is used
- Drafting and editing prose. Guides, explanatory copy and pages like this one may be drafted or edited with AI assistance, then reviewed by a human before publication.
- Engineering. AI assists in writing and reviewing the code that runs this site. That code is tested and reviewed like any other.
- Structuring and summarising. Turning long source documents — fare rules, tax schedules, terms — into structured inputs a human then verifies against the original.
- The assistant. Our conversational assistant answers questions in natural language, strictly over TrueTotal’s own sourced figures.
Where AI is never used
- Generating a price, cost or specification. Never. If we do not have a sourced figure, the interface says so rather than producing one.
- Deciding a ranking. Rankings come from the published deterministic calculation in the Methodology — not from a model’s judgement, and not from anything that could be nudged.
- Writing reviews or ratings. We publish no AI-written reviews, testimonials or scores presented as human experience.
- Generating product imagery. No AI-generated or stock images stand in for a real item. See Sources & Data.
- Making decisions about you. No automated decision-making with legal or similarly significant effects, and no profiling — see the Privacy Policy.
How the assistant is constrained
The assistant is deliberately built to be less capable than a general chatbot. It answers from TrueTotal’s own dataset, cites the figures it used, and declines to answer rather than guessing when it has no data. If you ask it for a price we do not hold, the correct behaviour — and the behaviour we build for — is to say so.
It is not an adviser. Everything in No Professional Advice applies to it in full.
Human review
A human is accountable for anything published. AI-assisted drafts are reviewed before they go live; figures are checked against their sources, not against the model’s recollection; and methodology changes — which move every figure at once — get the most scrutiny of all. Responsibility for an error is ours regardless of what produced it. “The AI wrote it” is not an excuse we will ever offer you.
Limitations we know about
- Language models can be confidently wrong. This is why they are barred from originating figures — the constraint exists precisely because the failure mode is fluent and hard to spot.
- Summarisation can drop nuance. A condensed fare rule can lose the exception that mattered. Where a rule is critical we link the primary source so you can read it yourself.
- Training data has a cutoff and no awareness of today’s prices. Nothing time-sensitive comes from a model — it comes from a dated observation.
- Bias can enter through phrasing. Editorial review looks for it, and we would rather hear about it from you than not.
Your data and AI
We do not train models on your personal data, and we do not sell your data to anyone for any purpose including model training. Conversations with the assistant are processed to answer your question; retention and your rights are covered in the Privacy Policy.
Continuous improvement
This statement describes today. As we use these tools differently we will update it and change the effective date rather than letting the description drift out of line with the practice. Material changes are noted in the Trust Center.
Report a mistake
If the assistant said something wrong, or any AI-assisted content contains an error, please tell us — via Report an Error or corrections@truetotalcalculator.com. Include what you asked and what it answered. Mistakes of this kind are especially useful to us, because they usually reveal a missing constraint rather than a one-off slip.