TaxOrch versus a general-purpose AI assistant
A general assistant will answer any tax question confidently. This one computes with a calculator, cites a real section, and refuses when it cannot.
An approach, not a product.
No product is named here, because a statement about a named competitor needs a public source rendered on the page — and without one it is hearsay published as fact, which also ages into a lie the moment they ship a release.
Where each one actually differs.
Where a number comes from
- A general-purpose AI assistant
- Generated by the model. Fluent, frequently right, and impossible to distinguish from a fluent wrong one.
- TaxOrch
- Calculated exactly from the official rates for the tax year, with the working shown.
Every figure is worked out exactly from official rates, never estimated.
What a legal statement rests on
- A general-purpose AI assistant
- Whatever the model has absorbed, sometimes with a citation that is topical rather than supporting.
- TaxOrch
- A specific passage in official sources.
Every legal claim cites a specific section of a real official document.
Whether the law is current
- A general-purpose AI assistant
- Often out of date, and confident either way. It rarely knows which year it is answering for.
- TaxOrch
- Answers use the law and rates in force for the date in question, and every answer carries its law-as-of date.
A 2023 question is answered with 2023 law.
What happens outside its knowledge
- A general-purpose AI assistant
- An answer, usually. Refusal is a prompt instruction, and a user's message is also input.
- TaxOrch
- A clear statement when a country or topic is not covered yet.
An answer ships only when it is verifiably certain — otherwise it asks a specific question or escalates to a human, with a stated reason.
Whether an old answer can be checked
- A general-purpose AI assistant
- No. The same prompt on a later model gives a different answer, and there is nothing recording which one you got.
- TaxOrch
- Every answer records its sources, so it can be checked again later.
Every answer records the sources it used, so it can be checked and reproduced later.
Breadth
- A general-purpose AI assistant
- Everything. It will attempt any question in any jurisdiction.
- TaxOrch
- Declared coverage, published per country including the gaps. Narrower on purpose.
Where a general-purpose ai assistant wins: Genuinely wider. If your question is outside declared coverage, a general assistant will answer it and this will not.Exact inside declared coverage.
A general-purpose AI assistant suits you if
- Drafting, summarising and rephrasing, where a plausible answer is a good answer
- Exploring a topic before you know what you are looking for
- Work where being wrong is cheap and obvious
TaxOrch suits you if
- Figures that go into a return, a set of books or an invoice
- Positions that a reviewer or an enquiry will examine
- Anywhere a confident wrong answer costs more than no answer
Do not buy this if any of these is true.
A comparison page with no such section is an advertisement, and a reader discounts everything else on it accordingly. This one is required by the data model, so it cannot be quietly dropped when somebody is editing for length.
- Your questions are mostly outside declared coverage — check the coverage page before anything else
- You want a draft rather than a defensible figure, in which case a general assistant is faster and cheaper
- You need an answer for a jurisdiction with no calculator, where this will abstain and a general assistant will not
Everything here is meant to be verifiable.
The coverage matrix, the evaluation results and the answer contract are all published, including what they do not cover. Read them before you talk to anybody.
Bring the question you expect it to get wrong.
Decision support, not tax advice. A professional reviews before anything is filed.