How we test, score and fund this library
Every listing on Nexoralist is judged against the same scorecard, by the same process, whether the tool is a one-person project or a funded company. This page explains that process in full — including what we are paid for and what we are not.
What we cover
We cover AI tools that a cross-border trading, sourcing or marketing team could plausibly use in a working week: writing product listings, translating supplier mail, generating imagery, parsing freight documents, answering customer questions, analysing supplier data. That scope is deliberately narrow. A tool that is interesting but has no path to export work does not get covered, however popular it is.
The library currently holds 257 tools across 14 categories. We add tools when teams ask about them or when a release changes what is possible; we remove tools when they shut down, drop below the quality bar, or become actively misleading in their marketing.
The scorecard
Each tool is scored from 0 to 5 against four criteria. The overall editor score is a weighted judgement, not a strict average — a tool that is excellent on output but unusable for a five-person team will not score highly, however good the output is.
- Output quality. We run the same brief through each tool: a product description, a supplier email, a short video script, a spreadsheet of order data. We judge the raw output before any prompt-tuning, then again after a realistic amount of effort.
- Export readiness. Multilingual output, unit and currency handling, tolerance for Incoterms and HS codes, and whether the tone survives translation into a customer-facing message.
- Data handling. What the tool does with what you paste into it: retention, training on customer data, region of processing, and whether the terms are readable enough to answer a customer's question about them.
- Price-to-value. What the free tier actually allows, where the first real paywall sits, and whether the paid tier is priced for a small team or for an enterprise procurement process.
How a listing gets made
A tool enters the library in four steps: it is flagged (by a release, a reader email, or our own monitoring), the official site and documentation are read in full, the working brief is run against it, and the result is written up with the criteria above. The write-up records the version we tested and the date of the check, both of which appear on the tool's page.
We do not publish a listing for a tool we have not opened and used. Where a tool is gated behind a sales call and could not be tested, we say so on the page rather than write around the gap.
How often we re-check
Pricing changes, models get swapped, features get deprecated. Listings are revisited on a rolling cycle, prioritised by how much a change would affect the score — pricing and model changes get checked first. Each tool page shows the date of its last check, so you can judge how current it is rather than taking our word for it.
What we are not paid for
Rankings and editor scores are never for sale. No tool can buy a position, a score, a "pick" label, or a place in a category. We do not use affiliate links in tool listings: the "Visit" button on a tool page goes straight to the vendor with no tracking parameter of ours attached.
We do sell one thing: a single sponsored placement on the homepage, labelled as a sponsor, sold by the month. It is visually separated from editorial listings, it carries no score, and it does not influence what gets reviewed or how anything is scored. If that ever changes, this page changes first.
How we use automation
We use automated tooling to monitor releases — model launches on Hugging Face, vendor announcement feeds — and to capture the interface screenshots that appear on tool pages. That monitoring tells us what to look at. It does not write the reviews.
Review text is written and edited by a person who has used the tool. Where a page is largely compiled from vendor documentation rather than hands-on testing, the page says so. We would rather publish a thinner page that is accurate than a longer one that is padded.
Corrections
If a listing is wrong — a price has changed, a feature was removed, a tool has shut down — tell us and we will fix it. Corrections are made to the listing itself and the check date is updated; we do not silently rewrite a page and pretend the earlier version never existed.
Who runs this
Nexoralist is an independent publication. It is not owned by, funded by, or affiliated with any of the tools it covers. The person editing the library has spent years on the commercial side of cross-border trade, which is where the testing briefs come from.
For corrections, coverage requests or questions about this policy, write to the contact address on this site. We answer.